]> git.saurik.com Git - redis.git/blob - src/redis.c
SLAVEOF is not a write command.
[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 #include <time.h>
35 #include <signal.h>
36 #include <sys/wait.h>
37 #include <errno.h>
38 #include <assert.h>
39 #include <ctype.h>
40 #include <stdarg.h>
41 #include <arpa/inet.h>
42 #include <sys/stat.h>
43 #include <fcntl.h>
44 #include <sys/time.h>
45 #include <sys/resource.h>
46 #include <sys/uio.h>
47 #include <limits.h>
48 #include <float.h>
49 #include <math.h>
50 #include <sys/resource.h>
51
52 /* Our shared "common" objects */
53
54 struct sharedObjectsStruct shared;
55
56 /* Global vars that are actually used as constants. The following double
57 * values are used for double on-disk serialization, and are initialized
58 * at runtime to avoid strange compiler optimizations. */
59
60 double R_Zero, R_PosInf, R_NegInf, R_Nan;
61
62 /*================================= Globals ================================= */
63
64 /* Global vars */
65 struct redisServer server; /* server global state */
66 struct redisCommand *commandTable;
67
68 /* Our command table.
69 *
70 * Every entry is composed of the following fields:
71 *
72 * name: a string representing the command name.
73 * function: pointer to the C function implementing the command.
74 * arity: number of arguments, it is possible to use -N to say >= N
75 * sflags: command flags as string. See below for a table of flags.
76 * flags: flags as bitmask. Computed by Redis using the 'sflags' field.
77 * get_keys_proc: an optional function to get key arguments from a command.
78 * This is only used when the following three fields are not
79 * enough to specify what arguments are keys.
80 * first_key_index: first argument that is a key
81 * last_key_index: last argument that is a key
82 * key_step: step to get all the keys from first to last argument. For instance
83 * in MSET the step is two since arguments are key,val,key,val,...
84 * microseconds: microseconds of total execution time for this command.
85 * calls: total number of calls of this command.
86 *
87 * The flags, microseconds and calls fields are computed by Redis and should
88 * always be set to zero.
89 *
90 * Command flags are expressed using strings where every character represents
91 * a flag. Later the populateCommandTable() function will take care of
92 * populating the real 'flags' field using this characters.
93 *
94 * This is the meaning of the flags:
95 *
96 * w: write command (may modify the key space).
97 * r: read command (will never modify the key space).
98 * m: may increase memory usage once called. Don't allow if out of memory.
99 * a: admin command, like SAVE or SHUTDOWN.
100 * p: Pub/Sub related command.
101 * f: force replication of this command, regarless of server.dirty.
102 * s: command not allowed in scripts.
103 * R: random command. Command is not deterministic, that is, the same command
104 * with the same arguments, with the same key space, may have different
105 * results. For instance SPOP and RANDOMKEY are two random commands.
106 * S: Sort command output array if called from script, so that the output
107 * is deterministic.
108 */
109 struct redisCommand redisCommandTable[] = {
110 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
111 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
112 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
113 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
114 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
115 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
116 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
117 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
118 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
119 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
120 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
121 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
122 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
123 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
124 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
125 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
126 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
127 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
128 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
129 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
130 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
131 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
132 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
133 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
134 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
135 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
136 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
137 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
138 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
139 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
140 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
141 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
142 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
143 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
144 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
145 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
146 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
147 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
148 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
149 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
150 {"srandmember",srandmemberCommand,2,"rR",0,NULL,1,1,1,0,0},
151 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
152 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
153 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
154 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
155 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
156 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
157 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
158 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
159 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
160 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
161 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
162 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
163 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
164 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
165 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
166 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
167 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
168 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
169 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
170 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
171 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
172 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
173 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
174 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
175 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
176 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
177 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
178 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
179 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
180 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
182 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
183 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
184 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
185 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
186 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
187 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
188 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
189 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
190 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
191 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
192 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
193 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
194 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
195 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
196 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
197 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
198 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
199 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
200 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
201 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
202 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
203 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
204 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
205 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
206 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
207 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
208 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
209 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
210 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
211 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
212 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
213 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
214 {"exec",execCommand,1,"s",0,NULL,0,0,0,0,0},
215 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
216 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
217 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
218 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
219 {"sort",sortCommand,-2,"wmS",0,NULL,1,1,1,0,0},
220 {"info",infoCommand,-1,"r",0,NULL,0,0,0,0,0},
221 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
222 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
223 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
224 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
225 {"slaveof",slaveofCommand,3,"as",0,NULL,0,0,0,0,0},
226 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
227 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
228 {"subscribe",subscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
229 {"unsubscribe",unsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
230 {"psubscribe",psubscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
231 {"punsubscribe",punsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
232 {"publish",publishCommand,3,"pf",0,NULL,0,0,0,0,0},
233 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
234 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
235 {"cluster",clusterCommand,-2,"ar",0,NULL,0,0,0,0,0},
236 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
237 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
238 {"asking",askingCommand,1,"r",0,NULL,0,0,0,0,0},
239 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
240 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
241 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
242 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
243 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
244 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
245 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
246 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0}
247 };
248
249 /*============================ Utility functions ============================ */
250
251 /* Low level logging. To use only for very big messages, otherwise
252 * redisLog() is to prefer. */
253 void redisLogRaw(int level, const char *msg) {
254 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
255 const char *c = ".-*#";
256 FILE *fp;
257 char buf[64];
258 int rawmode = (level & REDIS_LOG_RAW);
259
260 level &= 0xff; /* clear flags */
261 if (level < server.verbosity) return;
262
263 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
264 if (!fp) return;
265
266 if (rawmode) {
267 fprintf(fp,"%s",msg);
268 } else {
269 int off;
270 struct timeval tv;
271
272 gettimeofday(&tv,NULL);
273 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
274 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
275 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
276 }
277 fflush(fp);
278
279 if (server.logfile) fclose(fp);
280
281 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
282 }
283
284 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
285 * is used across the code. The raw version is only used in order to dump
286 * the INFO output on crash. */
287 void redisLog(int level, const char *fmt, ...) {
288 va_list ap;
289 char msg[REDIS_MAX_LOGMSG_LEN];
290
291 if ((level&0xff) < server.verbosity) return;
292
293 va_start(ap, fmt);
294 vsnprintf(msg, sizeof(msg), fmt, ap);
295 va_end(ap);
296
297 redisLogRaw(level,msg);
298 }
299
300 /* Log a fixed message without printf-alike capabilities, in a way that is
301 * safe to call from a signal handler.
302 *
303 * We actually use this only for signals that are not fatal from the point
304 * of view of Redis. Signals that are going to kill the server anyway and
305 * where we need printf-alike features are served by redisLog(). */
306 void redisLogFromHandler(int level, const char *msg) {
307 int fd;
308 char buf[64];
309
310 if ((level&0xff) < server.verbosity ||
311 (server.logfile == NULL && server.daemonize)) return;
312 fd = server.logfile ?
313 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
314 STDOUT_FILENO;
315 if (fd == -1) return;
316 ll2string(buf,sizeof(buf),getpid());
317 write(fd,"[",1);
318 write(fd,buf,strlen(buf));
319 write(fd," | signal handler] (",20);
320 ll2string(buf,sizeof(buf),time(NULL));
321 write(fd,buf,strlen(buf));
322 write(fd,") ",2);
323 write(fd,msg,strlen(msg));
324 write(fd,"\n",1);
325 if (server.logfile) close(fd);
326 }
327
328 /* Redis generally does not try to recover from out of memory conditions
329 * when allocating objects or strings, it is not clear if it will be possible
330 * to report this condition to the client since the networking layer itself
331 * is based on heap allocation for send buffers, so we simply abort.
