]> git.saurik.com Git - redis.git/blob - src/redis.c
TODO updated with 2.6 things
[redis.git] / 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 actally 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 struct redisCommand redisCommandTable[] = {
73 {"get",getCommand,2,0,NULL,1,1,1,0,0},
74 {"set",setCommand,3,REDIS_CMD_DENYOOM,noPreloadGetKeys,1,1,1,0,0},
75 {"setnx",setnxCommand,3,REDIS_CMD_DENYOOM,noPreloadGetKeys,1,1,1,0,0},
76 {"setex",setexCommand,4,REDIS_CMD_DENYOOM,noPreloadGetKeys,2,2,1,0,0},
77 {"append",appendCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
78 {"strlen",strlenCommand,2,0,NULL,1,1,1,0,0},
79 {"del",delCommand,-2,0,noPreloadGetKeys,1,-1,1,0,0},
80 {"exists",existsCommand,2,0,NULL,1,1,1,0,0},
81 {"setbit",setbitCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
82 {"getbit",getbitCommand,3,0,NULL,1,1,1,0,0},
83 {"setrange",setrangeCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
84 {"getrange",getrangeCommand,4,0,NULL,1,1,1,0,0},
85 {"substr",getrangeCommand,4,0,NULL,1,1,1,0,0},
86 {"incr",incrCommand,2,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
87 {"decr",decrCommand,2,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
88 {"mget",mgetCommand,-2,0,NULL,1,-1,1,0,0},
89 {"rpush",rpushCommand,-3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
90 {"lpush",lpushCommand,-3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
91 {"rpushx",rpushxCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
92 {"lpushx",lpushxCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
93 {"linsert",linsertCommand,5,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
94 {"rpop",rpopCommand,2,0,NULL,1,1,1,0,0},
95 {"lpop",lpopCommand,2,0,NULL,1,1,1,0,0},
96 {"brpop",brpopCommand,-3,0,NULL,1,1,1,0,0},
97 {"brpoplpush",brpoplpushCommand,4,REDIS_CMD_DENYOOM,NULL,1,2,1,0,0},
98 {"blpop",blpopCommand,-3,0,NULL,1,-2,1,0,0},
99 {"llen",llenCommand,2,0,NULL,1,1,1,0,0},
100 {"lindex",lindexCommand,3,0,NULL,1,1,1,0,0},
101 {"lset",lsetCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
102 {"lrange",lrangeCommand,4,0,NULL,1,1,1,0,0},
103 {"ltrim",ltrimCommand,4,0,NULL,1,1,1,0,0},
104 {"lrem",lremCommand,4,0,NULL,1,1,1,0,0},
105 {"rpoplpush",rpoplpushCommand,3,REDIS_CMD_DENYOOM,NULL,1,2,1,0,0},
106 {"sadd",saddCommand,-3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
107 {"srem",sremCommand,-3,0,NULL,1,1,1,0,0},
108 {"smove",smoveCommand,4,0,NULL,1,2,1,0,0},
109 {"sismember",sismemberCommand,3,0,NULL,1,1,1,0,0},
110 {"scard",scardCommand,2,0,NULL,1,1,1,0,0},
111 {"spop",spopCommand,2,0,NULL,1,1,1,0,0},
112 {"srandmember",srandmemberCommand,2,0,NULL,1,1,1,0,0},
113 {"sinter",sinterCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1,0,0},
114 {"sinterstore",sinterstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1,0,0},
115 {"sunion",sunionCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1,0,0},
116 {"sunionstore",sunionstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1,0,0},
117 {"sdiff",sdiffCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1,0,0},
118 {"sdiffstore",sdiffstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1,0,0},
119 {"smembers",sinterCommand,2,0,NULL,1,1,1,0,0},
120 {"zadd",zaddCommand,-4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
121 {"zincrby",zincrbyCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
122 {"zrem",zremCommand,-3,0,NULL,1,1,1,0,0},
123 {"zremrangebyscore",zremrangebyscoreCommand,4,0,NULL,1,1,1,0,0},
124 {"zremrangebyrank",zremrangebyrankCommand,4,0,NULL,1,1,1,0,0},
125 {"zunionstore",zunionstoreCommand,-4,REDIS_CMD_DENYOOM,zunionInterGetKeys,0,0,0,0,0},
126 {"zinterstore",zinterstoreCommand,-4,REDIS_CMD_DENYOOM,zunionInterGetKeys,0,0,0,0,0},
127 {"zrange",zrangeCommand,-4,0,NULL,1,1,1,0,0},
128 {"zrangebyscore",zrangebyscoreCommand,-4,0,NULL,1,1,1,0,0},
129 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,0,NULL,1,1,1,0,0},
130 {"zcount",zcountCommand,4,0,NULL,1,1,1,0,0},
131 {"zrevrange",zrevrangeCommand,-4,0,NULL,1,1,1,0,0},
132 {"zcard",zcardCommand,2,0,NULL,1,1,1,0,0},
133 {"zscore",zscoreCommand,3,0,NULL,1,1,1,0,0},
134 {"zrank",zrankCommand,3,0,NULL,1,1,1,0,0},
135 {"zrevrank",zrevrankCommand,3,0,NULL,1,1,1,0,0},
136 {"hset",hsetCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
137 {"hsetnx",hsetnxCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
138 {"hget",hgetCommand,3,0,NULL,1,1,1,0,0},
139 {"hmset",hmsetCommand,-4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
140 {"hmget",hmgetCommand,-3,0,NULL,1,1,1,0,0},
141 {"hincrby",hincrbyCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
142 {"hdel",hdelCommand,-3,0,NULL,1,1,1,0,0},
143 {"hlen",hlenCommand,2,0,NULL,1,1,1,0,0},
144 {"hkeys",hkeysCommand,2,0,NULL,1,1,1,0,0},
145 {"hvals",hvalsCommand,2,0,NULL,1,1,1,0,0},
146 {"hgetall",hgetallCommand,2,0,NULL,1,1,1,0,0},
147 {"hexists",hexistsCommand,3,0,NULL,1,1,1,0,0},
148 {"incrby",incrbyCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
149 {"decrby",decrbyCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
150 {"getset",getsetCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
151 {"mset",msetCommand,-3,REDIS_CMD_DENYOOM,NULL,1,-1,2,0,0},
152 {"msetnx",msetnxCommand,-3,REDIS_CMD_DENYOOM,NULL,1,-1,2,0,0},
153 {"randomkey",randomkeyCommand,1,0,NULL,0,0,0,0,0},
154 {"select",selectCommand,2,0,NULL,0,0,0,0,0},
155 {"move",moveCommand,3,0,NULL,1,1,1,0,0},
156 {"rename",renameCommand,3,0,renameGetKeys,1,2,1,0,0},
157 {"renamenx",renamenxCommand,3,0,renameGetKeys,1,2,1,0,0},
158 {"expire",expireCommand,3,0,NULL,1,1,1,0,0},
159 {"expireat",expireatCommand,3,0,NULL,1,1,1,0,0},
160 {"keys",keysCommand,2,0,NULL,0,0,0,0,0},
161 {"dbsize",dbsizeCommand,1,0,NULL,0,0,0,0,0},
162 {"auth",authCommand,2,0,NULL,0,0,0,0,0},
163 {"ping",pingCommand,1,0,NULL,0,0,0,0,0},
164 {"echo",echoCommand,2,0,NULL,0,0,0,0,0},
165 {"save",saveCommand,1,0,NULL,0,0,0,0,0},
166 {"bgsave",bgsaveCommand,1,0,NULL,0,0,0,0,0},
167 {"bgrewriteaof",bgrewriteaofCommand,1,0,NULL,0,0,0,0,0},
168 {"shutdown",shutdownCommand,1,0,NULL,0,0,0,0,0},
169 {"lastsave",lastsaveCommand,1,0,NULL,0,0,0,0,0},
170 {"type",typeCommand,2,0,NULL,1,1,1,0,0},
171 {"multi",multiCommand,1,0,NULL,0,0,0,0,0},
172 {"exec",execCommand,1,REDIS_CMD_DENYOOM,NULL,0,0,0,0,0},
173 {"discard",discardCommand,1,0,NULL,0,0,0,0,0},
174 {"sync",syncCommand,1,0,NULL,0,0,0,0,0},
175 {"flushdb",flushdbCommand,1,0,NULL,0,0,0,0,0},
176 {"flushall",flushallCommand,1,0,NULL,0,0,0,0,0},
177 {"sort",sortCommand,-2,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0},
178 {"info",infoCommand,-1,0,NULL,0,0,0,0,0},
179 {"monitor",monitorCommand,1,0,NULL,0,0,0,0,0},
180 {"ttl",ttlCommand,2,0,NULL,1,1,1,0,0},
181 {"persist",persistCommand,2,0,NULL,1,1,1,0,0},
182 {"slaveof",slaveofCommand,3,0,NULL,0,0,0,0,0},
183 {"debug",debugCommand,-2,0,NULL,0,0,0,0,0},
184 {"config",configCommand,-2,0,NULL,0,0,0,0,0},
185 {"subscribe",subscribeCommand,-2,0,NULL,0,0,0,0,0},
186 {"unsubscribe",unsubscribeCommand,-1,0,NULL,0,0,0,0,0},
187 {"psubscribe",psubscribeCommand,-2,0,NULL,0,0,0,0,0},
188 {"punsubscribe",punsubscribeCommand,-1,0,NULL,0,0,0,0,0},
189 {"publish",publishCommand,3,REDIS_CMD_FORCE_REPLICATION,NULL,0,0,0,0,0},
190 {"watch",watchCommand,-2,0,noPreloadGetKeys,1,-1,1,0,0},
191 {"unwatch",unwatchCommand,1,0,NULL,0,0,0,0,0},
192 {"cluster",clusterCommand,-2,0,NULL,0,0,0,0,0},
193 {"restore",restoreCommand,4,0,NULL,0,0,0,0,0},
194 {"migrate",migrateCommand,6,0,NULL,0,0,0,0,0},
195 {"dump",dumpCommand,2,0,NULL,0,0,0,0,0},
196 {"object",objectCommand,-2,0,NULL,0,0,0,0,0},
197 {"client",clientCommand,-2,0,NULL,0,0,0,0,0},
198 {"eval",evalCommand,-3,REDIS_CMD_DENYOOM,zunionInterGetKeys,0,0,0,0,0},
199 {"evalsha",evalShaCommand,-3,REDIS_CMD_DENYOOM,zunionInterGetKeys,0,0,0,0,0},
