]> git.saurik.com Git - redis.git/blob - src/redis.c
zmalloc: kill unused __size parameter in update_zmalloc_stat_alloc() macro.
[redis.git] / src / redis.c
1 /*
2 * Copyright (c) 2009-2012, 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 #include <sys/utsname.h>
52
53 /* Our shared "common" objects */
54
55 struct sharedObjectsStruct shared;
56
57 /* Global vars that are actually used as constants. The following double
58 * values are used for double on-disk serialization, and are initialized
59 * at runtime to avoid strange compiler optimizations. */
60
61 double R_Zero, R_PosInf, R_NegInf, R_Nan;
62
63 /*================================= Globals ================================= */
64
65 /* Global vars */
66 struct redisServer server; /* server global state */
67 struct redisCommand *commandTable;
68
69 /* Our command table.
70 *
71 * Every entry is composed of the following fields:
72 *
73 * name: a string representing the command name.
74 * function: pointer to the C function implementing the command.
75 * arity: number of arguments, it is possible to use -N to say >= N
76 * sflags: command flags as string. See below for a table of flags.
77 * flags: flags as bitmask. Computed by Redis using the 'sflags' field.
78 * get_keys_proc: an optional function to get key arguments from a command.
79 * This is only used when the following three fields are not
80 * enough to specify what arguments are keys.
81 * first_key_index: first argument that is a key
82 * last_key_index: last argument that is a key
83 * key_step: step to get all the keys from first to last argument. For instance
84 * in MSET the step is two since arguments are key,val,key,val,...
85 * microseconds: microseconds of total execution time for this command.
86 * calls: total number of calls of this command.
87 *
88 * The flags, microseconds and calls fields are computed by Redis and should
89 * always be set to zero.
90 *
91 * Command flags are expressed using strings where every character represents
92 * a flag. Later the populateCommandTable() function will take care of
93 * populating the real 'flags' field using this characters.
94 *
95 * This is the meaning of the flags:
96 *
97 * w: write command (may modify the key space).
98 * r: read command (will never modify the key space).
99 * m: may increase memory usage once called. Don't allow if out of memory.
100 * a: admin command, like SAVE or SHUTDOWN.
101 * p: Pub/Sub related command.
102 * f: force replication of this command, regarless of server.dirty.
103 * s: command not allowed in scripts.
104 * R: random command. Command is not deterministic, that is, the same command
105 * with the same arguments, with the same key space, may have different
106 * results. For instance SPOP and RANDOMKEY are two random commands.
107 * S: Sort command output array if called from script, so that the output
108 * is deterministic.
109 * l: Allow command while loading the database.
110 * t: Allow command while a slave has stale data but is not allowed to
111 * server this data. Normally no command is accepted in this condition
112 * but just a few.
113 * M: Do not automatically propagate the command on MONITOR.
114 */
115 struct redisCommand redisCommandTable[] = {
116 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
117 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
118 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
119 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
120 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
121 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
122 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
123 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
124 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
125 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
126 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
127 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
128 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
129 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
130 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
131 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
132 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
133 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
134 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
135 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
136 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
137 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
138 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
139 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
140 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
141 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
142 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
143 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
144 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
145 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
146 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
147 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
148 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
149 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
150 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
151 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
152 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
153 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
154 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
155 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
156 {"srandmember",srandmemberCommand,-2,"rR",0,NULL,1,1,1,0,0},
157 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
158 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
159 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
160 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
161 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
162 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
163 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
164 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
165 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
166 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
167 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
168 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
169 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
170 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
171 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
172 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
173 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
174 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