332 * At least the code will be simpler to read... */
333 void oom(const char *msg) {
334 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
335 sleep(1);
336 abort();
337 }
338
339 /* Return the UNIX time in microseconds */
340 long long ustime(void) {
341 struct timeval tv;
342 long long ust;
343
344 gettimeofday(&tv, NULL);
345 ust = ((long long)tv.tv_sec)*1000000;
346 ust += tv.tv_usec;
347 return ust;
348 }
349
350 /* Return the UNIX time in milliseconds */
351 long long mstime(void) {
352 return ustime()/1000;
353 }
354
355 /*====================== Hash table type implementation ==================== */
356
357 /* This is an hash table type that uses the SDS dynamic strings libary as
358 * keys and radis objects as values (objects can hold SDS strings,
359 * lists, sets). */
360
361 void dictVanillaFree(void *privdata, void *val)
362 {
363 DICT_NOTUSED(privdata);
364 zfree(val);
365 }
366
367 void dictListDestructor(void *privdata, void *val)
368 {
369 DICT_NOTUSED(privdata);
370 listRelease((list*)val);
371 }
372
373 int dictSdsKeyCompare(void *privdata, const void *key1,
374 const void *key2)
375 {
376 int l1,l2;
377 DICT_NOTUSED(privdata);
378
379 l1 = sdslen((sds)key1);
380 l2 = sdslen((sds)key2);
381 if (l1 != l2) return 0;
382 return memcmp(key1, key2, l1) == 0;
383 }
384
385 /* A case insensitive version used for the command lookup table. */
386 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
387 const void *key2)
388 {
389 DICT_NOTUSED(privdata);
390
391 return strcasecmp(key1, key2) == 0;
392 }
393
394 void dictRedisObjectDestructor(void *privdata, void *val)
395 {
396 DICT_NOTUSED(privdata);
397
398 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
399 decrRefCount(val);
400 }
401
402 void dictSdsDestructor(void *privdata, void *val)
403 {
404 DICT_NOTUSED(privdata);
405
406 sdsfree(val);
407 }
408
409 int dictObjKeyCompare(void *privdata, const void *key1,
410 const void *key2)
411 {
412 const robj *o1 = key1, *o2 = key2;
413 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
414 }
415
416 unsigned int dictObjHash(const void *key) {
417 const robj *o = key;
418 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
419 }
420
421 unsigned int dictSdsHash(const void *key) {
422 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
423 }
424
425 unsigned int dictSdsCaseHash(const void *key) {
426 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
427 }
428
429 int dictEncObjKeyCompare(void *privdata, const void *key1,
430 const void *key2)
431 {
432 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
433 int cmp;
434
435 if (o1->encoding == REDIS_ENCODING_INT &&
436 o2->encoding == REDIS_ENCODING_INT)
437 return o1->ptr == o2->ptr;
438
439 o1 = getDecodedObject(o1);
440 o2 = getDecodedObject(o2);
441 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
442 decrRefCount(o1);
443 decrRefCount(o2);
444 return cmp;
445 }
446
447 unsigned int dictEncObjHash(const void *key) {
448 robj *o = (robj*) key;
449
450 if (o->encoding == REDIS_ENCODING_RAW) {
451 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
452 } else {
453 if (o->encoding == REDIS_ENCODING_INT) {
454 char buf[32];
455 int len;
456
457 len = ll2string(buf,32,(long)o->ptr);
458 return dictGenHashFunction((unsigned char*)buf, len);
459 } else {
460 unsigned int hash;
461
462 o = getDecodedObject(o);
463 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
464 decrRefCount(o);
465 return hash;
466 }
467 }
468 }
469
470 /* Sets type hash table */
471 dictType setDictType = {
472 dictEncObjHash, /* hash function */
473 NULL, /* key dup */
474 NULL, /* val dup */
475 dictEncObjKeyCompare, /* key compare */
476 dictRedisObjectDestructor, /* key destructor */
477 NULL /* val destructor */
478 };
479
480 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
481 dictType zsetDictType = {
482 dictEncObjHash, /* hash function */
483 NULL, /* key dup */
484 NULL, /* val dup */
485 dictEncObjKeyCompare, /* key compare */
486 dictRedisObjectDestructor, /* key destructor */
487 NULL /* val destructor */
488 };
489
490 /* Db->dict, keys are sds strings, vals are Redis objects. */
491 dictType dbDictType = {
492 dictSdsHash, /* hash function */
493 NULL, /* key dup */
494 NULL, /* val dup */
495 dictSdsKeyCompare, /* key compare */
496 dictSdsDestructor, /* key destructor */
497 dictRedisObjectDestructor /* val destructor */
498 };
499
500 /* Db->expires */
501 dictType keyptrDictType = {
502 dictSdsHash, /* hash function */
503 NULL, /* key dup */
504 NULL, /* val dup */
505 dictSdsKeyCompare, /* key compare */
506 NULL, /* key destructor */
507 NULL /* val destructor */
508 };
509
510 /* Command table. sds string -> command struct pointer. */
511 dictType commandTableDictType = {
512 dictSdsCaseHash, /* hash function */
513 NULL, /* key dup */
514 NULL, /* val dup */
515 dictSdsKeyCaseCompare, /* key compare */
516 dictSdsDestructor, /* key destructor */
517 NULL /* val destructor */
518 };
519
520 /* Hash type hash table (note that small hashes are represented with zimpaps) */
521 dictType hashDictType = {
522 dictEncObjHash, /* hash function */
523 NULL, /* key dup */
524 NULL, /* val dup */
525 dictEncObjKeyCompare, /* key compare */
526 dictRedisObjectDestructor, /* key destructor */
527 dictRedisObjectDestructor /* val destructor */
528 };
529
530 /* Keylist hash table type has unencoded redis objects as keys and
531 * lists as values. It's used for blocking operations (BLPOP) and to
532 * map swapped keys to a list of clients waiting for this keys to be loaded. */
533 dictType keylistDictType = {
534 dictObjHash, /* hash function */
535 NULL, /* key dup */
536 NULL, /* val dup */
537 dictObjKeyCompare, /* key compare */
538 dictRedisObjectDestructor, /* key destructor */
539 dictListDestructor /* val destructor */
540 };
541
542 /* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
543 * clusterNode structures. */
544 dictType clusterNodesDictType = {
545 dictSdsHash, /* hash function */
546 NULL, /* key dup */
547 NULL, /* val dup */
548 dictSdsKeyCompare, /* key compare */
549 dictSdsDestructor, /* key destructor */
550 NULL /* val destructor */
551 };
552
553 int htNeedsResize(dict *dict) {
554 long long size, used;
555
556 size = dictSlots(dict);
557 used = dictSize(dict);
558 return (size && used && size > DICT_HT_INITIAL_SIZE &&
559 (used*100/size < REDIS_HT_MINFILL));
560 }
561
562 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
563 * we resize the hash table to save memory */
564 void tryResizeHashTables(void) {
565 int j;
566
567 for (j = 0; j < server.dbnum; j++) {
568 if (htNeedsResize(server.db[j].dict))
569 dictResize(server.db[j].dict);
570 if (htNeedsResize(server.db[j].expires))
571 dictResize(server.db[j].expires);
572 }
573 }
574
575 /* Our hash table implementation performs rehashing incrementally while
576 * we write/read from the hash table. Still if the server is idle, the hash
577 * table will use two tables for a long time. So we try to use 1 millisecond
578 * of CPU time at every serverCron() loop in order to rehash some key. */
579 void incrementallyRehash(void) {
580 int j;
581
582 for (j = 0; j < server.dbnum; j++) {
583 if (dictIsRehashing(server.db[j].dict)) {
584 dictRehashMilliseconds(server.db[j].dict,1);
585 break; /* already used our millisecond for this loop... */
586 }
587 }
588 }
589
590 /* This function is called once a background process of some kind terminates,
591 * as we want to avoid resizing the hash tables when there is a child in order
592 * to play well with copy-on-write (otherwise when a resize happens lots of
593 * memory pages are copied). The goal of this function is to update the ability
594 * for dict.c to resize the hash tables accordingly to the fact we have o not
595 * running childs. */
596 void updateDictResizePolicy(void) {
597 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
598 dictEnableResize();
599 else
600 dictDisableResize();
601 }
602
603 /* ======================= Cron: called every 100 ms ======================== */
604
605 /* Try to expire a few timed out keys. The algorithm used is adaptive and
606 * will use few CPU cycles if there are few expiring keys, otherwise
607 * it will get more aggressive to avoid that too much memory is used by
608 * keys that can be removed from the keyspace. */
609 void activeExpireCycle(void) {
610 int j;
611
612 for (j = 0; j < server.dbnum; j++) {
613 int expired;
614 redisDb *db = server.db+j;
615
616 /* Continue to expire if at the end of the cycle more than 25%
617 * of the keys were expired. */
618 do {
619 long num = dictSize(db->expires);
620 long long now = mstime();
621
622 expired = 0;
623 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
624 num = REDIS_EXPIRELOOKUPS_PER_CRON;
625 while (num--) {
626 dictEntry *de;
627 long long t;
628
629 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
630 t = dictGetSignedIntegerVal(de);
631 if (now > t) {
632 sds key = dictGetKey(de);
633 robj *keyobj = createStringObject(key,sdslen(key));
634
635 propagateExpire(db,keyobj);
636 dbDelete(db,keyobj);
637 decrRefCount(keyobj);
638 expired++;
639 server.stat_expiredkeys++;
640 }
641 }
642 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
643 }
644 }
645
646 void updateLRUClock(void) {
647 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
648 REDIS_LRU_CLOCK_MAX;
649 }
650
651
652 /* Add a sample to the operations per second array of samples. */
653 void trackOperationsPerSecond(void) {
654 long long t = mstime() - server.ops_sec_last_sample_time;
655 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
656 long long ops_sec;
657
658 ops_sec = t > 0 ? (ops*1000/t) : 0;
659
660 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
661 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
662 server.ops_sec_last_sample_time = mstime();
663 server.ops_sec_last_sample_ops = server.stat_numcommands;
664 }
665
666 /* Return the mean of all the samples. */
667 long long getOperationsPerSecond(void) {
668 int j;
669 long long sum = 0;
670
671 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
672 sum += server.ops_sec_samples[j];
673 return sum / REDIS_OPS_SEC_SAMPLES;
674 }
675
676 /* Check for timeouts. Returns non-zero if the client was terminated */
677 int clientsCronHandleTimeout(redisClient *c) {
678 time_t now = server.unixtime;
679
680 if (server.maxidletime &&
681 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
682 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
683 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
684 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
685 listLength(c->pubsub_patterns) == 0 &&
686 (now - c->lastinteraction > server.maxidletime))
687 {
688 redisLog(REDIS_VERBOSE,"Closing idle client");
689 freeClient(c);
690 return 1;
691 } else if (c->flags & REDIS_BLOCKED) {
692 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
693 addReply(c,shared.nullmultibulk);
694 unblockClientWaitingData(c);
695 }
696 }
697 return 0;
698 }
699
700 /* The client query buffer is an sds.c string that can end with a lot of
701 * free space not used, this function reclaims space if needed.
702 *
703 * The funciton always returns 0 as it never terminates the client. */
704 int clientsCronResizeQueryBuffer(redisClient *c) {
705 size_t querybuf_size = sdsAllocSize(c->querybuf);
706 time_t idletime = server.unixtime - c->lastinteraction;
707
708 /* There are two conditions to resize the query buffer:
709 * 1) Query buffer is > BIG_ARG and too big for latest peak.
