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