175 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
176 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
177 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
178 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
179 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
180 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
182 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
183 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
184 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
185 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
186 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
187 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
188 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
189 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
190 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
191 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
192 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
193 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
194 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
195 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
196 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
197 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
198 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
199 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
200 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
201 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
202 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
203 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
204 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
205 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
206 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
207 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
208 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
209 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
210 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
211 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
212 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
213 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
214 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
215 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
216 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
217 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
218 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
219 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
220 {"exec",execCommand,1,"sM",0,NULL,0,0,0,0,0},
221 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
222 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
223 {"replconf",replconfCommand,-1,"ars",0,NULL,0,0,0,0,0},
224 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
225 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
226 {"sort",sortCommand,-2,"wm",0,NULL,1,1,1,0,0},
227 {"info",infoCommand,-1,"rlt",0,NULL,0,0,0,0,0},
228 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
229 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
230 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
231 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
232 {"slaveof",slaveofCommand,3,"ast",0,NULL,0,0,0,0,0},
233 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
234 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
235 {"subscribe",subscribeCommand,-2,"rpslt",0,NULL,0,0,0,0,0},
236 {"unsubscribe",unsubscribeCommand,-1,"rpslt",0,NULL,0,0,0,0,0},
237 {"psubscribe",psubscribeCommand,-2,"rpslt",0,NULL,0,0,0,0,0},
238 {"punsubscribe",punsubscribeCommand,-1,"rpslt",0,NULL,0,0,0,0,0},
239 {"publish",publishCommand,3,"pflt",0,NULL,0,0,0,0,0},
240 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
241 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
242 {"cluster",clusterCommand,-2,"ar",0,NULL,0,0,0,0,0},
243 {"restore",restoreCommand,-4,"awm",0,NULL,1,1,1,0,0},
244 {"migrate",migrateCommand,-6,"aw",0,NULL,0,0,0,0,0},
245 {"asking",askingCommand,1,"r",0,NULL,0,0,0,0,0},
246 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
247 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
248 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
249 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
250 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
251 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
252 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
253 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0},
254 {"bitop",bitopCommand,-4,"wm",0,NULL,2,-1,1,0,0},
255 {"bitcount",bitcountCommand,-2,"r",0,NULL,1,1,1,0,0}
256 };
257
258 /*============================ Utility functions ============================ */
259
260 /* Low level logging. To use only for very big messages, otherwise
261 * redisLog() is to prefer. */
262 void redisLogRaw(int level, const char *msg) {
263 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
264 const char *c = ".-*#";
265 FILE *fp;
266 char buf[64];
267 int rawmode = (level & REDIS_LOG_RAW);
268
269 level &= 0xff; /* clear flags */
270 if (level < server.verbosity) return;
271
272 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
273 if (!fp) return;
274
275 if (rawmode) {
276 fprintf(fp,"%s",msg);
277 } else {
278 int off;
279 struct timeval tv;
280
281 gettimeofday(&tv,NULL);
282 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
283 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
284 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
285 }
286 fflush(fp);
287
288 if (server.logfile) fclose(fp);
289
290 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
291 }
292
293 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
294 * is used across the code. The raw version is only used in order to dump
295 * the INFO output on crash. */
296 void redisLog(int level, const char *fmt, ...) {
297 va_list ap;
298 char msg[REDIS_MAX_LOGMSG_LEN];
299
300 if ((level&0xff) < server.verbosity) return;
301
302 va_start(ap, fmt);
303 vsnprintf(msg, sizeof(msg), fmt, ap);
304 va_end(ap);
305
306 redisLogRaw(level,msg);
307 }
308
309 /* Log a fixed message without printf-alike capabilities, in a way that is
310 * safe to call from a signal handler.
311 *
312 * We actually use this only for signals that are not fatal from the point
313 * of view of Redis. Signals that are going to kill the server anyway and
314 * where we need printf-alike features are served by redisLog(). */
315 void redisLogFromHandler(int level, const char *msg) {
316 int fd;
317 char buf[64];
318
319 if ((level&0xff) < server.verbosity ||
320 (server.logfile == NULL && server.daemonize)) return;
321 fd = server.logfile ?