710 * 2) Client is inactive and the buffer is bigger than 1k. */
711 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
712 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
713 (querybuf_size > 1024 && idletime > 2))
714 {
715 /* Only resize the query buffer if it is actually wasting space. */
716 if (sdsavail(c->querybuf) > 1024) {
717 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
718 }
719 }
720 /* Reset the peak again to capture the peak memory usage in the next
721 * cycle. */
722 c->querybuf_peak = 0;
723 return 0;
724 }
725
726 void clientsCron(void) {
727 /* Make sure to process at least 1/100 of clients per call.
728 * Since this function is called 10 times per second we are sure that
729 * in the worst case we process all the clients in 10 seconds.
730 * In normal conditions (a reasonable number of clients) we process
731 * all the clients in a shorter time. */
732 int numclients = listLength(server.clients);
733 int iterations = numclients/100;
734
735 if (iterations < 50)
736 iterations = (numclients < 50) ? numclients : 50;
737 while(listLength(server.clients) && iterations--) {
738 redisClient *c;
739 listNode *head;
740
741 /* Rotate the list, take the current head, process.
742 * This way if the client must be removed from the list it's the
743 * first element and we don't incur into O(N) computation. */
744 listRotate(server.clients);
745 head = listFirst(server.clients);
746 c = listNodeValue(head);
747 /* The following functions do different service checks on the client.
748 * The protocol is that they return non-zero if the client was
749 * terminated. */
750 if (clientsCronHandleTimeout(c)) continue;
751 if (clientsCronResizeQueryBuffer(c)) continue;
752 }
753 }
754
755 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
756 int j, loops = server.cronloops;
757 REDIS_NOTUSED(eventLoop);
758 REDIS_NOTUSED(id);
759 REDIS_NOTUSED(clientData);
760
761 /* Software watchdog: deliver the SIGALRM that will reach the signal
762 * handler if we don't return here fast enough. */
763 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
764
765 /* We take a cached value of the unix time in the global state because
766 * with virtual memory and aging there is to store the current time
767 * in objects at every object access, and accuracy is not needed.
768 * To access a global var is faster than calling time(NULL) */
769 server.unixtime = time(NULL);
770
771 trackOperationsPerSecond();
772
773 /* We have just 22 bits per object for LRU information.
774 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
775 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
776 *
777 * Note that even if this will wrap after 1.5 years it's not a problem,
778 * everything will still work but just some object will appear younger
779 * to Redis. But for this to happen a given object should never be touched
780 * for 1.5 years.
781 *
782 * Note that you can change the resolution altering the
783 * REDIS_LRU_CLOCK_RESOLUTION define.
784 */
785 updateLRUClock();
786
787 /* Record the max memory used since the server was started. */
788 if (zmalloc_used_memory() > server.stat_peak_memory)
789 server.stat_peak_memory = zmalloc_used_memory();
790
791 /* We received a SIGTERM, shutting down here in a safe way, as it is
792 * not ok doing so inside the signal handler. */
793 if (server.shutdown_asap) {
794 if (prepareForShutdown(0) == REDIS_OK) exit(0);
795 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
796 }
797
798 /* Show some info about non-empty databases */
799 for (j = 0; j < server.dbnum; j++) {
800 long long size, used, vkeys;
801
802 size = dictSlots(server.db[j].dict);
803 used = dictSize(server.db[j].dict);
804 vkeys = dictSize(server.db[j].expires);
805 if (!(loops % 50) && (used || vkeys)) {
806 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
807 /* dictPrintStats(server.dict); */
808 }
809 }
810
811 /* We don't want to resize the hash tables while a bacground saving
812 * is in progress: the saving child is created using fork() that is
813 * implemented with a copy-on-write semantic in most modern systems, so
814 * if we resize the HT while there is the saving child at work actually
815 * a lot of memory movements in the parent will cause a lot of pages
816 * copied. */
817 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
818 if (!(loops % 10)) tryResizeHashTables();
819 if (server.activerehashing) incrementallyRehash();
820 }
821
822 /* Show information about connected clients */
823 if (!(loops % 50)) {
824 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
825 listLength(server.clients)-listLength(server.slaves),
826 listLength(server.slaves),
827 zmalloc_used_memory());
828 }
829
830 /* We need to do a few operations on clients asynchronously. */
831 clientsCron();
832
833 /* Start a scheduled AOF rewrite if this was requested by the user while
834 * a BGSAVE was in progress. */
835 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
836 server.aof_rewrite_scheduled)
837 {
838 rewriteAppendOnlyFileBackground();
839 }
840
841 /* Check if a background saving or AOF rewrite in progress terminated. */
842 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
843 int statloc;
844 pid_t pid;
845
846 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
847 int exitcode = WEXITSTATUS(statloc);
848 int bysignal = 0;
849
850 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
851
852 if (pid == server.rdb_child_pid) {
853 backgroundSaveDoneHandler(exitcode,bysignal);
854 } else {
855 backgroundRewriteDoneHandler(exitcode,bysignal);
856 }
857 updateDictResizePolicy();
858 }
859 } else {
860 /* If there is not a background saving/rewrite in progress check if
861 * we have to save/rewrite now */
862 for (j = 0; j < server.saveparamslen; j++) {
863 struct saveparam *sp = server.saveparams+j;
864
865 if (server.dirty >= sp->changes &&
866 server.unixtime-server.lastsave > sp->seconds) {
867 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
868 sp->changes, sp->seconds);
869 rdbSaveBackground(server.rdb_filename);
870 break;
871 }
872 }
873
874 /* Trigger an AOF rewrite if needed */
875 if (server.rdb_child_pid == -1 &&
876 server.aof_child_pid == -1 &&
877 server.aof_rewrite_perc &&
878 server.aof_current_size > server.aof_rewrite_min_size)
879 {
880 long long base = server.aof_rewrite_base_size ?
881 server.aof_rewrite_base_size : 1;
882 long long growth = (server.aof_current_size*100/base) - 100;
883 if (growth >= server.aof_rewrite_perc) {
884 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
885 rewriteAppendOnlyFileBackground();
886 }
887 }
888 }
889
890
891 /* If we postponed an AOF buffer flush, let's try to do it every time the
892 * cron function is called. */
893 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
894
895 /* Expire a few keys per cycle, only if this is a master.
896 * On slaves we wait for DEL operations synthesized by the master
897 * in order to guarantee a strict consistency. */
898 if (server.masterhost == NULL) activeExpireCycle();
899
900 /* Close clients that need to be closed asynchronous */
901 freeClientsInAsyncFreeQueue();
902
903 /* Replication cron function -- used to reconnect to master and
904 * to detect transfer failures. */
905 if (!(loops % 10)) replicationCron();
906
907 /* Run other sub-systems specific cron jobs */
908 if (server.cluster_enabled && !(loops % 10)) clusterCron();
909
910 server.cronloops++;
911 return 100;
912 }
913
914 /* This function gets called every time Redis is entering the
915 * main loop of the event driven library, that is, before to sleep
916 * for ready file descriptors. */
917 void beforeSleep(struct aeEventLoop *eventLoop) {
918 REDIS_NOTUSED(eventLoop);
919 listNode *ln;
920 redisClient *c;
921
922 /* Try to process pending commands for clients that were just unblocked. */
923 while (listLength(server.unblocked_clients)) {
924 ln = listFirst(server.unblocked_clients);
925 redisAssert(ln != NULL);
926 c = ln->value;
927 listDelNode(server.unblocked_clients,ln);
928 c->flags &= ~REDIS_UNBLOCKED;
929
930 /* Process remaining data in the input buffer. */
931 if (c->querybuf && sdslen(c->querybuf) > 0) {
932 server.current_client = c;
933 processInputBuffer(c);
934 server.current_client = NULL;
935 }
936 }
937
938 /* Write the AOF buffer on disk */
939 flushAppendOnlyFile(0);
940 }
941
942 /* =========================== Server initialization ======================== */
943
944 void createSharedObjects(void) {
945 int j;
946
947 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
948 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
949 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
950 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
951 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
952 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
953 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
954 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
955 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
956 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
957 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
958 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
959 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
960 "-ERR Operation against a key holding the wrong kind of value\r\n"));
961 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
962 "-ERR no such key\r\n"));
963 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
964 "-ERR syntax error\r\n"));
965 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
966 "-ERR source and destination objects are the same\r\n"));
967 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
968 "-ERR index out of range\r\n"));
969 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
970 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
971 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
972 "-LOADING Redis is loading the dataset in memory\r\n"));
973 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
974 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
975 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
976 "-MISCONF Redis is configured to save RDB snapshots, but is currently not able to persist on disk. Commands that may modify the data set are disabled. Please check Redis logs for details about the error.\r\n"));
977 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
978 "-READONLY You can't write against a read only slave.\r\n"));
979 shared.oomerr = createObject(REDIS_STRING,sdsnew(
980 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
981 shared.space = createObject(REDIS_STRING,sdsnew(" "));
982 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
983 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
984
985 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
986 shared.select[j] = createObject(REDIS_STRING,
987 sdscatprintf(sdsempty(),"select %d\r\n", j));
988 }
989 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
990 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
991 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
992 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
993 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
994 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
995 shared.del = createStringObject("DEL",3);
996 shared.rpop = createStringObject("RPOP",4);
997 shared.lpop = createStringObject("LPOP",4);
998 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
999 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1000 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1001 }
1002 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1003 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1004 sdscatprintf(sdsempty(),"*%d\r\n",j));
1005 shared.bulkhdr[j] = createObject(REDIS_STRING,
1006 sdscatprintf(sdsempty(),"$%d\r\n",j));
1007 }
1008 }
1009
1010 void initServerConfig() {
1011 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1012 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1013 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1014 server.port = REDIS_SERVERPORT;
1015 server.bindaddr = NULL;
1016 server.unixsocket = NULL;
1017 server.unixsocketperm = 0;
1018 server.ipfd = -1;
1019 server.sofd = -1;
1020 server.dbnum = REDIS_DEFAULT_DBNUM;
1021 server.verbosity = REDIS_NOTICE;
1022 server.maxidletime = REDIS_MAXIDLETIME;
1023 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1024 server.saveparams = NULL;
1025 server.loading = 0;