200 {"slowlog",slowlogCommand,-2,0,NULL,0,0,0,0,0}
201 };
202
203 /*============================ Utility functions ============================ */
204
205 /* Low level logging. To use only for very big messages, otherwise
206 * redisLog() is to prefer. */
207 void redisLogRaw(int level, const char *msg) {
208 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
209 const char *c = ".-*#";
210 time_t now = time(NULL);
211 FILE *fp;
212 char buf[64];
213 int rawmode = (level & REDIS_LOG_RAW);
214
215 level &= 0xff; /* clear flags */
216 if (level < server.verbosity) return;
217
218 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
219 if (!fp) return;
220
221 if (rawmode) {
222 fprintf(fp,"%s",msg);
223 } else {
224 strftime(buf,sizeof(buf),"%d %b %H:%M:%S",localtime(&now));
225 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
226 }
227 fflush(fp);
228
229 if (server.logfile) fclose(fp);
230
231 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
232 }
233
234 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
235 * is used across the code. The raw version is only used in order to dump
236 * the INFO output on crash. */
237 void redisLog(int level, const char *fmt, ...) {
238 va_list ap;
239 char msg[REDIS_MAX_LOGMSG_LEN];
240
241 if ((level&0xff) < server.verbosity) return;
242
243 va_start(ap, fmt);
244 vsnprintf(msg, sizeof(msg), fmt, ap);
245 va_end(ap);
246
247 redisLogRaw(level,msg);
248 }
249
250 /* Redis generally does not try to recover from out of memory conditions
251 * when allocating objects or strings, it is not clear if it will be possible
252 * to report this condition to the client since the networking layer itself
253 * is based on heap allocation for send buffers, so we simply abort.
254 * At least the code will be simpler to read... */
255 void oom(const char *msg) {
256 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
257 sleep(1);
258 abort();
259 }
260
261 /* Return the UNIX time in microseconds */
262 long long ustime(void) {
263 struct timeval tv;
264 long long ust;
265
266 gettimeofday(&tv, NULL);
267 ust = ((long long)tv.tv_sec)*1000000;
268 ust += tv.tv_usec;
269 return ust;
270 }
271
272 /*====================== Hash table type implementation ==================== */
273
274 /* This is an hash table type that uses the SDS dynamic strings libary as
275 * keys and radis objects as values (objects can hold SDS strings,
276 * lists, sets). */
277
278 void dictVanillaFree(void *privdata, void *val)
279 {
280 DICT_NOTUSED(privdata);
281 zfree(val);
282 }
283
284 void dictListDestructor(void *privdata, void *val)
285 {
286 DICT_NOTUSED(privdata);
287 listRelease((list*)val);
288 }
289
290 int dictSdsKeyCompare(void *privdata, const void *key1,
291 const void *key2)
292 {
293 int l1,l2;
294 DICT_NOTUSED(privdata);
295
296 l1 = sdslen((sds)key1);
297 l2 = sdslen((sds)key2);
298 if (l1 != l2) return 0;
299 return memcmp(key1, key2, l1) == 0;
300 }
301
302 /* A case insensitive version used for the command lookup table. */
303 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
304 const void *key2)
305 {
306 DICT_NOTUSED(privdata);
307
308 return strcasecmp(key1, key2) == 0;
309 }
310
311 void dictRedisObjectDestructor(void *privdata, void *val)
312 {
313 DICT_NOTUSED(privdata);
314
315 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
316 decrRefCount(val);
317 }
318
319 void dictSdsDestructor(void *privdata, void *val)
320 {
321 DICT_NOTUSED(privdata);
322
323 sdsfree(val);
324 }
325
326 int dictObjKeyCompare(void *privdata, const void *key1,
327 const void *key2)
328 {
329 const robj *o1 = key1, *o2 = key2;
330 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
331 }
332
333 unsigned int dictObjHash(const void *key) {
334 const robj *o = key;
335 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
336 }
337
338 unsigned int dictSdsHash(const void *key) {
339 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
340 }
341
342 unsigned int dictSdsCaseHash(const void *key) {
343 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
344 }
345
346 int dictEncObjKeyCompare(void *privdata, const void *key1,
347 const void *key2)
348 {
349 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
350 int cmp;
351
352 if (o1->encoding == REDIS_ENCODING_INT &&
353 o2->encoding == REDIS_ENCODING_INT)
354 return o1->ptr == o2->ptr;
355
356 o1 = getDecodedObject(o1);
357 o2 = getDecodedObject(o2);
358 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
359 decrRefCount(o1);
360 decrRefCount(o2);
361 return cmp;
362 }
363
364 unsigned int dictEncObjHash(const void *key) {
365 robj *o = (robj*) key;
366
367 if (o->encoding == REDIS_ENCODING_RAW) {
368 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
369 } else {
370 if (o->encoding == REDIS_ENCODING_INT) {
371 char buf[32];
372 int len;
373
374 len = ll2string(buf,32,(long)o->ptr);
375 return dictGenHashFunction((unsigned char*)buf, len);
376 } else {
377 unsigned int hash;
378
379 o = getDecodedObject(o);
380 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
381 decrRefCount(o);
382 return hash;
383 }
384 }
385 }
386
387 /* Sets type hash table */
388 dictType setDictType = {
389 dictEncObjHash, /* hash function */
390 NULL, /* key dup */
391 NULL, /* val dup */
392 dictEncObjKeyCompare, /* key compare */
393 dictRedisObjectDestructor, /* key destructor */
394 NULL /* val destructor */
395 };
396
397 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
398 dictType zsetDictType = {
399 dictEncObjHash, /* hash function */
400 NULL, /* key dup */
401 NULL, /* val dup */
402 dictEncObjKeyCompare, /* key compare */
403 dictRedisObjectDestructor, /* key destructor */
404 NULL /* val destructor */
405 };
406
407 /* Db->dict, keys are sds strings, vals are Redis objects. */
408 dictType dbDictType = {
409 dictSdsHash, /* hash function */
410 NULL, /* key dup */
411 NULL, /* val dup */
412 dictSdsKeyCompare, /* key compare */
413 dictSdsDestructor, /* key destructor */
414 dictRedisObjectDestructor /* val destructor */
415 };
416
417 /* Db->expires */
418 dictType keyptrDictType = {
419 dictSdsHash, /* hash function */
420 NULL, /* key dup */
421 NULL, /* val dup */
422 dictSdsKeyCompare, /* key compare */
423 NULL, /* key destructor */
424 NULL /* val destructor */
425 };
426
427 /* Command table. sds string -> command struct pointer. */
428 dictType commandTableDictType = {
429 dictSdsCaseHash, /* hash function */
430 NULL, /* key dup */
431 NULL, /* val dup */
432 dictSdsKeyCaseCompare, /* key compare */
433 dictSdsDestructor, /* key destructor */
434 NULL /* val destructor */
435 };
436
437 /* Hash type hash table (note that small hashes are represented with zimpaps) */
438 dictType hashDictType = {
439 dictEncObjHash, /* hash function */
440 NULL, /* key dup */
441 NULL, /* val dup */
442 dictEncObjKeyCompare, /* key compare */
443 dictRedisObjectDestructor, /* key destructor */
444 dictRedisObjectDestructor /* val destructor */
445 };
446
447 /* Keylist hash table type has unencoded redis objects as keys and
448 * lists as values. It's used for blocking operations (BLPOP) and to
449 * map swapped keys to a list of clients waiting for this keys to be loaded. */
450 dictType keylistDictType = {
451 dictObjHash, /* hash function */
452 NULL, /* key dup */
453 NULL, /* val dup */
454 dictObjKeyCompare, /* key compare */
455 dictRedisObjectDestructor, /* key destructor */
456 dictListDestructor /* val destructor */
457 };
458
459 /* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
460 * clusterNode structures. */
461 dictType clusterNodesDictType = {
462 dictSdsHash, /* hash function */
463 NULL, /* key dup */
464 NULL, /* val dup */
465 dictSdsKeyCompare, /* key compare */
466 dictSdsDestructor, /* key destructor */