322 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
323 STDOUT_FILENO;
324 if (fd == -1) return;
325 ll2string(buf,sizeof(buf),getpid());
326 if (write(fd,"[",1) == -1) goto err;
327 if (write(fd,buf,strlen(buf)) == -1) goto err;
328 if (write(fd," | signal handler] (",20) == -1) goto err;
329 ll2string(buf,sizeof(buf),time(NULL));
330 if (write(fd,buf,strlen(buf)) == -1) goto err;
331 if (write(fd,") ",2) == -1) goto err;
332 if (write(fd,msg,strlen(msg)) == -1) goto err;
333 if (write(fd,"\n",1) == -1) goto err;
334 err:
335 if (server.logfile) close(fd);
336 }
337
338 /* Return the UNIX time in microseconds */
339 long long ustime(void) {
340 struct timeval tv;
341 long long ust;
342
343 gettimeofday(&tv, NULL);
344 ust = ((long long)tv.tv_sec)*1000000;
345 ust += tv.tv_usec;
346 return ust;
347 }
348
349 /* Return the UNIX time in milliseconds */
350 long long mstime(void) {
351 return ustime()/1000;
352 }
353
354 /* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
355 * exit(), because the latter may interact with the same file objects used by
356 * the parent process. However if we are testing the coverage normal exit() is
357 * used in order to obtain the right coverage information. */
358 void exitFromChild(int retcode) {
359 #ifdef COVERAGE_TEST
360 exit(retcode);
361 #else
362 _exit(retcode);
363 #endif
364 }
365
366 /*====================== Hash table type implementation ==================== */
367
368 /* This is an hash table type that uses the SDS dynamic strings libary as
369 * keys and radis objects as values (objects can hold SDS strings,
370 * lists, sets). */
371
372 void dictVanillaFree(void *privdata, void *val)
373 {
374 DICT_NOTUSED(privdata);
375 zfree(val);
376 }
377
378 void dictListDestructor(void *privdata, void *val)
379 {
380 DICT_NOTUSED(privdata);
381 listRelease((list*)val);
382 }
383
384 int dictSdsKeyCompare(void *privdata, const void *key1,
385 const void *key2)
386 {
387 int l1,l2;
388 DICT_NOTUSED(privdata);
389
390 l1 = sdslen((sds)key1);
391 l2 = sdslen((sds)key2);
392 if (l1 != l2) return 0;
393 return memcmp(key1, key2, l1) == 0;
394 }
395
396 /* A case insensitive version used for the command lookup table. */
397 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
398 const void *key2)
399 {
400 DICT_NOTUSED(privdata);
401
402 return strcasecmp(key1, key2) == 0;
403 }
404
405 void dictRedisObjectDestructor(void *privdata, void *val)
406 {
407 DICT_NOTUSED(privdata);
408
409 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
410 decrRefCount(val);
411 }
412
413 void dictSdsDestructor(void *privdata, void *val)
414 {
415 DICT_NOTUSED(privdata);
416
417 sdsfree(val);
418 }
419
420 int dictObjKeyCompare(void *privdata, const void *key1,
421 const void *key2)
422 {
423 const robj *o1 = key1, *o2 = key2;
424 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
425 }
426
427 unsigned int dictObjHash(const void *key) {
428 const robj *o = key;
429 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
430 }
431
432 unsigned int dictSdsHash(const void *key) {
433 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
434 }
435
436 unsigned int dictSdsCaseHash(const void *key) {
437 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
438 }
439
440 int dictEncObjKeyCompare(void *privdata, const void *key1,
441 const void *key2)
442 {
443 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
444 int cmp;
445
446 if (o1->encoding == REDIS_ENCODING_INT &&
447 o2->encoding == REDIS_ENCODING_INT)
448 return o1->ptr == o2->ptr;
449
450 o1 = getDecodedObject(o1);
451 o2 = getDecodedObject(o2);
452 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
453 decrRefCount(o1);
454 decrRefCount(o2);
455 return cmp;
456 }
457
458 unsigned int dictEncObjHash(const void *key) {
459 robj *o = (robj*) key;
460
461 if (o->encoding == REDIS_ENCODING_RAW) {
462 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
463 } else {
464 if (o->encoding == REDIS_ENCODING_INT) {
465 char buf[32];
466 int len;
467
468 len = ll2string(buf,32,(long)o->ptr);
469 return dictGenHashFunction((unsigned char*)buf, len);
470 } else {
471 unsigned int hash;
472
473 o = getDecodedObject(o);
474 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
475 decrRefCount(o);
476 return hash;
477 }
478 }
479 }
480
481 /* Sets type hash table */
482 dictType setDictType = {
483 dictEncObjHash, /* hash function */
484 NULL, /* key dup */
485 NULL, /* val dup */
486 dictEncObjKeyCompare, /* key compare */
487 dictRedisObjectDestructor, /* key destructor */
488 NULL /* val destructor */