1026 server.logfile = NULL; /* NULL = log on standard output */
1027 server.syslog_enabled = 0;
1028 server.syslog_ident = zstrdup("redis");
1029 server.syslog_facility = LOG_LOCAL0;
1030 server.daemonize = 0;
1031 server.aof_state = REDIS_AOF_OFF;
1032 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1033 server.aof_no_fsync_on_rewrite = 0;
1034 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1035 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1036 server.aof_rewrite_base_size = 0;
1037 server.aof_rewrite_scheduled = 0;
1038 server.aof_last_fsync = time(NULL);
1039 server.aof_delayed_fsync = 0;
1040 server.aof_fd = -1;
1041 server.aof_selected_db = -1; /* Make sure the first time will not match */
1042 server.aof_flush_postponed_start = 0;
1043 server.pidfile = zstrdup("/var/run/redis.pid");
1044 server.rdb_filename = zstrdup("dump.rdb");
1045 server.aof_filename = zstrdup("appendonly.aof");
1046 server.requirepass = NULL;
1047 server.rdb_compression = 1;
1048 server.activerehashing = 1;
1049 server.maxclients = REDIS_MAX_CLIENTS;
1050 server.bpop_blocked_clients = 0;
1051 server.maxmemory = 0;
1052 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1053 server.maxmemory_samples = 3;
1054 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1055 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1056 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1057 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1058 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1059 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1060 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1061 server.shutdown_asap = 0;
1062 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1063 server.repl_timeout = REDIS_REPL_TIMEOUT;
1064 server.cluster_enabled = 0;
1065 server.cluster.configfile = zstrdup("nodes.conf");
1066 server.lua_caller = NULL;
1067 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1068 server.lua_client = NULL;
1069 server.lua_timedout = 0;
1070
1071 updateLRUClock();
1072 resetServerSaveParams();
1073
1074 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1075 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1076 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1077 /* Replication related */
1078 server.masterauth = NULL;
1079 server.masterhost = NULL;
1080 server.masterport = 6379;
1081 server.master = NULL;
1082 server.repl_state = REDIS_REPL_NONE;
1083 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1084 server.repl_serve_stale_data = 1;
1085 server.repl_slave_ro = 1;
1086 server.repl_down_since = -1;
1087
1088 /* Client output buffer limits */
1089 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1090 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1091 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1092 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1093 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1094 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1095 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1096 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1097 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1098
1099 /* Double constants initialization */
1100 R_Zero = 0.0;
1101 R_PosInf = 1.0/R_Zero;
1102 R_NegInf = -1.0/R_Zero;
1103 R_Nan = R_Zero/R_Zero;
1104
1105 /* Command table -- we intiialize it here as it is part of the
1106 * initial configuration, since command names may be changed via
1107 * redis.conf using the rename-command directive. */
1108 server.commands = dictCreate(&commandTableDictType,NULL);
1109 populateCommandTable();
1110 server.delCommand = lookupCommandByCString("del");
1111 server.multiCommand = lookupCommandByCString("multi");
1112 server.lpushCommand = lookupCommandByCString("lpush");
1113
1114 /* Slow log */
1115 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1116 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1117
1118 /* Debugging */
1119 server.assert_failed = "<no assertion failed>";
1120 server.assert_file = "<no file>";
1121 server.assert_line = 0;
1122 server.bug_report_start = 0;
1123 server.watchdog_period = 0;
1124 }
1125
1126 /* This function will try to raise the max number of open files accordingly to
1127 * the configured max number of clients. It will also account for 32 additional
1128 * file descriptors as we need a few more for persistence, listening
1129 * sockets, log files and so forth.
1130 *
1131 * If it will not be possible to set the limit accordingly to the configured
1132 * max number of clients, the function will do the reverse setting
1133 * server.maxclients to the value that we can actually handle. */
1134 void adjustOpenFilesLimit(void) {
1135 rlim_t maxfiles = server.maxclients+32;
1136 struct rlimit limit;
1137
1138 if (maxfiles < 1024) maxfiles = 1024;
1139 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1140 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1141 strerror(errno));
1142 server.maxclients = 1024-32;
1143 } else {
1144 rlim_t oldlimit = limit.rlim_cur;
1145
1146 /* Set the max number of files if the current limit is not enough
1147 * for our needs. */
1148 if (oldlimit < maxfiles) {
1149 rlim_t f;
1150
1151 f = maxfiles;
1152 while(f > oldlimit) {
1153 limit.rlim_cur = f;
1154 limit.rlim_max = f;
1155 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1156 f -= 128;
1157 }
1158 if (f < oldlimit) f = oldlimit;
1159 if (f != maxfiles) {
1160 server.maxclients = f-32;
1161 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1162 (int) maxfiles, strerror(errno), (int) server.maxclients);
1163 } else {
1164 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1165 (int) maxfiles);
1166 }
1167 }
1168 }
1169 }
1170
1171 void initServer() {
1172 int j;
1173
1174 signal(SIGHUP, SIG_IGN);
1175 signal(SIGPIPE, SIG_IGN);
1176 setupSignalHandlers();
1177
1178 if (server.syslog_enabled) {
1179 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1180 server.syslog_facility);
1181 }
1182
1183 server.current_client = NULL;
1184 server.clients = listCreate();
1185 server.clients_to_close = listCreate();
1186 server.slaves = listCreate();
1187 server.monitors = listCreate();
1188 server.unblocked_clients = listCreate();
1189
1190 createSharedObjects();
1191 adjustOpenFilesLimit();
1192 server.el = aeCreateEventLoop(server.maxclients+1024);
1193 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1194
1195 if (server.port != 0) {
1196 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1197 if (server.ipfd == ANET_ERR) {
1198 redisLog(REDIS_WARNING, "Opening port %d: %s",
1199 server.port, server.neterr);
1200 exit(1);
1201 }
1202 }
1203 if (server.unixsocket != NULL) {
1204 unlink(server.unixsocket); /* don't care if this fails */
1205 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1206 if (server.sofd == ANET_ERR) {
1207 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1208 exit(1);
1209 }
1210 }
1211 if (server.ipfd < 0 && server.sofd < 0) {
1212 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1213 exit(1);
1214 }
1215 for (j = 0; j < server.dbnum; j++) {
1216 server.db[j].dict = dictCreate(&dbDictType,NULL);
1217 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1218 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1219 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1220 server.db[j].id = j;
1221 }
1222 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1223 server.pubsub_patterns = listCreate();
1224 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1225 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1226 server.cronloops = 0;
1227 server.rdb_child_pid = -1;
1228 server.aof_child_pid = -1;
1229 server.aof_rewrite_buf = sdsempty();
1230 server.aof_buf = sdsempty();
1231 server.lastsave = time(NULL);
1232 server.dirty = 0;
1233 server.stat_numcommands = 0;
1234 server.stat_numconnections = 0;
1235 server.stat_expiredkeys = 0;
1236 server.stat_evictedkeys = 0;
1237 server.stat_starttime = time(NULL);
1238 server.stat_keyspace_misses = 0;
1239 server.stat_keyspace_hits = 0;
1240 server.stat_peak_memory = 0;
1241 server.stat_fork_time = 0;
1242 server.stat_rejected_conn = 0;
1243 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1244 server.ops_sec_idx = 0;
1245 server.ops_sec_last_sample_time = mstime();
1246 server.ops_sec_last_sample_ops = 0;
1247 server.unixtime = time(NULL);
1248 server.lastbgsave_status = REDIS_OK;
1249 server.stop_writes_on_bgsave_err = 1;
1250 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1251 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1252 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1253 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1254 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1255
1256 if (server.aof_state == REDIS_AOF_ON) {
1257 server.aof_fd = open(server.aof_filename,
1258 O_WRONLY|O_APPEND|O_CREAT,0644);
1259 if (server.aof_fd == -1) {
1260 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1261 strerror(errno));
1262 exit(1);
1263 }
1264 }
1265
1266 /* 32 bit instances are limited to 4GB of address space, so if there is
1267 * no explicit limit in the user provided configuration we set a limit
1268 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1269 * useless crashes of the Redis instance. */
1270 if (server.arch_bits == 32 && server.maxmemory == 0) {
1271 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1272 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1273 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1274 }
1275
1276 if (server.cluster_enabled) clusterInit();
1277 scriptingInit();
1278 slowlogInit();
1279 bioInit();
1280 }
1281
1282 /* Populates the Redis Command Table starting from the hard coded list
1283 * we have on top of redis.c file. */
1284 void populateCommandTable(void) {
1285 int j;
1286 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1287
1288 for (j = 0; j < numcommands; j++) {
1289 struct redisCommand *c = redisCommandTable+j;
1290 char *f = c->sflags;
1291 int retval;
1292
1293 while(*f != '\0') {
1294 switch(*f) {
1295 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1296 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1297 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1298 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1299 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1300 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1301 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1302 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1303 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1304 default: redisPanic("Unsupported command flag"); break;
1305 }
1306 f++;
1307 }
1308
1309 retval = dictAdd(server.commands, sdsnew(c->name), c);
1310 assert(retval == DICT_OK);
1311 }
1312 }
1313
1314 void resetCommandTableStats(void) {
1315 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1316 int j;
1317
1318 for (j = 0; j < numcommands; j++) {
1319 struct redisCommand *c = redisCommandTable+j;
1320
1321 c->microseconds = 0;
1322 c->calls = 0;
1323 }
1324 }
1325
1326 /* ========================== Redis OP Array API ============================ */
1327
1328 void redisOpArrayInit(redisOpArray *oa) {
1329 oa->ops = NULL;
1330 oa->numops = 0;
1331 }
1332
1333 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1334 robj **argv, int argc, int target)
1335 {
1336 redisOp *op;
1337
1338 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1339 op = oa->ops+oa->numops;
1340 op->cmd = cmd;
1341 op->dbid = dbid;
1342 op->argv = argv;
1343 op->argc = argc;
1344 op->target = target;
1345 oa->numops++;
1346 return oa->numops;
1347 }
1348
1349 void redisOpArrayFree(redisOpArray *oa) {
1350 while(oa->numops) {
1351 int j;
1352 redisOp *op;
1353
1354 oa->numops--;
1355 op = oa->ops+oa->numops;
1356 for (j = 0; j < op->argc; j++)
1357 decrRefCount(op->argv[j]);
1358 zfree(op->argv);
1359 }
1360 zfree(oa->ops);
1361 }
1362
1363 /* ====================== Commands lookup and execution ===================== */
1364
1365 struct redisCommand *lookupCommand(sds name) {
1366 return dictFetchValue(server.commands, name);
1367 }
1368
1369 struct redisCommand *lookupCommandByCString(char *s) {
1370 struct redisCommand *cmd;
1371 sds name = sdsnew(s);
1372
1373 cmd = dictFetchValue(server.commands, name);
1374 sdsfree(name);
1375 return cmd;
1376 }
1377
1378 /* Propagate the specified command (in the context of the specified database id)
1379 * to AOF, Slaves and Monitors.