467 NULL /* val destructor */
468 };
469
470 int htNeedsResize(dict *dict) {
471 long long size, used;
472
473 size = dictSlots(dict);
474 used = dictSize(dict);
475 return (size && used && size > DICT_HT_INITIAL_SIZE &&
476 (used*100/size < REDIS_HT_MINFILL));
477 }
478
479 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
480 * we resize the hash table to save memory */
481 void tryResizeHashTables(void) {
482 int j;
483
484 for (j = 0; j < server.dbnum; j++) {
485 if (htNeedsResize(server.db[j].dict))
486 dictResize(server.db[j].dict);
487 if (htNeedsResize(server.db[j].expires))
488 dictResize(server.db[j].expires);
489 }
490 }
491
492 /* Our hash table implementation performs rehashing incrementally while
493 * we write/read from the hash table. Still if the server is idle, the hash
494 * table will use two tables for a long time. So we try to use 1 millisecond
495 * of CPU time at every serverCron() loop in order to rehash some key. */
496 void incrementallyRehash(void) {
497 int j;
498
499 for (j = 0; j < server.dbnum; j++) {
500 if (dictIsRehashing(server.db[j].dict)) {
501 dictRehashMilliseconds(server.db[j].dict,1);
502 break; /* already used our millisecond for this loop... */
503 }
504 }
505 }
506
507 /* This function is called once a background process of some kind terminates,
508 * as we want to avoid resizing the hash tables when there is a child in order
509 * to play well with copy-on-write (otherwise when a resize happens lots of
510 * memory pages are copied). The goal of this function is to update the ability
511 * for dict.c to resize the hash tables accordingly to the fact we have o not
512 * running childs. */
513 void updateDictResizePolicy(void) {
514 if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1)
515 dictEnableResize();
516 else
517 dictDisableResize();
518 }
519
520 /* ======================= Cron: called every 100 ms ======================== */
521
522 /* Try to expire a few timed out keys. The algorithm used is adaptive and
523 * will use few CPU cycles if there are few expiring keys, otherwise
524 * it will get more aggressive to avoid that too much memory is used by
525 * keys that can be removed from the keyspace. */
526 void activeExpireCycle(void) {
527 int j;
528
529 for (j = 0; j < server.dbnum; j++) {
530 int expired;
531 redisDb *db = server.db+j;
532
533 /* Continue to expire if at the end of the cycle more than 25%
534 * of the keys were expired. */
535 do {
536 long num = dictSize(db->expires);
537 time_t now = time(NULL);
538
539 expired = 0;
540 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
541 num = REDIS_EXPIRELOOKUPS_PER_CRON;
542 while (num--) {
543 dictEntry *de;
544 time_t t;
545
546 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
547 t = (time_t) dictGetEntryVal(de);
548 if (now > t) {
549 sds key = dictGetEntryKey(de);
550 robj *keyobj = createStringObject(key,sdslen(key));
551
552 propagateExpire(db,keyobj);
553 dbDelete(db,keyobj);
554 decrRefCount(keyobj);
555 expired++;
556 server.stat_expiredkeys++;
557 }
558 }
559 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
560 }
561 }
562
563 void updateLRUClock(void) {
564 server.lruclock = (time(NULL)/REDIS_LRU_CLOCK_RESOLUTION) &
565 REDIS_LRU_CLOCK_MAX;
566 }
567
568 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
569 int j, loops = server.cronloops;
570 REDIS_NOTUSED(eventLoop);
571 REDIS_NOTUSED(id);
572 REDIS_NOTUSED(clientData);
573
574 /* We take a cached value of the unix time in the global state because
575 * with virtual memory and aging there is to store the current time
576 * in objects at every object access, and accuracy is not needed.
577 * To access a global var is faster than calling time(NULL) */
578 server.unixtime = time(NULL);
579
580 /* We have just 22 bits per object for LRU information.
581 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
582 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
583 *
584 * Note that even if this will wrap after 1.5 years it's not a problem,
585 * everything will still work but just some object will appear younger
586 * to Redis. But for this to happen a given object should never be touched
587 * for 1.5 years.
588 *
589 * Note that you can change the resolution altering the
590 * REDIS_LRU_CLOCK_RESOLUTION define.
591 */
592 updateLRUClock();
593
594 /* Record the max memory used since the server was started. */
595 if (zmalloc_used_memory() > server.stat_peak_memory)
596 server.stat_peak_memory = zmalloc_used_memory();
597
598 /* We received a SIGTERM, shutting down here in a safe way, as it is
599 * not ok doing so inside the signal handler. */
600 if (server.shutdown_asap) {
601 if (prepareForShutdown() == REDIS_OK) exit(0);
602 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
603 }
604
605 /* Show some info about non-empty databases */
606 for (j = 0; j < server.dbnum; j++) {
607 long long size, used, vkeys;
608
609 size = dictSlots(server.db[j].dict);
610 used = dictSize(server.db[j].dict);
611 vkeys = dictSize(server.db[j].expires);
612 if (!(loops % 50) && (used || vkeys)) {
613 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
614 /* dictPrintStats(server.dict); */
615 }
616 }
617
618 /* We don't want to resize the hash tables while a bacground saving
619 * is in progress: the saving child is created using fork() that is
620 * implemented with a copy-on-write semantic in most modern systems, so
621 * if we resize the HT while there is the saving child at work actually
622 * a lot of memory movements in the parent will cause a lot of pages
623 * copied. */
624 if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1) {
625 if (!(loops % 10)) tryResizeHashTables();
626 if (server.activerehashing) incrementallyRehash();
627 }
628
629 /* Show information about connected clients */
630 if (!(loops % 50)) {
631 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
632 listLength(server.clients)-listLength(server.slaves),
633 listLength(server.slaves),
634 zmalloc_used_memory());
635 }
636
637 /* Close connections of timedout clients */
638 if ((server.maxidletime && !(loops % 100)) || server.bpop_blocked_clients)
639 closeTimedoutClients();
640
641 /* Start a scheduled AOF rewrite if this was requested by the user while
642 * a BGSAVE was in progress. */
643 if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1 &&
644 server.aofrewrite_scheduled)
645 {
646 rewriteAppendOnlyFileBackground();
647 }
648
649 /* Check if a background saving or AOF rewrite in progress terminated. */
650 if (server.bgsavechildpid != -1 || server.bgrewritechildpid != -1) {
651 int statloc;
652 pid_t pid;
653
654 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
655 int exitcode = WEXITSTATUS(statloc);
656 int bysignal = 0;
657
658 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
659
660 if (pid == server.bgsavechildpid) {
661 backgroundSaveDoneHandler(exitcode,bysignal);
662 } else {
663 backgroundRewriteDoneHandler(exitcode,bysignal);
664 }
665 updateDictResizePolicy();
666 }
667 } else {
668 time_t now = time(NULL);
669
670 /* If there is not a background saving/rewrite in progress check if
671 * we have to save/rewrite now */
672 for (j = 0; j < server.saveparamslen; j++) {
673 struct saveparam *sp = server.saveparams+j;
674
675 if (server.dirty >= sp->changes &&
676 now-server.lastsave > sp->seconds) {
677 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
678 sp->changes, sp->seconds);
679 rdbSaveBackground(server.dbfilename);
680 break;
681 }
682 }
683
684 /* Trigger an AOF rewrite if needed */
685 if (server.bgsavechildpid == -1 &&
686 server.bgrewritechildpid == -1 &&
687 server.auto_aofrewrite_perc &&
688 server.appendonly_current_size > server.auto_aofrewrite_min_size)
689 {
690 long long base = server.auto_aofrewrite_base_size ?