489 };
490
491 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
492 dictType zsetDictType = {
493 dictEncObjHash, /* hash function */
494 NULL, /* key dup */
495 NULL, /* val dup */
496 dictEncObjKeyCompare, /* key compare */
497 dictRedisObjectDestructor, /* key destructor */
498 NULL /* val destructor */
499 };
500
501 /* Db->dict, keys are sds strings, vals are Redis objects. */
502 dictType dbDictType = {
503 dictSdsHash, /* hash function */
504 NULL, /* key dup */
505 NULL, /* val dup */
506 dictSdsKeyCompare, /* key compare */
507 dictSdsDestructor, /* key destructor */
508 dictRedisObjectDestructor /* val destructor */
509 };
510
511 /* Db->expires */
512 dictType keyptrDictType = {
513 dictSdsHash, /* hash function */
514 NULL, /* key dup */
515 NULL, /* val dup */
516 dictSdsKeyCompare, /* key compare */
517 NULL, /* key destructor */
518 NULL /* val destructor */
519 };
520
521 /* Command table. sds string -> command struct pointer. */
522 dictType commandTableDictType = {
523 dictSdsCaseHash, /* hash function */
524 NULL, /* key dup */
525 NULL, /* val dup */
526 dictSdsKeyCaseCompare, /* key compare */
527 dictSdsDestructor, /* key destructor */
528 NULL /* val destructor */
529 };
530
531 /* Hash type hash table (note that small hashes are represented with zimpaps) */
532 dictType hashDictType = {
533 dictEncObjHash, /* hash function */
534 NULL, /* key dup */
535 NULL, /* val dup */
536 dictEncObjKeyCompare, /* key compare */
537 dictRedisObjectDestructor, /* key destructor */
538 dictRedisObjectDestructor /* val destructor */
539 };
540
541 /* Keylist hash table type has unencoded redis objects as keys and
542 * lists as values. It's used for blocking operations (BLPOP) and to
543 * map swapped keys to a list of clients waiting for this keys to be loaded. */
544 dictType keylistDictType = {
545 dictObjHash, /* hash function */
546 NULL, /* key dup */
547 NULL, /* val dup */
548 dictObjKeyCompare, /* key compare */
549 dictRedisObjectDestructor, /* key destructor */
550 dictListDestructor /* val destructor */
551 };
552
553 /* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
554 * clusterNode structures. */
555 dictType clusterNodesDictType = {
556 dictSdsHash, /* hash function */
557 NULL, /* key dup */
558 NULL, /* val dup */
559 dictSdsKeyCompare, /* key compare */
560 dictSdsDestructor, /* key destructor */
561 NULL /* val destructor */
562 };
563
564 /* Migrate cache dict type. */
565 dictType migrateCacheDictType = {
566 dictSdsHash, /* hash function */
567 NULL, /* key dup */
568 NULL, /* val dup */
569 dictSdsKeyCompare, /* key compare */
570 dictSdsDestructor, /* key destructor */
571 NULL /* val destructor */
572 };
573
574 int htNeedsResize(dict *dict) {
575 long long size, used;
576
577 size = dictSlots(dict);
578 used = dictSize(dict);
579 return (size && used && size > DICT_HT_INITIAL_SIZE &&
580 (used*100/size < REDIS_HT_MINFILL));
581 }
582
583 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
584 * we resize the hash table to save memory */
585 void tryResizeHashTables(void) {
586 int j;
587
588 for (j = 0; j < server.dbnum; j++) {
589 if (htNeedsResize(server.db[j].dict))
590 dictResize(server.db[j].dict);
591 if (htNeedsResize(server.db[j].expires))
592 dictResize(server.db[j].expires);
593 }
594 }
595
596 /* Our hash table implementation performs rehashing incrementally while
597 * we write/read from the hash table. Still if the server is idle, the hash
598 * table will use two tables for a long time. So we try to use 1 millisecond
599 * of CPU time at every serverCron() loop in order to rehash some key. */
600 void incrementallyRehash(void) {
601 int j;
602
603 for (j = 0; j < server.dbnum; j++) {
604 /* Keys dictionary */
605 if (dictIsRehashing(server.db[j].dict)) {
606 dictRehashMilliseconds(server.db[j].dict,1);
607 break; /* already used our millisecond for this loop... */
608 }
609 /* Expires */
610 if (dictIsRehashing(server.db[j].expires)) {
611 dictRehashMilliseconds(server.db[j].expires,1);
612 break; /* already used our millisecond for this loop... */
613 }
614 }
615 }
616
617 /* This function is called once a background process of some kind terminates,
618 * as we want to avoid resizing the hash tables when there is a child in order
619 * to play well with copy-on-write (otherwise when a resize happens lots of
620 * memory pages are copied). The goal of this function is to update the ability
621 * for dict.c to resize the hash tables accordingly to the fact we have o not
622 * running childs. */