1380 *
1381 * flags are an xor between:
1382 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1383 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1384 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1385 */
1386 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1387 int flags)
1388 {
1389 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1390 feedAppendOnlyFile(cmd,dbid,argv,argc);
1391 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1392 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1393 }
1394
1395 /* Used inside commands to schedule the propagation of additional commands
1396 * after the current command is propagated to AOF / Replication. */
1397 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1398 int target)
1399 {
1400 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1401 }
1402
1403 /* Call() is the core of Redis execution of a command */
1404 void call(redisClient *c, int flags) {
1405 long long dirty, start = ustime(), duration;
1406
1407 /* Sent the command to clients in MONITOR mode, only if the commands are
1408 * not geneated from reading an AOF. */
1409 if (listLength(server.monitors) && !server.loading)
1410 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1411
1412 /* Call the command. */
1413 redisOpArrayInit(&server.also_propagate);
1414 dirty = server.dirty;
1415 c->cmd->proc(c);
1416 dirty = server.dirty-dirty;
1417 duration = ustime()-start;
1418
1419 /* When EVAL is called loading the AOF we don't want commands called
1420 * from Lua to go into the slowlog or to populate statistics. */
1421 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1422 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1423
1424 /* Log the command into the Slow log if needed, and populate the
1425 * per-command statistics that we show in INFO commandstats. */
1426 if (flags & REDIS_CALL_SLOWLOG)
1427 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1428 if (flags & REDIS_CALL_STATS) {
1429 c->cmd->microseconds += duration;
1430 c->cmd->calls++;
1431 }
1432
1433 /* Propagate the command into the AOF and replication link */
1434 if (flags & REDIS_CALL_PROPAGATE) {
1435 int flags = REDIS_PROPAGATE_NONE;
1436
1437 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1438 flags |= REDIS_PROPAGATE_REPL;
1439 if (dirty)
1440 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1441 if (flags != REDIS_PROPAGATE_NONE)
1442 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1443 }
1444 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1445 * PUSH command. */
1446 if (server.also_propagate.numops) {
1447 int j;
1448 redisOp *rop;
1449
1450 for (j = 0; j < server.also_propagate.numops; j++) {
1451 rop = &server.also_propagate.ops[j];
1452 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1453 }
1454 redisOpArrayFree(&server.also_propagate);
1455 }
1456 server.stat_numcommands++;
1457 }
1458
1459 /* If this function gets called we already read a whole
1460 * command, argments are in the client argv/argc fields.
1461 * processCommand() execute the command or prepare the
1462 * server for a bulk read from the client.
1463 *
1464 * If 1 is returned the client is still alive and valid and
1465 * and other operations can be performed by the caller. Otherwise
1466 * if 0 is returned the client was destroied (i.e. after QUIT). */
1467 int processCommand(redisClient *c) {
1468 /* The QUIT command is handled separately. Normal command procs will
1469 * go through checking for replication and QUIT will cause trouble
1470 * when FORCE_REPLICATION is enabled and would be implemented in
1471 * a regular command proc. */
1472 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1473 addReply(c,shared.ok);
1474 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1475 return REDIS_ERR;
1476 }
1477
1478 /* Now lookup the command and check ASAP about trivial error conditions
1479 * such as wrong arity, bad command name and so forth. */
1480 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1481 if (!c->cmd) {
1482 addReplyErrorFormat(c,"unknown command '%s'",
1483 (char*)c->argv[0]->ptr);
1484 return REDIS_OK;
1485 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1486 (c->argc < -c->cmd->arity)) {
1487 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1488 c->cmd->name);
1489 return REDIS_OK;
1490 }
1491
1492 /* Check if the user is authenticated */
1493 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1494 {
1495 addReplyError(c,"operation not permitted");
1496 return REDIS_OK;
1497 }
1498
1499 /* If cluster is enabled, redirect here */
1500 if (server.cluster_enabled &&
1501 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1502 int hashslot;
1503
1504 if (server.cluster.state != REDIS_CLUSTER_OK) {
1505 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1506 return REDIS_OK;
1507 } else {
1508 int ask;
1509 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1510 if (n == NULL) {
1511 addReplyError(c,"Multi keys request invalid in cluster");
1512 return REDIS_OK;
1513 } else if (n != server.cluster.myself) {
1514 addReplySds(c,sdscatprintf(sdsempty(),
1515 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1516 hashslot,n->ip,n->port));
1517 return REDIS_OK;
1518 }
1519 }
1520 }
1521
1522 /* Handle the maxmemory directive.
1523 *
1524 * First we try to free some memory if possible (if there are volatile
1525 * keys in the dataset). If there are not the only thing we can do
1526 * is returning an error. */
1527 if (server.maxmemory) {
1528 int retval = freeMemoryIfNeeded();
1529 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1530 addReply(c, shared.oomerr);
1531 return REDIS_OK;
1532 }
1533 }
1534
1535 /* Don't accept write commands if there are problems persisting on disk. */
1536 if (server.stop_writes_on_bgsave_err &&
1537 server.saveparamslen > 0
1538 && server.lastbgsave_status == REDIS_ERR &&
1539 c->cmd->flags & REDIS_CMD_WRITE)
1540 {
1541 addReply(c, shared.bgsaveerr);
1542 return REDIS_OK;
1543 }
1544
1545 /* Don't accept wirte commands if this is a read only slave. But
1546 * accept write commands if this is our master. */
1547 if (server.masterhost && server.repl_slave_ro &&
1548 !(c->flags & REDIS_MASTER) &&
1549 c->cmd->flags & REDIS_CMD_WRITE)
1550 {
1551 addReply(c, shared.roslaveerr);
1552 return REDIS_OK;
1553 }
1554
1555 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1556 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1557 &&
1558 c->cmd->proc != subscribeCommand &&
1559 c->cmd->proc != unsubscribeCommand &&
1560 c->cmd->proc != psubscribeCommand &&
1561 c->cmd->proc != punsubscribeCommand) {
1562 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1563 return REDIS_OK;
1564 }
1565
1566 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1567 * we are a slave with a broken link with master. */
1568 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1569 server.repl_serve_stale_data == 0 &&
1570 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1571 {
1572 addReplyError(c,
1573 "link with MASTER is down and slave-serve-stale-data is set to no");
1574 return REDIS_OK;
1575 }
1576
1577 /* Loading DB? Return an error if the command is not INFO */
1578 if (server.loading && c->cmd->proc != infoCommand) {
1579 addReply(c, shared.loadingerr);
1580 return REDIS_OK;
1581 }
1582
1583 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1584 if (server.lua_timedout &&
1585 !(c->cmd->proc != shutdownCommand &&
1586 c->argc == 2 &&
1587 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1588 !(c->cmd->proc == scriptCommand &&
1589 c->argc == 2 &&
1590 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1591 {
1592 addReply(c, shared.slowscripterr);
1593 return REDIS_OK;
1594 }
1595
1596 /* Exec the command */
1597 if (c->flags & REDIS_MULTI &&
1598 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1599 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1600 {
1601 queueMultiCommand(c);
1602 addReply(c,shared.queued);
1603 } else {
1604 call(c,REDIS_CALL_FULL);
1605 }
1606 return REDIS_OK;
1607 }
1608
1609 /*================================== Shutdown =============================== */
1610
1611 int prepareForShutdown(int flags) {
1612 int save = flags & REDIS_SHUTDOWN_SAVE;
1613 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1614
1615 redisLog(REDIS_WARNING,"User requested shutdown...");
1616 /* Kill the saving child if there is a background saving in progress.