691 server.auto_aofrewrite_base_size : 1;
692 long long growth = (server.appendonly_current_size*100/base) - 100;
693 if (growth >= server.auto_aofrewrite_perc) {
694 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
695 rewriteAppendOnlyFileBackground();
696 }
697 }
698 }
699
700 /* Expire a few keys per cycle, only if this is a master.
701 * On slaves we wait for DEL operations synthesized by the master
702 * in order to guarantee a strict consistency. */
703 if (server.masterhost == NULL) activeExpireCycle();
704
705 /* Replication cron function -- used to reconnect to master and
706 * to detect transfer failures. */
707 if (!(loops % 10)) replicationCron();
708
709 /* Run other sub-systems specific cron jobs */
710 if (server.cluster_enabled && !(loops % 10)) clusterCron();
711
712 server.cronloops++;
713 return 100;
714 }
715
716 /* This function gets called every time Redis is entering the
717 * main loop of the event driven library, that is, before to sleep
718 * for ready file descriptors. */
719 void beforeSleep(struct aeEventLoop *eventLoop) {
720 REDIS_NOTUSED(eventLoop);
721 listNode *ln;
722 redisClient *c;
723
724 /* Try to process pending commands for clients that were just unblocked. */
725 while (listLength(server.unblocked_clients)) {
726 ln = listFirst(server.unblocked_clients);
727 redisAssert(ln != NULL);
728 c = ln->value;
729 listDelNode(server.unblocked_clients,ln);
730 c->flags &= ~REDIS_UNBLOCKED;
731
732 /* Process remaining data in the input buffer. */
733 if (c->querybuf && sdslen(c->querybuf) > 0)
734 processInputBuffer(c);
735 }
736
737 /* Write the AOF buffer on disk */
738 flushAppendOnlyFile();
739 }
740
741 /* =========================== Server initialization ======================== */
742
743 void createSharedObjects(void) {
744 int j;
745
746 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
747 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
748 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
749 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
750 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
751 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
752 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
753 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
754 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
755 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
756 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
757 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
758 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
759 "-ERR Operation against a key holding the wrong kind of value\r\n"));
760 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
761 "-ERR no such key\r\n"));
762 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
763 "-ERR syntax error\r\n"));
764 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
765 "-ERR source and destination objects are the same\r\n"));
766 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
767 "-ERR index out of range\r\n"));
768 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
769 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
770 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
771 "-LOADING Redis is loading the dataset in memory\r\n"));
772 shared.space = createObject(REDIS_STRING,sdsnew(" "));
773 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
774 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
775 shared.select0 = createStringObject("select 0\r\n",10);
776 shared.select1 = createStringObject("select 1\r\n",10);
777 shared.select2 = createStringObject("select 2\r\n",10);
778 shared.select3 = createStringObject("select 3\r\n",10);
779 shared.select4 = createStringObject("select 4\r\n",10);
780 shared.select5 = createStringObject("select 5\r\n",10);
781 shared.select6 = createStringObject("select 6\r\n",10);
782 shared.select7 = createStringObject("select 7\r\n",10);
783 shared.select8 = createStringObject("select 8\r\n",10);
784 shared.select9 = createStringObject("select 9\r\n",10);
785 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
786 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
787 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
788 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
789 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
790 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
791 shared.mbulk3 = createStringObject("*3\r\n",4);
792 shared.mbulk4 = createStringObject("*4\r\n",4);
793 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
794 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
795 shared.integers[j]->encoding = REDIS_ENCODING_INT;
796 }
797 }
798
799 void initServerConfig() {
800 server.port = REDIS_SERVERPORT;
801 server.bindaddr = NULL;
802 server.unixsocket = NULL;
803 server.ipfd = -1;
804 server.sofd = -1;
805 server.dbnum = REDIS_DEFAULT_DBNUM;
806 server.verbosity = REDIS_VERBOSE;
807 server.maxidletime = REDIS_MAXIDLETIME;
808 server.saveparams = NULL;
809 server.loading = 0;
810 server.logfile = NULL; /* NULL = log on standard output */
811 server.syslog_enabled = 0;
812 server.syslog_ident = zstrdup("redis");
813 server.syslog_facility = LOG_LOCAL0;
814 server.daemonize = 0;
815 server.appendonly = 0;
816 server.appendfsync = APPENDFSYNC_EVERYSEC;
817 server.no_appendfsync_on_rewrite = 0;
818 server.auto_aofrewrite_perc = REDIS_AUTO_AOFREWRITE_PERC;
819 server.auto_aofrewrite_min_size = REDIS_AUTO_AOFREWRITE_MIN_SIZE;
820 server.auto_aofrewrite_base_size = 0;
821 server.aofrewrite_scheduled = 0;
822 server.lastfsync = time(NULL);
823 server.appendfd = -1;
824 server.appendseldb = -1; /* Make sure the first time will not match */
825 server.pidfile = zstrdup("/var/run/redis.pid");
826 server.dbfilename = zstrdup("dump.rdb");
827 server.appendfilename = zstrdup("appendonly.aof");
828 server.requirepass = NULL;
829 server.rdbcompression = 1;
830 server.activerehashing = 1;
831 server.maxclients = 0;
832 server.bpop_blocked_clients = 0;
833 server.maxmemory = 0;
834 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
835 server.maxmemory_samples = 3;
836 server.hash_max_zipmap_entries = REDIS_HASH_MAX_ZIPMAP_ENTRIES;
837 server.hash_max_zipmap_value = REDIS_HASH_MAX_ZIPMAP_VALUE;
838 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
839 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
840 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
841 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
842 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
843 server.shutdown_asap = 0;
844 server.cluster_enabled = 0;
845 server.cluster.configfile = zstrdup("nodes.conf");
846 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
847
848 updateLRUClock();
849 resetServerSaveParams();
850
851 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
852 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
853 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
854 /* Replication related */
855 server.isslave = 0;
856 server.masterauth = NULL;
857 server.masterhost = NULL;
858 server.masterport = 6379;
859 server.master = NULL;
860 server.replstate = REDIS_REPL_NONE;
861 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
862 server.repl_serve_stale_data = 1;
863 server.repl_down_since = -1;
864
865 /* Double constants initialization */
866 R_Zero = 0.0;
867 R_PosInf = 1.0/R_Zero;
868 R_NegInf = -1.0/R_Zero;
869 R_Nan = R_Zero/R_Zero;
870
871 /* Command table -- we intiialize it here as it is part of the
872 * initial configuration, since command names may be changed via
873 * redis.conf using the rename-command directive. */
874 server.commands = dictCreate(&commandTableDictType,NULL);
875 populateCommandTable();
876 server.delCommand = lookupCommandByCString("del");
877 server.multiCommand = lookupCommandByCString("multi");
878
879 /* Slow log */
880 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
881 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
882 }
883
884 void initServer() {
885 int j;
886
887 signal(SIGHUP, SIG_IGN);
888 signal(SIGPIPE, SIG_IGN);
889 setupSignalHandlers();
890
891 if (server.syslog_enabled) {
892 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
893 server.syslog_facility);
894 }
895
896 server.clients = listCreate();
897 server.slaves = listCreate();
898 server.monitors = listCreate();
899 server.unblocked_clients = listCreate();
900
901 createSharedObjects();
902 server.el = aeCreateEventLoop();
903 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
904
905 if (server.port != 0) {
906 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
907 if (server.ipfd == ANET_ERR) {