623 void updateDictResizePolicy(void) {
624 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
625 dictEnableResize();
626 else
627 dictDisableResize();
628 }
629
630 /* ======================= Cron: called every 100 ms ======================== */
631
632 /* Try to expire a few timed out keys. The algorithm used is adaptive and
633 * will use few CPU cycles if there are few expiring keys, otherwise
634 * it will get more aggressive to avoid that too much memory is used by
635 * keys that can be removed from the keyspace. */
636 void activeExpireCycle(void) {
637 int j, iteration = 0;
638 long long start = ustime(), timelimit;
639
640 /* We can use at max REDIS_EXPIRELOOKUPS_TIME_PERC percentage of CPU time
641 * per iteration. Since this function gets called with a frequency of
642 * REDIS_HZ times per second, the following is the max amount of
643 * microseconds we can spend in this function. */
644 timelimit = 1000000*REDIS_EXPIRELOOKUPS_TIME_PERC/REDIS_HZ/100;
645 if (timelimit <= 0) timelimit = 1;
646
647 for (j = 0; j < server.dbnum; j++) {
648 int expired;
649 redisDb *db = server.db+j;
650
651 /* Continue to expire if at the end of the cycle more than 25%
652 * of the keys were expired. */
653 do {
654 unsigned long num = dictSize(db->expires);
655 unsigned long slots = dictSlots(db->expires);
656 long long now = mstime();
657
658 /* When there are less than 1% filled slots getting random
659 * keys is expensive, so stop here waiting for better times...
660 * The dictionary will be resized asap. */
661 if (num && slots > DICT_HT_INITIAL_SIZE &&
662 (num*100/slots < 1)) break;
663
664 /* The main collection cycle. Sample random keys among keys
665 * with an expire set, checking for expired ones. */
666 expired = 0;
667 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
668 num = REDIS_EXPIRELOOKUPS_PER_CRON;
669 while (num--) {
670 dictEntry *de;
671 long long t;
672
673 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
674 t = dictGetSignedIntegerVal(de);
675 if (now > t) {
676 sds key = dictGetKey(de);
677 robj *keyobj = createStringObject(key,sdslen(key));
678
679 propagateExpire(db,keyobj);
680 dbDelete(db,keyobj);
681 decrRefCount(keyobj);
682 expired++;
683 server.stat_expiredkeys++;
684 }
685 }
686 /* We can't block forever here even if there are many keys to
687 * expire. So after a given amount of milliseconds return to the
688 * caller waiting for the other active expire cycle. */
689 iteration++;
690 if ((iteration & 0xf) == 0 && /* check once every 16 cycles. */
691 (ustime()-start) > timelimit) return;
692 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
693 }
694 }
695
696 void updateLRUClock(void) {
697 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
698 REDIS_LRU_CLOCK_MAX;
699 }
700
701
702 /* Add a sample to the operations per second array of samples. */
703 void trackOperationsPerSecond(void) {
704 long long t = mstime() - server.ops_sec_last_sample_time;
705 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
706 long long ops_sec;
707
708 ops_sec = t > 0 ? (ops*1000/t) : 0;
709
710 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
711 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
712 server.ops_sec_last_sample_time = mstime();
713 server.ops_sec_last_sample_ops = server.stat_numcommands;
714 }
715
716 /* Return the mean of all the samples. */
717 long long getOperationsPerSecond(void) {
718 int j;
719 long long sum = 0;
720
721 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
722 sum += server.ops_sec_samples[j];
723 return sum / REDIS_OPS_SEC_SAMPLES;
724 }
725
726 /* Check for timeouts. Returns non-zero if the client was terminated */
727 int clientsCronHandleTimeout(redisClient *c) {
728 time_t now = server.unixtime;
729
730 if (server.maxidletime &&
731 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
732 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
733 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
734 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
735 listLength(c->pubsub_patterns) == 0 &&
736 (now - c->lastinteraction > server.maxidletime))
737 {
738 redisLog(REDIS_VERBOSE,"Closing idle client");
739 freeClient(c);
740 return 1;
741 } else if (c->flags & REDIS_BLOCKED) {
742 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
743 addReply(c,shared.nullmultibulk);
744 unblockClientWaitingData(c);
745 }
746 }
747 return 0;
748 }
749
750 /* The client query buffer is an sds.c string that can end with a lot of
751 * free space not used, this function reclaims space if needed.