1617 We want to avoid race conditions, for instance our saving child may
1618 overwrite the synchronous saving did by SHUTDOWN. */
1619 if (server.rdb_child_pid != -1) {
1620 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1621 kill(server.rdb_child_pid,SIGKILL);
1622 rdbRemoveTempFile(server.rdb_child_pid);
1623 }
1624 if (server.aof_state != REDIS_AOF_OFF) {
1625 /* Kill the AOF saving child as the AOF we already have may be longer
1626 * but contains the full dataset anyway. */
1627 if (server.aof_child_pid != -1) {
1628 redisLog(REDIS_WARNING,
1629 "There is a child rewriting the AOF. Killing it!");
1630 kill(server.aof_child_pid,SIGKILL);
1631 }
1632 /* Append only file: fsync() the AOF and exit */
1633 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1634 aof_fsync(server.aof_fd);
1635 }
1636 if ((server.saveparamslen > 0 && !nosave) || save) {
1637 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1638 /* Snapshotting. Perform a SYNC SAVE and exit */
1639 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1640 /* Ooops.. error saving! The best we can do is to continue
1641 * operating. Note that if there was a background saving process,
1642 * in the next cron() Redis will be notified that the background
1643 * saving aborted, handling special stuff like slaves pending for
1644 * synchronization... */
1645 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1646 return REDIS_ERR;
1647 }
1648 }
1649 if (server.daemonize) {
1650 redisLog(REDIS_NOTICE,"Removing the pid file.");
1651 unlink(server.pidfile);
1652 }
1653 /* Close the listening sockets. Apparently this allows faster restarts. */
1654 if (server.ipfd != -1) close(server.ipfd);
1655 if (server.sofd != -1) close(server.sofd);
1656 if (server.unixsocket) {
1657 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1658 unlink(server.unixsocket); /* don't care if this fails */
1659 }
1660
1661 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1662 return REDIS_OK;
1663 }
1664
1665 /*================================== Commands =============================== */
1666
1667 void authCommand(redisClient *c) {
1668 if (!server.requirepass) {
1669 addReplyError(c,"Client sent AUTH, but no password is set");
1670 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1671 c->authenticated = 1;
1672 addReply(c,shared.ok);
1673 } else {
1674 c->authenticated = 0;
1675 addReplyError(c,"invalid password");
1676 }
1677 }
1678
1679 void pingCommand(redisClient *c) {
1680 addReply(c,shared.pong);
1681 }
1682
1683 void echoCommand(redisClient *c) {
1684 addReplyBulk(c,c->argv[1]);
1685 }
1686
1687 void timeCommand(redisClient *c) {
1688 struct timeval tv;
1689
1690 /* gettimeofday() can only fail if &tv is a bad addresss so we
1691 * don't check for errors. */
1692 gettimeofday(&tv,NULL);
1693 addReplyMultiBulkLen(c,2);
1694 addReplyBulkLongLong(c,tv.tv_sec);
1695 addReplyBulkLongLong(c,tv.tv_usec);
1696 }
1697
1698 /* Convert an amount of bytes into a human readable string in the form
1699 * of 100B, 2G, 100M, 4K, and so forth. */
1700 void bytesToHuman(char *s, unsigned long long n) {
1701 double d;
1702
1703 if (n < 1024) {
1704 /* Bytes */
1705 sprintf(s,"%lluB",n);
1706 return;
1707 } else if (n < (1024*1024)) {
1708 d = (double)n/(1024);
1709 sprintf(s,"%.2fK",d);
1710 } else if (n < (1024LL*1024*1024)) {
1711 d = (double)n/(1024*1024);
1712 sprintf(s,"%.2fM",d);
1713 } else if (n < (1024LL*1024*1024*1024)) {
1714 d = (double)n/(1024LL*1024*1024);
1715 sprintf(s,"%.2fG",d);
1716 }
1717 }
1718
1719 /* Create the string returned by the INFO command. This is decoupled
1720 * by the INFO command itself as we need to report the same information
1721 * on memory corruption problems. */
1722 sds genRedisInfoString(char *section) {
1723 sds info = sdsempty();
1724 time_t uptime = server.unixtime-server.stat_starttime;
1725 int j, numcommands;
1726 struct rusage self_ru, c_ru;
1727 unsigned long lol, bib;
1728 int allsections = 0, defsections = 0;
1729 int sections = 0;
1730
1731 if (section) {
1732 allsections = strcasecmp(section,"all") == 0;
1733 defsections = strcasecmp(section,"default") == 0;
1734 }
1735
1736 getrusage(RUSAGE_SELF, &self_ru);
1737 getrusage(RUSAGE_CHILDREN, &c_ru);
1738 getClientsMaxBuffers(&lol,&bib);
1739
1740 /* Server */
1741 if (allsections || defsections || !strcasecmp(section,"server")) {
1742 if (sections++) info = sdscat(info,"\r\n");
1743 info = sdscatprintf(info,
1744 "# Server\r\n"
1745 "redis_version:%s\r\n"
1746 "redis_git_sha1:%s\r\n"
1747 "redis_git_dirty:%d\r\n"
1748 "arch_bits:%d\r\n"
1749 "multiplexing_api:%s\r\n"
1750 "gcc_version:%d.%d.%d\r\n"
1751 "process_id:%ld\r\n"
1752 "run_id:%s\r\n"
1753 "tcp_port:%d\r\n"
1754 "uptime_in_seconds:%ld\r\n"
1755 "uptime_in_days:%ld\r\n"
1756 "lru_clock:%ld\r\n",
1757 REDIS_VERSION,
1758 redisGitSHA1(),
1759 strtol(redisGitDirty(),NULL,10) > 0,
1760 server.arch_bits,
1761 aeGetApiName(),
1762 #ifdef __GNUC__
1763 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1764 #else
1765 0,0,0,
1766 #endif
1767 (long) getpid(),
1768 server.runid,
1769 server.port,
1770 uptime,
1771 uptime/(3600*24),
1772 (unsigned long) server.lruclock);
1773 }
1774
1775 /* Clients */
1776 if (allsections || defsections || !strcasecmp(section,"clients")) {
1777 if (sections++) info = sdscat(info,"\r\n");
1778 info = sdscatprintf(info,
1779 "# Clients\r\n"
1780 "connected_clients:%lu\r\n"
1781 "client_longest_output_list:%lu\r\n"
1782 "client_biggest_input_buf:%lu\r\n"
1783 "blocked_clients:%d\r\n",
1784 listLength(server.clients)-listLength(server.slaves),
1785 lol, bib,
1786 server.bpop_blocked_clients);
1787 }
1788
1789 /* Memory */
1790 if (allsections || defsections || !strcasecmp(section,"memory")) {
1791 char hmem[64];
1792 char peak_hmem[64];
1793
1794 bytesToHuman(hmem,zmalloc_used_memory());
1795 bytesToHuman(peak_hmem,server.stat_peak_memory);
1796 if (sections++) info = sdscat(info,"\r\n");
1797 info = sdscatprintf(info,
1798 "# Memory\r\n"
1799 "used_memory:%zu\r\n"
1800 "used_memory_human:%s\r\n"
1801 "used_memory_rss:%zu\r\n"
1802 "used_memory_peak:%zu\r\n"
1803 "used_memory_peak_human:%s\r\n"
1804 "used_memory_lua:%lld\r\n"
1805 "mem_fragmentation_ratio:%.2f\r\n"
1806 "mem_allocator:%s\r\n",
1807 zmalloc_used_memory(),
1808 hmem,
1809 zmalloc_get_rss(),
1810 server.stat_peak_memory,
1811 peak_hmem,
1812 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1813 zmalloc_get_fragmentation_ratio(),
1814 ZMALLOC_LIB
1815 );
1816 }
1817
1818 /* Persistence */
1819 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1820 if (sections++) info = sdscat(info,"\r\n");
1821 info = sdscatprintf(info,
1822 "# Persistence\r\n"
1823 "loading:%d\r\n"
1824 "aof_enabled:%d\r\n"
1825 "changes_since_last_save:%lld\r\n"
1826 "bgsave_in_progress:%d\r\n"
1827 "last_save_time:%ld\r\n"
1828 "last_bgsave_status:%s\r\n"
1829 "bgrewriteaof_in_progress:%d\r\n",
1830 server.loading,
1831 server.aof_state != REDIS_AOF_OFF,
1832 server.dirty,
1833 server.rdb_child_pid != -1,
1834 server.lastsave,
1835 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1836 server.aof_child_pid != -1);
1837
1838 if (server.aof_state != REDIS_AOF_OFF) {
1839 info = sdscatprintf(info,
1840 "aof_current_size:%lld\r\n"
1841 "aof_base_size:%lld\r\n"
1842 "aof_pending_rewrite:%d\r\n"
1843 "aof_buffer_length:%zu\r\n"
1844 "aof_pending_bio_fsync:%llu\r\n"
1845 "aof_delayed_fsync:%lu\r\n",
1846 (long long) server.aof_current_size,
1847 (long long) server.aof_rewrite_base_size,
1848 server.aof_rewrite_scheduled,
1849 sdslen(server.aof_buf),
1850 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1851 server.aof_delayed_fsync);
1852 }
1853
1854 if (server.loading) {
1855 double perc;
1856 time_t eta, elapsed;
1857 off_t remaining_bytes = server.loading_total_bytes-
1858 server.loading_loaded_bytes;
1859
1860 perc = ((double)server.loading_loaded_bytes /
1861 server.loading_total_bytes) * 100;
1862
1863 elapsed = server.unixtime-server.loading_start_time;
1864 if (elapsed == 0) {
1865 eta = 1; /* A fake 1 second figure if we don't have
1866 enough info */
1867 } else {
1868 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1869 }
1870
1871 info = sdscatprintf(info,
1872 "loading_start_time:%ld\r\n"
1873 "loading_total_bytes:%llu\r\n"
1874 "loading_loaded_bytes:%llu\r\n"
1875 "loading_loaded_perc:%.2f\r\n"
1876 "loading_eta_seconds:%ld\r\n"
1877 ,(unsigned long) server.loading_start_time,
1878 (unsigned long long) server.loading_total_bytes,
1879 (unsigned long long) server.loading_loaded_bytes,
1880 perc,
1881 eta
1882 );
1883 }
1884 }
1885
1886 /* Stats */
1887 if (allsections || defsections || !strcasecmp(section,"stats")) {
1888 if (sections++) info = sdscat(info,"\r\n");
1889 info = sdscatprintf(info,
1890 "# Stats\r\n"
1891 "total_connections_received:%lld\r\n"
1892 "total_commands_processed:%lld\r\n"
1893 "instantaneous_ops_per_sec:%lld\r\n"
1894 "rejected_connections:%lld\r\n"
1895 "expired_keys:%lld\r\n"
1896 "evicted_keys:%lld\r\n"
1897 "keyspace_hits:%lld\r\n"
1898 "keyspace_misses:%lld\r\n"
1899 "pubsub_channels:%ld\r\n"
1900 "pubsub_patterns:%lu\r\n"
1901 "latest_fork_usec:%lld\r\n",
1902 server.stat_numconnections,
1903 server.stat_numcommands,
1904 getOperationsPerSecond(),
1905 server.stat_rejected_conn,
1906 server.stat_expiredkeys,
1907 server.stat_evictedkeys,
1908 server.stat_keyspace_hits,
1909 server.stat_keyspace_misses,
1910 dictSize(server.pubsub_channels),
1911 listLength(server.pubsub_patterns),
1912 server.stat_fork_time);
1913 }
1914
1915 /* Replication */
1916 if (allsections || defsections || !strcasecmp(section,"replication")) {
1917 if (sections++) info = sdscat(info,"\r\n");
1918 info = sdscatprintf(info,
1919 "# Replication\r\n"
1920 "role:%s\r\n",
1921 server.masterhost == NULL ? "master" : "slave");
1922 if (server.masterhost) {
1923 info = sdscatprintf(info,
1924 "master_host:%s\r\n"
1925 "master_port:%d\r\n"
1926 "master_link_status:%s\r\n"
1927 "master_last_io_seconds_ago:%d\r\n"
1928 "master_sync_in_progress:%d\r\n"
1929 ,server.masterhost,
1930 server.masterport,
1931 (server.repl_state == REDIS_REPL_CONNECTED) ?