908 redisLog(REDIS_WARNING, "Opening port: %s", server.neterr);
909 exit(1);
910 }
911 }
912 if (server.unixsocket != NULL) {
913 unlink(server.unixsocket); /* don't care if this fails */
914 server.sofd = anetUnixServer(server.neterr,server.unixsocket);
915 if (server.sofd == ANET_ERR) {
916 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
917 exit(1);
918 }
919 }
920 if (server.ipfd < 0 && server.sofd < 0) {
921 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
922 exit(1);
923 }
924 for (j = 0; j < server.dbnum; j++) {
925 server.db[j].dict = dictCreate(&dbDictType,NULL);
926 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
927 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
928 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
929 server.db[j].id = j;
930 }
931 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
932 server.pubsub_patterns = listCreate();
933 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
934 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
935 server.cronloops = 0;
936 server.bgsavechildpid = -1;
937 server.bgrewritechildpid = -1;
938 server.bgrewritebuf = sdsempty();
939 server.aofbuf = sdsempty();
940 server.lastsave = time(NULL);
941 server.dirty = 0;
942 server.stat_numcommands = 0;
943 server.stat_numconnections = 0;
944 server.stat_expiredkeys = 0;
945 server.stat_evictedkeys = 0;
946 server.stat_starttime = time(NULL);
947 server.stat_keyspace_misses = 0;
948 server.stat_keyspace_hits = 0;
949 server.stat_peak_memory = 0;
950 server.stat_fork_time = 0;
951 server.unixtime = time(NULL);
952 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
953 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
954 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
955 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
956 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
957
958 if (server.appendonly) {
959 server.appendfd = open(server.appendfilename,O_WRONLY|O_APPEND|O_CREAT,0644);
960 if (server.appendfd == -1) {
961 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
962 strerror(errno));
963 exit(1);
964 }
965 }
966
967 if (server.cluster_enabled) clusterInit();
968 scriptingInit();
969 slowlogInit();
970 bioInit();
971 srand(time(NULL)^getpid());
972 }
973
974 /* Populates the Redis Command Table starting from the hard coded list
975 * we have on top of redis.c file. */
976 void populateCommandTable(void) {
977 int j;
978 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
979
980 for (j = 0; j < numcommands; j++) {
981 struct redisCommand *c = redisCommandTable+j;
982 int retval;
983
984 retval = dictAdd(server.commands, sdsnew(c->name), c);
985 assert(retval == DICT_OK);
986 }
987 }
988
989 void resetCommandTableStats(void) {
990 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
991 int j;
992
993 for (j = 0; j < numcommands; j++) {
994 struct redisCommand *c = redisCommandTable+j;
995
996 c->microseconds = 0;
997 c->calls = 0;
998 }
999 }
1000
1001 /* ====================== Commands lookup and execution ===================== */
1002
1003 struct redisCommand *lookupCommand(sds name) {
1004 return dictFetchValue(server.commands, name);
1005 }
1006
1007 struct redisCommand *lookupCommandByCString(char *s) {
1008 struct redisCommand *cmd;
1009 sds name = sdsnew(s);
1010
1011 cmd = dictFetchValue(server.commands, name);
1012 sdsfree(name);
1013 return cmd;
1014 }
1015
1016 /* Call() is the core of Redis execution of a command */
1017 void call(redisClient *c) {
1018 long long dirty, start = ustime(), duration;
1019
1020 dirty = server.dirty;
1021 c->cmd->proc(c);
1022 dirty = server.dirty-dirty;
1023 duration = ustime()-start;
1024 c->cmd->microseconds += duration;
1025 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1026 c->cmd->calls++;
1027
1028 if (server.appendonly && dirty > 0)
1029 feedAppendOnlyFile(c->cmd,c->db->id,c->argv,c->argc);
1030 if ((dirty > 0 || c->cmd->flags & REDIS_CMD_FORCE_REPLICATION) &&
1031 listLength(server.slaves))
1032 replicationFeedSlaves(server.slaves,c->db->id,c->argv,c->argc);
1033 if (listLength(server.monitors))
1034 replicationFeedMonitors(server.monitors,c->db->id,c->argv,c->argc);
1035 server.stat_numcommands++;
1036 }
1037
1038 /* If this function gets called we already read a whole
1039 * command, argments are in the client argv/argc fields.
1040 * processCommand() execute the command or prepare the
1041 * server for a bulk read from the client.
1042 *
1043 * If 1 is returned the client is still alive and valid and
1044 * and other operations can be performed by the caller. Otherwise
1045 * if 0 is returned the client was destroied (i.e. after QUIT). */
1046 int processCommand(redisClient *c) {
1047 /* The QUIT command is handled separately. Normal command procs will
1048 * go through checking for replication and QUIT will cause trouble
1049 * when FORCE_REPLICATION is enabled and would be implemented in
1050 * a regular command proc. */
1051 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1052 addReply(c,shared.ok);
1053 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1054 return REDIS_ERR;
1055 }
1056
1057 /* Now lookup the command and check ASAP about trivial error conditions
1058 * such as wrong arity, bad command name and so forth. */
1059 c->cmd = lookupCommand(c->argv[0]->ptr);
1060 if (!c->cmd) {
1061 addReplyErrorFormat(c,"unknown command '%s'",
1062 (char*)c->argv[0]->ptr);
1063 return REDIS_OK;
1064 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1065 (c->argc < -c->cmd->arity)) {
1066 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1067 c->cmd->name);
1068 return REDIS_OK;
1069 }
1070
1071 /* Check if the user is authenticated */
1072 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1073 {
1074 addReplyError(c,"operation not permitted");
1075 return REDIS_OK;
1076 }
1077
1078 /* If cluster is enabled, redirect here */
1079 if (server.cluster_enabled &&
1080 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1081 int hashslot;
1082
1083 if (server.cluster.state != REDIS_CLUSTER_OK) {
1084 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1085 return REDIS_OK;
1086 } else {
1087 int ask;
1088 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1089 if (n == NULL) {
1090 addReplyError(c,"Multi keys request invalid in cluster");
1091 return REDIS_OK;
1092 } else if (n != server.cluster.myself) {
1093 addReplySds(c,sdscatprintf(sdsempty(),
1094 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1095 hashslot,n->ip,n->port));
1096 return REDIS_OK;
1097 }
1098 }
1099 }
1100
1101 /* Handle the maxmemory directive.
1102 *
1103 * First we try to free some memory if possible (if there are volatile
1104 * keys in the dataset). If there are not the only thing we can do
1105 * is returning an error. */
1106 if (server.maxmemory) freeMemoryIfNeeded();
1107 if (server.maxmemory && (c->cmd->flags & REDIS_CMD_DENYOOM) &&
1108 zmalloc_used_memory() > server.maxmemory)
1109 {
1110 addReplyError(c,"command not allowed when used memory > 'maxmemory'");
1111 return REDIS_OK;
1112 }
1113
1114 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1115 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1116 &&
1117 c->cmd->proc != subscribeCommand &&
1118 c->cmd->proc != unsubscribeCommand &&
1119 c->cmd->proc != psubscribeCommand &&
1120 c->cmd->proc != punsubscribeCommand) {
1121 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1122 return REDIS_OK;
1123 }
1124
1125 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1126 * we are a slave with a broken link with master. */
1127 if (server.masterhost && server.replstate != REDIS_REPL_CONNECTED &&
1128 server.repl_serve_stale_data == 0 &&
1129 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1130 {
1131 addReplyError(c,
1132 "link with MASTER is down and slave-serve-stale-data is set to no");
1133 return REDIS_OK;
1134 }
1135
1136 /* Loading DB? Return an error if the command is not INFO */
1137 if (server.loading && c->cmd->proc != infoCommand) {
1138 addReply(c, shared.loadingerr);
1139 return REDIS_OK;
1140 }
1141
1142 /* Exec the command */
1143 if (c->flags & REDIS_MULTI &&
1144 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1145 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1146 {
1147 queueMultiCommand(c);
1148 addReply(c,shared.queued);
1149 } else {
1150 call(c);
1151 }
1152 return REDIS_OK;
1153 }
1154
1155 /*================================== Shutdown =============================== */
1156
1157 int prepareForShutdown() {
1158 redisLog(REDIS_WARNING,"User requested shutdown...");
1159 /* Kill the saving child if there is a background saving in progress.
1160 We want to avoid race conditions, for instance our saving child may
1161 overwrite the synchronous saving did by SHUTDOWN. */
1162 if (server.bgsavechildpid != -1) {
1163 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1164 kill(server.bgsavechildpid,SIGKILL);
1165 rdbRemoveTempFile(server.bgsavechildpid);
1166 }
1167 if (server.appendonly) {
1168 /* Kill the AOF saving child as the AOF we already have may be longer
1169 * but contains the full dataset anyway. */
1170 if (server.bgrewritechildpid != -1) {
1171 redisLog(REDIS_WARNING,
1172 "There is a child rewriting the AOF. Killing it!");
1173 kill(server.bgrewritechildpid,SIGKILL);
1174 }
1175 /* Append only file: fsync() the AOF and exit */
1176 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1177 aof_fsync(server.appendfd);
1178 }
1179 if (server.saveparamslen > 0) {
1180 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1181 /* Snapshotting. Perform a SYNC SAVE and exit */
1182 if (rdbSave(server.dbfilename) != REDIS_OK) {
1183 /* Ooops.. error saving! The best we can do is to continue
1184 * operating. Note that if there was a background saving process,
1185 * in the next cron() Redis will be notified that the background
1186 * saving aborted, handling special stuff like slaves pending for
1187 * synchronization... */
1188 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1189 return REDIS_ERR;
1190 }
1191 }
1192 if (server.daemonize) {
1193 redisLog(REDIS_NOTICE,"Removing the pid file.");
1194 unlink(server.pidfile);
1195 }
1196 /* Close the listening sockets. Apparently this allows faster restarts. */
1197 if (server.ipfd != -1) close(server.ipfd);
1198 if (server.sofd != -1) close(server.sofd);
1199
1200 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1201 return REDIS_OK;
1202 }
1203
1204 /*================================== Commands =============================== */
1205
1206 void authCommand(redisClient *c) {
1207 if (!server.requirepass || !strcmp(c->argv[1]->ptr, server.requirepass)) {
1208 c->authenticated = 1;
1209 addReply(c,shared.ok);
1210 } else {
1211 c->authenticated = 0;
1212 addReplyError(c,"invalid password");
1213 }
1214 }
1215
1216 void pingCommand(redisClient *c) {
1217 addReply(c,shared.pong);
1218 }
1219
1220 void echoCommand(redisClient *c) {
1221 addReplyBulk(c,c->argv[1]);
1222 }
1223
1224 /* Convert an amount of bytes into a human readable string in the form
1225 * of 100B, 2G, 100M, 4K, and so forth. */
1226 void bytesToHuman(char *s, unsigned long long n) {
1227 double d;
1228
1229 if (n < 1024) {
1230 /* Bytes */
1231 sprintf(s,"%lluB",n);
1232 return;
1233 } else if (n < (1024*1024)) {
1234 d = (double)n/(1024);
1235 sprintf(s,"%.2fK",d);
1236 } else if (n < (1024LL*1024*1024)) {
1237 d = (double)n/(1024*1024);
1238 sprintf(s,"%.2fM",d);
1239 } else if (n < (1024LL*1024*1024*1024)) {
1240 d = (double)n/(1024LL*1024*1024);
1241 sprintf(s,"%.2fG",d);
1242 }
1243 }
1244
1245 /* Create the string returned by the INFO command. This is decoupled
1246 * by the INFO command itself as we need to report the same information
1247 * on memory corruption problems. */
1248 sds genRedisInfoString(char *section) {
1249 sds info = sdsempty();
1250 time_t uptime = time(NULL)-server.stat_starttime;
1251 int j, numcommands;
1252 struct rusage self_ru, c_ru;
1253 unsigned long lol, bib;
1254 int allsections = 0, defsections = 0;
1255 int sections = 0;
1256
1257 if (section) {
1258 allsections = strcasecmp(section,"all") == 0;
1259 defsections = strcasecmp(section,"default") == 0;
1260 }
1261
1262 getrusage(RUSAGE_SELF, &self_ru);
1263 getrusage(RUSAGE_CHILDREN, &c_ru);
1264 getClientsMaxBuffers(&lol,&bib);
1265
1266 /* Server */
1267 if (allsections || defsections || !strcasecmp(section,"server")) {
1268 if (sections++) info = sdscat(info,"\r\n");
1269 info = sdscatprintf(info,
1270 "# Server\r\n"
1271 "redis_version:%s\r\n"
1272 "redis_git_sha1:%s\r\n"
1273 "redis_git_dirty:%d\r\n"
1274 "arch_bits:%s\r\n"
1275 "multiplexing_api:%s\r\n"
1276 "process_id:%ld\r\n"
1277 "tcp_port:%d\r\n"
1278 "uptime_in_seconds:%ld\r\n"
1279 "uptime_in_days:%ld\r\n"
1280 "lru_clock:%ld\r\n",
1281 REDIS_VERSION,
1282 redisGitSHA1(),
1283 strtol(redisGitDirty(),NULL,10) > 0,
1284 (sizeof(long) == 8) ? "64" : "32",
1285 aeGetApiName(),
1286 (long) getpid(),
1287 server.port,
1288 uptime,
1289 uptime/(3600*24),
1290 (unsigned long) server.lruclock);
1291 }
1292
1293 /* Clients */
1294 if (allsections || defsections || !strcasecmp(section,"clients")) {
1295 if (sections++) info = sdscat(info,"\r\n");
1296 info = sdscatprintf(info,
1297 "# Clients\r\n"
1298 "connected_clients:%d\r\n"
1299 "client_longest_output_list:%lu\r\n"
1300 "client_biggest_input_buf:%lu\r\n"
1301 "blocked_clients:%d\r\n",
1302 listLength(server.clients)-listLength(server.slaves),
1303 lol, bib,
1304 server.bpop_blocked_clients);
1305 }
1306
1307 /* Memory */
1308 if (allsections || defsections || !strcasecmp(section,"memory")) {
1309 char hmem[64];
1310 char peak_hmem[64];
1311
1312 bytesToHuman(hmem,zmalloc_used_memory());
1313 bytesToHuman(peak_hmem,server.stat_peak_memory);
1314 if (sections++) info = sdscat(info,"\r\n");
1315 info = sdscatprintf(info,
1316 "# Memory\r\n"
1317 "used_memory:%zu\r\n"
1318 "used_memory_human:%s\r\n"
1319 "used_memory_rss:%zu\r\n"
1320 "used_memory_peak:%zu\r\n"
1321 "used_memory_peak_human:%s\r\n"
1322 "used_memory_lua:%lld\r\n"
1323 "mem_fragmentation_ratio:%.2f\r\n"
1324 "mem_allocator:%s\r\n",
1325 zmalloc_used_memory(),
1326 hmem,
1327 zmalloc_get_rss(),
1328 server.stat_peak_memory,
1329 peak_hmem,
1330 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1331 zmalloc_get_fragmentation_ratio(),
1332 ZMALLOC_LIB
1333 );
1334 }
1335
1336 /* Persistence */
1337 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1338 if (sections++) info = sdscat(info,"\r\n");
1339 info = sdscatprintf(info,
1340 "# Persistence\r\n"
1341 "loading:%d\r\n"
1342 "aof_enabled:%d\r\n"
1343 "changes_since_last_save:%lld\r\n"
1344 "bgsave_in_progress:%d\r\n"
1345 "last_save_time:%ld\r\n"
1346 "bgrewriteaof_in_progress:%d\r\n",
1347 server.loading,
1348 server.appendonly,
1349 server.dirty,
1350 server.bgsavechildpid != -1,
1351 server.lastsave,
1352 server.bgrewritechildpid != -1);
1353
1354 if (server.appendonly) {
1355 info = sdscatprintf(info,
1356 "aof_current_size:%lld\r\n"
1357 "aof_base_size:%lld\r\n"
1358 "aof_pending_rewrite:%d\r\n",
1359 (long long) server.appendonly_current_size,
1360 (long long) server.auto_aofrewrite_base_size,
1361 server.aofrewrite_scheduled);
1362 }
1363
1364 if (server.loading) {
1365 double perc;
1366 time_t eta, elapsed;
1367 off_t remaining_bytes = server.loading_total_bytes-
1368 server.loading_loaded_bytes;
1369
1370 perc = ((double)server.loading_loaded_bytes /
1371 server.loading_total_bytes) * 100;
1372
1373 elapsed = time(NULL)-server.loading_start_time;
1374 if (elapsed == 0) {
1375 eta = 1; /* A fake 1 second figure if we don't have
1376 enough info */
1377 } else {
1378 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1379 }
1380
1381 info = sdscatprintf(info,
1382 "loading_start_time:%ld\r\n"
1383 "loading_total_bytes:%llu\r\n"
1384 "loading_loaded_bytes:%llu\r\n"
1385 "loading_loaded_perc:%.2f\r\n"
1386 "loading_eta_seconds:%ld\r\n"
1387 ,(unsigned long) server.loading_start_time,
1388 (unsigned long long) server.loading_total_bytes,
1389 (unsigned long long) server.loading_loaded_bytes,
1390 perc,
1391 eta
1392 );
1393 }
1394 }
1395
1396 /* Stats */
1397 if (allsections || defsections || !strcasecmp(section,"stats")) {
1398 if (sections++) info = sdscat(info,"\r\n");
1399 info = sdscatprintf(info,
1400 "# Stats\r\n"
1401 "total_connections_received:%lld\r\n"
1402 "total_commands_processed:%lld\r\n"
1403 "expired_keys:%lld\r\n"
1404 "evicted_keys:%lld\r\n"
1405 "keyspace_hits:%lld\r\n"
1406 "keyspace_misses:%lld\r\n"
1407 "pubsub_channels:%ld\r\n"
1408 "pubsub_patterns:%u\r\n"
1409 "latest_fork_usec:%lld\r\n",
1410 server.stat_numconnections,
1411 server.stat_numcommands,
1412 server.stat_expiredkeys,
1413 server.stat_evictedkeys,
1414 server.stat_keyspace_hits,
1415 server.stat_keyspace_misses,
1416 dictSize(server.pubsub_channels),
1417 listLength(server.pubsub_patterns),
1418 server.stat_fork_time);
1419 }
1420
1421 /* Replication */
1422 if (allsections || defsections || !strcasecmp(section,"replication")) {
1423 if (sections++) info = sdscat(info,"\r\n");
1424 info = sdscatprintf(info,
1425 "# Replication\r\n"
1426 "role:%s\r\n",
1427 server.masterhost == NULL ? "master" : "slave");
1428 if (server.masterhost) {
1429 info = sdscatprintf(info,
1430 "master_host:%s\r\n"
1431 "master_port:%d\r\n"
1432 "master_link_status:%s\r\n"
1433 "master_last_io_seconds_ago:%d\r\n"
1434 "master_sync_in_progress:%d\r\n"
1435 ,server.masterhost,
1436 server.masterport,
1437 (server.replstate == REDIS_REPL_CONNECTED) ?