1932 "up" : "down",
1933 server.master ?
1934 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
1935 server.repl_state == REDIS_REPL_TRANSFER
1936 );
1937
1938 if (server.repl_state == REDIS_REPL_TRANSFER) {
1939 info = sdscatprintf(info,
1940 "master_sync_left_bytes:%ld\r\n"
1941 "master_sync_last_io_seconds_ago:%d\r\n"
1942 ,(long)server.repl_transfer_left,
1943 (int)(server.unixtime-server.repl_transfer_lastio)
1944 );
1945 }
1946
1947 if (server.repl_state != REDIS_REPL_CONNECTED) {
1948 info = sdscatprintf(info,
1949 "master_link_down_since_seconds:%ld\r\n",
1950 (long)server.unixtime-server.repl_down_since);
1951 }
1952 }
1953 info = sdscatprintf(info,
1954 "connected_slaves:%lu\r\n",
1955 listLength(server.slaves));
1956 if (listLength(server.slaves)) {
1957 int slaveid = 0;
1958 listNode *ln;
1959 listIter li;
1960
1961 listRewind(server.slaves,&li);
1962 while((ln = listNext(&li))) {
1963 redisClient *slave = listNodeValue(ln);
1964 char *state = NULL;
1965 char ip[32];
1966 int port;
1967
1968 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
1969 switch(slave->replstate) {
1970 case REDIS_REPL_WAIT_BGSAVE_START:
1971 case REDIS_REPL_WAIT_BGSAVE_END:
1972 state = "wait_bgsave";
1973 break;
1974 case REDIS_REPL_SEND_BULK:
1975 state = "send_bulk";
1976 break;
1977 case REDIS_REPL_ONLINE:
1978 state = "online";
1979 break;
1980 }
1981 if (state == NULL) continue;
1982 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
1983 slaveid,ip,port,state);
1984 slaveid++;
1985 }
1986 }
1987 }
1988
1989 /* CPU */
1990 if (allsections || defsections || !strcasecmp(section,"cpu")) {
1991 if (sections++) info = sdscat(info,"\r\n");
1992 info = sdscatprintf(info,
1993 "# CPU\r\n"
1994 "used_cpu_sys:%.2f\r\n"
1995 "used_cpu_user:%.2f\r\n"
1996 "used_cpu_sys_children:%.2f\r\n"
1997 "used_cpu_user_children:%.2f\r\n",
1998 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
1999 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2000 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2001 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2002 }
2003
2004 /* cmdtime */
2005 if (allsections || !strcasecmp(section,"commandstats")) {
2006 if (sections++) info = sdscat(info,"\r\n");
2007 info = sdscatprintf(info, "# Commandstats\r\n");
2008 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2009 for (j = 0; j < numcommands; j++) {
2010 struct redisCommand *c = redisCommandTable+j;
2011
2012 if (!c->calls) continue;
2013 info = sdscatprintf(info,
2014 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2015 c->name, c->calls, c->microseconds,
2016 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2017 }
2018 }
2019
2020 /* Clusetr */
2021 if (allsections || defsections || !strcasecmp(section,"cluster")) {
2022 if (sections++) info = sdscat(info,"\r\n");
2023 info = sdscatprintf(info,
2024 "# Cluster\r\n"
2025 "cluster_enabled:%d\r\n",
2026 server.cluster_enabled);
2027 }
2028
2029 /* Key space */
2030 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2031 if (sections++) info = sdscat(info,"\r\n");
2032 info = sdscatprintf(info, "# Keyspace\r\n");
2033 for (j = 0; j < server.dbnum; j++) {
2034 long long keys, vkeys;
2035
2036 keys = dictSize(server.db[j].dict);
2037 vkeys = dictSize(server.db[j].expires);
2038 if (keys || vkeys) {
2039 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2040 j, keys, vkeys);
2041 }
2042 }
2043 }
2044 return info;
2045 }
2046
2047 void infoCommand(redisClient *c) {
2048 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2049
2050 if (c->argc > 2) {
2051 addReply(c,shared.syntaxerr);
2052 return;
2053 }
2054 sds info = genRedisInfoString(section);
2055 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2056 (unsigned long)sdslen(info)));
2057 addReplySds(c,info);
2058 addReply(c,shared.crlf);
2059 }
2060
2061 void monitorCommand(redisClient *c) {
2062 /* ignore MONITOR if aleady slave or in monitor mode */
2063 if (c->flags & REDIS_SLAVE) return;
2064
2065 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2066 c->slaveseldb = 0;
2067 listAddNodeTail(server.monitors,c);
2068 addReply(c,shared.ok);
2069 }
2070
2071 /* ============================ Maxmemory directive ======================== */
2072
2073 /* This function gets called when 'maxmemory' is set on the config file to limit
2074 * the max memory used by the server, before processing a command.
2075 *
2076 * The goal of the function is to free enough memory to keep Redis under the
2077 * configured memory limit.
2078 *
2079 * The function starts calculating how many bytes should be freed to keep
2080 * Redis under the limit, and enters a loop selecting the best keys to
2081 * evict accordingly to the configured policy.
2082 *
2083 * If all the bytes needed to return back under the limit were freed the
2084 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2085 * should block the execution of commands that will result in more memory
2086 * used by the server.