1438 "up" : "down",
1439 server.master ?
1440 ((int)(time(NULL)-server.master->lastinteraction)) : -1,
1441 server.replstate == REDIS_REPL_TRANSFER
1442 );
1443
1444 if (server.replstate == REDIS_REPL_TRANSFER) {
1445 info = sdscatprintf(info,
1446 "master_sync_left_bytes:%ld\r\n"
1447 "master_sync_last_io_seconds_ago:%d\r\n"
1448 ,(long)server.repl_transfer_left,
1449 (int)(time(NULL)-server.repl_transfer_lastio)
1450 );
1451 }
1452
1453 if (server.replstate != REDIS_REPL_CONNECTED) {
1454 info = sdscatprintf(info,
1455 "master_link_down_since_seconds:%ld\r\n",
1456 (long)time(NULL)-server.repl_down_since);
1457 }
1458 }
1459 info = sdscatprintf(info,
1460 "connected_slaves:%d\r\n",
1461 listLength(server.slaves));
1462 }
1463
1464 /* CPU */
1465 if (allsections || defsections || !strcasecmp(section,"cpu")) {
1466 if (sections++) info = sdscat(info,"\r\n");
1467 info = sdscatprintf(info,
1468 "# CPU\r\n"
1469 "used_cpu_sys:%.2f\r\n"
1470 "used_cpu_user:%.2f\r\n"
1471 "used_cpu_sys_children:%.2f\r\n"
1472 "used_cpu_user_children:%.2f\r\n",
1473 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
1474 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
1475 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000,
1476 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000);
1477 }
1478
1479 /* cmdtime */
1480 if (allsections || !strcasecmp(section,"commandstats")) {
1481 if (sections++) info = sdscat(info,"\r\n");
1482 info = sdscatprintf(info, "# Commandstats\r\n");
1483 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1484 for (j = 0; j < numcommands; j++) {
1485 struct redisCommand *c = redisCommandTable+j;
1486
1487 if (!c->calls) continue;
1488 info = sdscatprintf(info,
1489 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
1490 c->name, c->calls, c->microseconds,
1491 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
1492 }
1493 }
1494
1495 /* Clusetr */
1496 if (allsections || defsections || !strcasecmp(section,"cluster")) {
1497 if (sections++) info = sdscat(info,"\r\n");
1498 info = sdscatprintf(info,
1499 "# Cluster\r\n"
1500 "cluster_enabled:%d\r\n",
1501 server.cluster_enabled);
1502 }
1503
1504 /* Key space */
1505 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
1506 if (sections++) info = sdscat(info,"\r\n");
1507 info = sdscatprintf(info, "# Keyspace\r\n");
1508 for (j = 0; j < server.dbnum; j++) {
1509 long long keys, vkeys;
1510
1511 keys = dictSize(server.db[j].dict);
1512 vkeys = dictSize(server.db[j].expires);
1513 if (keys || vkeys) {
1514 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
1515 j, keys, vkeys);
1516 }
1517 }
1518 }
1519 return info;
1520 }
1521
1522 void infoCommand(redisClient *c) {
1523 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
1524
1525 if (c->argc > 2) {
1526 addReply(c,shared.syntaxerr);
1527 return;
1528 }
1529 sds info = genRedisInfoString(section);
1530 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
1531 (unsigned long)sdslen(info)));
1532 addReplySds(c,info);
1533 addReply(c,shared.crlf);
1534 }
1535
1536 void monitorCommand(redisClient *c) {
1537 /* ignore MONITOR if aleady slave or in monitor mode */
1538 if (c->flags & REDIS_SLAVE) return;
1539
1540 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
1541 c->slaveseldb = 0;
1542 listAddNodeTail(server.monitors,c);
1543 addReply(c,shared.ok);
1544 }
1545
1546 /* ============================ Maxmemory directive ======================== */
1547
1548 /* This function gets called when 'maxmemory' is set on the config file to limit
1549 * the max memory used by the server, and we are out of memory.
1550 * This function will try to, in order:
1551 *
1552 * - Free objects from the free list
1553 * - Try to remove keys with an EXPIRE set
1554 *
1555 * It is not possible to free enough memory to reach used-memory < maxmemory
1556 * the server will start refusing commands that will enlarge even more the
1557 * memory usage.