2087 */
2088 int freeMemoryIfNeeded(void) {
2089 size_t mem_used, mem_tofree, mem_freed;
2090 int slaves = listLength(server.slaves);
2091
2092 /* Remove the size of slaves output buffers and AOF buffer from the
2093 * count of used memory. */
2094 mem_used = zmalloc_used_memory();
2095 if (slaves) {
2096 listIter li;
2097 listNode *ln;
2098
2099 listRewind(server.slaves,&li);
2100 while((ln = listNext(&li))) {
2101 redisClient *slave = listNodeValue(ln);
2102 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2103 if (obuf_bytes > mem_used)
2104 mem_used = 0;
2105 else
2106 mem_used -= obuf_bytes;
2107 }
2108 }
2109 if (server.aof_state != REDIS_AOF_OFF) {
2110 mem_used -= sdslen(server.aof_buf);
2111 mem_used -= sdslen(server.aof_rewrite_buf);
2112 }
2113
2114 /* Check if we are over the memory limit. */
2115 if (mem_used <= server.maxmemory) return REDIS_OK;
2116
2117 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2118 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2119
2120 /* Compute how much memory we need to free. */
2121 mem_tofree = mem_used - server.maxmemory;
2122 mem_freed = 0;
2123 while (mem_freed < mem_tofree) {
2124 int j, k, keys_freed = 0;
2125
2126 for (j = 0; j < server.dbnum; j++) {
2127 long bestval = 0; /* just to prevent warning */
2128 sds bestkey = NULL;
2129 struct dictEntry *de;
2130 redisDb *db = server.db+j;
2131 dict *dict;
2132
2133 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2134 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2135 {
2136 dict = server.db[j].dict;
2137 } else {
2138 dict = server.db[j].expires;
2139 }
2140 if (dictSize(dict) == 0) continue;
2141
2142 /* volatile-random and allkeys-random policy */
2143 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2144 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2145 {
2146 de = dictGetRandomKey(dict);
2147 bestkey = dictGetKey(de);
2148 }
2149
2150 /* volatile-lru and allkeys-lru policy */
2151 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2152 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2153 {
2154 for (k = 0; k < server.maxmemory_samples; k++) {
2155 sds thiskey;
2156 long thisval;
2157 robj *o;
2158
2159 de = dictGetRandomKey(dict);
2160 thiskey = dictGetKey(de);
2161 /* When policy is volatile-lru we need an additonal lookup
2162 * to locate the real key, as dict is set to db->expires. */
2163 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2164 de = dictFind(db->dict, thiskey);
2165 o = dictGetVal(de);
2166 thisval = estimateObjectIdleTime(o);
2167
2168 /* Higher idle time is better candidate for deletion */
2169 if (bestkey == NULL || thisval > bestval) {
2170 bestkey = thiskey;
2171 bestval = thisval;
2172 }
2173 }
2174 }
2175
2176 /* volatile-ttl */
2177 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2178 for (k = 0; k < server.maxmemory_samples; k++) {
2179 sds thiskey;
2180 long thisval;
2181
2182 de = dictGetRandomKey(dict);
2183 thiskey = dictGetKey(de);
2184 thisval = (long) dictGetVal(de);
2185
2186 /* Expire sooner (minor expire unix timestamp) is better
2187 * candidate for deletion */
2188 if (bestkey == NULL || thisval < bestval) {
2189 bestkey = thiskey;
2190 bestval = thisval;
2191 }
2192 }
2193 }
2194
2195 /* Finally remove the selected key. */
2196 if (bestkey) {
2197 long long delta;
2198
2199 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2200 propagateExpire(db,keyobj);
2201 /* We compute the amount of memory freed by dbDelete() alone.
2202 * It is possible that actually the memory needed to propagate
2203 * the DEL in AOF and replication link is greater than the one
2204 * we are freeing removing the key, but we can't account for
2205 * that otherwise we would never exit the loop.
2206 *
2207 * AOF and Output buffer memory will be freed eventually so
2208 * we only care about memory used by the key space. */
2209 delta = (long long) zmalloc_used_memory();
2210 dbDelete(db,keyobj);
2211 delta -= (long long) zmalloc_used_memory();
2212 mem_freed += delta;
2213 server.stat_evictedkeys++;
2214 decrRefCount(keyobj);
2215 keys_freed++;
2216
2217 /* When the memory to free starts to be big enough, we may
2218 * start spending so much time here that is impossible to
2219 * deliver data to the slaves fast enough, so we force the
2220 * transmission here inside the loop. */
2221 if (slaves) flushSlavesOutputBuffers();
2222 }
2223 }
2224 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2225 }
2226 return REDIS_OK;
2227 }
2228
2229 /* =================================== Main! ================================ */
2230
2231 #ifdef __linux__
2232 int linuxOvercommitMemoryValue(void) {
2233 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2234 char buf[64];
2235
2236 if (!fp) return -1;
2237 if (fgets(buf,64,fp) == NULL) {
2238 fclose(fp);
2239 return -1;
2240 }
2241 fclose(fp);
2242
2243 return atoi(buf);
2244 }
2245
2246 void linuxOvercommitMemoryWarning(void) {
2247 if (linuxOvercommitMemoryValue() == 0) {
2248 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.");
2249 }
2250 }
2251 #endif /* __linux__ */
2252
2253 void createPidFile(void) {
2254 /* Try to write the pid file in a best-effort way. */
2255 FILE *fp = fopen(server.pidfile,"w");
2256 if (fp) {
2257 fprintf(fp,"%d\n",(int)getpid());
2258 fclose(fp);
2259 }
2260 }
2261
2262 void daemonize(void) {
2263 int fd;
2264
2265 if (fork() != 0) exit(0); /* parent exits */
2266 setsid(); /* create a new session */
2267
2268 /* Every output goes to /dev/null. If Redis is daemonized but
2269 * the 'logfile' is set to 'stdout' in the configuration file
2270 * it will not log at all. */
2271 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2272 dup2(fd, STDIN_FILENO);
2273 dup2(fd, STDOUT_FILENO);
2274 dup2(fd, STDERR_FILENO);
2275 if (fd > STDERR_FILENO) close(fd);
2276 }
2277 }
2278
2279 void version() {
2280 printf("Redis server v=%s sha=%s:%d malloc=%s\n", REDIS_VERSION,
2281 redisGitSHA1(), atoi(redisGitDirty()) > 0, ZMALLOC_LIB);
2282 exit(0);
2283 }
2284
2285 void usage() {
2286 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2287 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2288 fprintf(stderr," ./redis-server -v or --version\n");
2289 fprintf(stderr," ./redis-server -h or --help\n");
2290 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2291 fprintf(stderr,"Examples:\n");
2292 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2293 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2294 fprintf(stderr," ./redis-server --port 7777\n");
2295 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2296 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2297 exit(1);
2298 }
2299
2300 void redisAsciiArt(void) {
2301 #include "asciilogo.h"
2302 char *buf = zmalloc(1024*16);
2303
2304 snprintf(buf,1024*16,ascii_logo,
2305 REDIS_VERSION,
2306 redisGitSHA1(),
2307 strtol(redisGitDirty(),NULL,10) > 0,
2308 (sizeof(long) == 8) ? "64" : "32",
2309 server.cluster_enabled ? "cluster" : "stand alone",
2310 server.port,
2311 (long) getpid()
2312 );
2313 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2314 zfree(buf);
2315 }
2316
2317 static void sigtermHandler(int sig) {
2318 REDIS_NOTUSED(sig);
2319
2320 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2321 server.shutdown_asap = 1;
2322 }
2323
2324 void setupSignalHandlers(void) {
2325 struct sigaction act;
2326
2327 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2328 * Otherwise, sa_handler is used. */
2329 sigemptyset(&act.sa_mask);
2330 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
2331 act.sa_handler = sigtermHandler;
2332 sigaction(SIGTERM, &act, NULL);
2333
2334 #ifdef HAVE_BACKTRACE
2335 sigemptyset(&act.sa_mask);
2336 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
2337 act.sa_sigaction = sigsegvHandler;
2338 sigaction(SIGSEGV, &act, NULL);
2339 sigaction(SIGBUS, &act, NULL);
2340 sigaction(SIGFPE, &act, NULL);
2341 sigaction(SIGILL, &act, NULL);
2342 #endif
2343 return;
2344 }
2345
2346 void memtest(size_t megabytes, int passes);
2347
2348 int main(int argc, char **argv) {
2349 long long start;
2350 struct timeval tv;
2351
2352 /* We need to initialize our libraries, and the server configuration. */
2353 zmalloc_enable_thread_safeness();
2354 srand(time(NULL)^getpid());
2355 gettimeofday(&tv,NULL);
2356 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2357 initServerConfig();
2358
2359 if (argc >= 2) {
2360 int j = 1; /* First option to parse in argv[] */
2361 sds options = sdsempty();
2362 char *configfile = NULL;
2363
2364 /* Handle special options --help and --version */
2365 if (strcmp(argv[1], "-v") == 0 ||
2366 strcmp(argv[1], "--version") == 0) version();
2367 if (strcmp(argv[1], "--help") == 0 ||
2368 strcmp(argv[1], "-h") == 0) usage();
2369 if (strcmp(argv[1], "--test-memory") == 0) {
2370 if (argc == 3) {
2371 memtest(atoi(argv[2]),50);
2372 exit(0);
2373 } else {
2374 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2375 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2376 exit(1);
2377 }
2378 }
2379
2380 /* First argument is the config file name? */
2381 if (argv[j][0] != '-' || argv[j][1] != '-')
2382 configfile = argv[j++];
2383 /* All the other options are parsed and conceptually appended to the
2384 * configuration file. For instance --port 6380 will generate the
2385 * string "port 6380\n" to be parsed after the actual file name
2386 * is parsed, if any. */
2387 while(j != argc) {
2388 if (argv[j][0] == '-' && argv[j][1] == '-') {
2389 /* Option name */
2390 if (sdslen(options)) options = sdscat(options,"\n");
2391 options = sdscat(options,argv[j]+2);
2392 options = sdscat(options," ");
2393 } else {
2394 /* Option argument */
2395 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2396 options = sdscat(options," ");
2397 }
2398 j++;
2399 }
2400 resetServerSaveParams();
2401 loadServerConfig(configfile,options);
2402 sdsfree(options);
2403 } else {
2404 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'");
2405 }
2406 if (server.daemonize) daemonize();
2407 initServer();
2408 if (server.daemonize) createPidFile();
2409 redisAsciiArt();
2410 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2411 #ifdef __linux__
2412 linuxOvercommitMemoryWarning();
2413 #endif
2414 start = ustime();
2415 if (server.aof_state == REDIS_AOF_ON) {
2416 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2417 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2418 } else {
2419 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2420 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2421 (float)(ustime()-start)/1000000);
2422 } else if (errno != ENOENT) {
2423 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2424 exit(1);
2425 }
2426 }
2427 if (server.ipfd > 0)
2428 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2429 if (server.sofd > 0)
2430 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2431 aeSetBeforeSleepProc(server.el,beforeSleep);
2432 aeMain(server.el);
2433 aeDeleteEventLoop(server.el);
2434 return 0;
2435 }
2436
2437 /* The End */