1558 */
1559 void freeMemoryIfNeeded(void) {
1560 /* Remove keys accordingly to the active policy as long as we are
1561 * over the memory limit. */
1562 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION) return;
1563
1564 while (server.maxmemory && zmalloc_used_memory() > server.maxmemory) {
1565 int j, k, freed = 0;
1566
1567 for (j = 0; j < server.dbnum; j++) {
1568 long bestval = 0; /* just to prevent warning */
1569 sds bestkey = NULL;
1570 struct dictEntry *de;
1571 redisDb *db = server.db+j;
1572 dict *dict;
1573
1574 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
1575 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
1576 {
1577 dict = server.db[j].dict;
1578 } else {
1579 dict = server.db[j].expires;
1580 }
1581 if (dictSize(dict) == 0) continue;
1582
1583 /* volatile-random and allkeys-random policy */
1584 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
1585 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
1586 {
1587 de = dictGetRandomKey(dict);
1588 bestkey = dictGetEntryKey(de);
1589 }
1590
1591 /* volatile-lru and allkeys-lru policy */
1592 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
1593 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
1594 {
1595 for (k = 0; k < server.maxmemory_samples; k++) {
1596 sds thiskey;
1597 long thisval;
1598 robj *o;
1599
1600 de = dictGetRandomKey(dict);
1601 thiskey = dictGetEntryKey(de);
1602 /* When policy is volatile-lru we need an additonal lookup
1603 * to locate the real key, as dict is set to db->expires. */
1604 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
1605 de = dictFind(db->dict, thiskey);
1606 o = dictGetEntryVal(de);
1607 thisval = estimateObjectIdleTime(o);
1608
1609 /* Higher idle time is better candidate for deletion */
1610 if (bestkey == NULL || thisval > bestval) {
1611 bestkey = thiskey;
1612 bestval = thisval;
1613 }
1614 }
1615 }
1616
1617 /* volatile-ttl */
1618 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
1619 for (k = 0; k < server.maxmemory_samples; k++) {
1620 sds thiskey;
1621 long thisval;
1622
1623 de = dictGetRandomKey(dict);
1624 thiskey = dictGetEntryKey(de);
1625 thisval = (long) dictGetEntryVal(de);
1626
1627 /* Expire sooner (minor expire unix timestamp) is better
1628 * candidate for deletion */
1629 if (bestkey == NULL || thisval < bestval) {
1630 bestkey = thiskey;
1631 bestval = thisval;
1632 }
1633 }
1634 }
1635
1636 /* Finally remove the selected key. */
1637 if (bestkey) {
1638 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
1639 propagateExpire(db,keyobj);
1640 dbDelete(db,keyobj);
1641 server.stat_evictedkeys++;
1642 decrRefCount(keyobj);
1643 freed++;
1644 }
1645 }
1646 if (!freed) return; /* nothing to free... */
1647 }
1648 }
1649
1650 /* =================================== Main! ================================ */
1651
1652 #ifdef __linux__
1653 int linuxOvercommitMemoryValue(void) {
1654 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
1655 char buf[64];
1656
1657 if (!fp) return -1;
1658 if (fgets(buf,64,fp) == NULL) {
1659 fclose(fp);
1660 return -1;
1661 }
1662 fclose(fp);
1663
1664 return atoi(buf);
1665 }
1666
1667 void linuxOvercommitMemoryWarning(void) {
1668 if (linuxOvercommitMemoryValue() == 0) {
1669 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.");
1670 }
1671 }
1672 #endif /* __linux__ */
1673
1674 void createPidFile(void) {
1675 /* Try to write the pid file in a best-effort way. */
1676 FILE *fp = fopen(server.pidfile,"w");
1677 if (fp) {
1678 fprintf(fp,"%d\n",(int)getpid());
1679 fclose(fp);
1680 }
1681 }
1682
1683 void daemonize(void) {
1684 int fd;
1685
1686 if (fork() != 0) exit(0); /* parent exits */
1687 setsid(); /* create a new session */
1688
1689 /* Every output goes to /dev/null. If Redis is daemonized but
1690 * the 'logfile' is set to 'stdout' in the configuration file
1691 * it will not log at all. */
1692 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
1693 dup2(fd, STDIN_FILENO);
1694 dup2(fd, STDOUT_FILENO);
1695 dup2(fd, STDERR_FILENO);
1696 if (fd > STDERR_FILENO) close(fd);
1697 }
1698 }
1699
1700 void version() {
1701 printf("Redis server version %s (%s:%d)\n", REDIS_VERSION,
1702 redisGitSHA1(), atoi(redisGitDirty()) > 0);
1703 exit(0);
1704 }
1705
1706 void usage() {
1707 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf]\n");
1708 fprintf(stderr," ./redis-server - (read config from stdin)\n");
1709 exit(1);
1710 }
1711
1712 void redisAsciiArt(void) {
1713 #include "asciilogo.h"
1714 char *buf = zmalloc(1024*16);
1715
1716 snprintf(buf,1024*16,ascii_logo,
1717 REDIS_VERSION,
1718 redisGitSHA1(),
1719 strtol(redisGitDirty(),NULL,10) > 0,
1720 (sizeof(long) == 8) ? "64" : "32",
1721 server.cluster_enabled ? "cluster" : "stand alone",
1722 server.port,
1723 (long) getpid()
1724 );
1725 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
1726 zfree(buf);
1727 }
1728
1729 int main(int argc, char **argv) {
1730 long long start;
1731
1732 zmalloc_enable_thread_safeness();
1733 initServerConfig();
1734 if (argc == 2) {
1735 if (strcmp(argv[1], "-v") == 0 ||
1736 strcmp(argv[1], "--version") == 0) version();
1737 if (strcmp(argv[1], "--help") == 0) usage();
1738 resetServerSaveParams();
1739 loadServerConfig(argv[1]);
1740 } else if ((argc > 2)) {
1741 usage();
1742 } else {
1743 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'");
1744 }
1745 if (server.daemonize) daemonize();
1746 initServer();
1747 if (server.daemonize) createPidFile();
1748 redisAsciiArt();
1749 redisLog(REDIS_NOTICE,"Server started, Redis version " REDIS_VERSION);
1750 #ifdef __linux__
1751 linuxOvercommitMemoryWarning();
1752 #endif
1753 start = ustime();
1754 if (server.appendonly) {
1755 if (loadAppendOnlyFile(server.appendfilename) == REDIS_OK)
1756 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
1757 } else {
1758 if (rdbLoad(server.dbfilename) == REDIS_OK)
1759 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",(float)(ustime()-start)/1000000);
1760 }
1761 if (server.ipfd > 0)
1762 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
1763 if (server.sofd > 0)
1764 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
1765 aeSetBeforeSleepProc(server.el,beforeSleep);
1766 aeMain(server.el);
1767 aeDeleteEventLoop(server.el);
1768 return 0;
1769 }
1770
1771 #ifdef HAVE_BACKTRACE
1772 static void *getMcontextEip(ucontext_t *uc) {
1773 #if defined(__FreeBSD__)
1774 return (void*) uc->uc_mcontext.mc_eip;
1775 #elif defined(__dietlibc__)
1776 return (void*) uc->uc_mcontext.eip;
1777 #elif defined(__APPLE__) && !defined(MAC_OS_X_VERSION_10_6)
1778 #if __x86_64__
1779 return (void*) uc->uc_mcontext->__ss.__rip;
1780 #else
1781 return (void*) uc->uc_mcontext->__ss.__eip;
1782 #endif
1783 #elif defined(__APPLE__) && defined(MAC_OS_X_VERSION_10_6)
1784 #if defined(_STRUCT_X86_THREAD_STATE64) && !defined(__i386__)
1785 return (void*) uc->uc_mcontext->__ss.__rip;
1786 #else
1787 return (void*) uc->uc_mcontext->__ss.__eip;
1788 #endif
1789 #elif defined(__i386__)
1790 return (void*) uc->uc_mcontext.gregs[14]; /* Linux 32 */
1791 #elif defined(__X86_64__) || defined(__x86_64__)
1792 return (void*) uc->uc_mcontext.gregs[16]; /* Linux 64 */
1793 #elif defined(__ia64__) /* Linux IA64 */
1794 return (void*) uc->uc_mcontext.sc_ip;
1795 #else
1796 return NULL;
1797 #endif
1798 }
1799
1800 static void sigsegvHandler(int sig, siginfo_t *info, void *secret) {
1801 void *trace[100];
1802 char **messages = NULL;
1803 int i, trace_size = 0;
1804 ucontext_t *uc = (ucontext_t*) secret;
1805 sds infostring;
1806 struct sigaction act;
1807 REDIS_NOTUSED(info);
1808
1809 redisLog(REDIS_WARNING,
1810 "======= Ooops! Redis %s got signal: -%d- =======", REDIS_VERSION, sig);
1811 infostring = genRedisInfoString("all");
1812 redisLogRaw(REDIS_WARNING, infostring);
1813 /* It's not safe to sdsfree() the returned string under memory
1814 * corruption conditions. Let it leak as we are going to abort */
1815
1816 trace_size = backtrace(trace, 100);
1817 /* overwrite sigaction with caller's address */
1818 if (getMcontextEip(uc) != NULL) {
1819 trace[1] = getMcontextEip(uc);
1820 }
1821 messages = backtrace_symbols(trace, trace_size);
1822
1823 for (i=1; i<trace_size; ++i)
1824 redisLog(REDIS_WARNING,"%s", messages[i]);
1825
1826 /* free(messages); Don't call free() with possibly corrupted memory. */
1827 if (server.daemonize) unlink(server.pidfile);
1828
1829 /* Make sure we exit with the right signal at the end. So for instance
1830 * the core will be dumped if enabled. */
1831 sigemptyset (&act.sa_mask);
1832 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction
1833 * is used. Otherwise, sa_handler is used */
1834 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
1835 act.sa_handler = SIG_DFL;
1836 sigaction (sig, &act, NULL);
1837 kill(getpid(),sig);
1838 }
1839 #endif /* HAVE_BACKTRACE */
1840
1841 static void sigtermHandler(int sig) {
1842 REDIS_NOTUSED(sig);
1843
1844 redisLog(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
1845 server.shutdown_asap = 1;
1846 }
1847
1848 void setupSignalHandlers(void) {
1849 struct sigaction act;
1850
1851 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
1852 * Otherwise, sa_handler is used. */
1853 sigemptyset(&act.sa_mask);
1854 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
1855 act.sa_handler = sigtermHandler;
1856 sigaction(SIGTERM, &act, NULL);
1857
1858 #ifdef HAVE_BACKTRACE
1859 sigemptyset(&act.sa_mask);
1860 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
1861 act.sa_sigaction = sigsegvHandler;
1862 sigaction(SIGSEGV, &act, NULL);
1863 sigaction(SIGBUS, &act, NULL);
1864 sigaction(SIGFPE, &act, NULL);
1865 sigaction(SIGILL, &act, NULL);
1866 #endif
1867 return;
1868 }
1869
1870 /* The End */