752 *
753 * The funciton always returns 0 as it never terminates the client. */
754 int clientsCronResizeQueryBuffer(redisClient *c) {
755 size_t querybuf_size = sdsAllocSize(c->querybuf);
756 time_t idletime = server.unixtime - c->lastinteraction;
757
758 /* There are two conditions to resize the query buffer:
759 * 1) Query buffer is > BIG_ARG and too big for latest peak.
760 * 2) Client is inactive and the buffer is bigger than 1k. */
761 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
762 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
763 (querybuf_size > 1024 && idletime > 2))
764 {
765 /* Only resize the query buffer if it is actually wasting space. */
766 if (sdsavail(c->querybuf) > 1024) {
767 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
768 }
769 }
770 /* Reset the peak again to capture the peak memory usage in the next
771 * cycle. */
772 c->querybuf_peak = 0;
773 return 0;
774 }
775
776 void clientsCron(void) {
777 /* Make sure to process at least 1/(REDIS_HZ*10) of clients per call.
778 * Since this function is called REDIS_HZ times per second we are sure that
779 * in the worst case we process all the clients in 10 seconds.
780 * In normal conditions (a reasonable number of clients) we process
781 * all the clients in a shorter time. */
782 int numclients = listLength(server.clients);
783 int iterations = numclients/(REDIS_HZ*10);
784
785 if (iterations < 50)
786 iterations = (numclients < 50) ? numclients : 50;
787 while(listLength(server.clients) && iterations--) {
788 redisClient *c;
789 listNode *head;
790
791 /* Rotate the list, take the current head, process.
792 * This way if the client must be removed from the list it's the
793 * first element and we don't incur into O(N) computation. */
794 listRotate(server.clients);
795 head = listFirst(server.clients);
796 c = listNodeValue(head);
797 /* The following functions do different service checks on the client.
798 * The protocol is that they return non-zero if the client was
799 * terminated. */
800 if (clientsCronHandleTimeout(c)) continue;
801 if (clientsCronResizeQueryBuffer(c)) continue;
802 }
803 }
804
805 /* This is our timer interrupt, called REDIS_HZ times per second.
806 * Here is where we do a number of things that need to be done asynchronously.
807 * For instance:
808 *
809 * - Active expired keys collection (it is also performed in a lazy way on
810 * lookup).
811 * - Software watchdong.
812 * - Update some statistic.
813 * - Incremental rehashing of the DBs hash tables.
814 * - Triggering BGSAVE / AOF rewrite, and handling of terminated children.
815 * - Clients timeout of differnet kinds.
816 * - Replication reconnection.
817 * - Many more...
818 *
819 * Everything directly called here will be called REDIS_HZ times per second,
820 * so in order to throttle execution of things we want to do less frequently
821 * a macro is used: run_with_period(milliseconds) { .... }
822 */
823
824 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
825 int j;
826 REDIS_NOTUSED(eventLoop);
827 REDIS_NOTUSED(id);
828 REDIS_NOTUSED(clientData);
829
830 /* Software watchdog: deliver the SIGALRM that will reach the signal
831 * handler if we don't return here fast enough. */
832 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
833
834 /* We take a cached value of the unix time in the global state because
835 * with virtual memory and aging there is to store the current time
836 * in objects at every object access, and accuracy is not needed.
837 * To access a global var is faster than calling time(NULL) */
838 server.unixtime = time(NULL);
839
840 run_with_period(100) trackOperationsPerSecond();
841
842 /* We have just 22 bits per object for LRU information.
843 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
844 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
845 *
846 * Note that even if this will wrap after 1.5 years it's not a problem,
847 * everything will still work but just some object will appear younger
848 * to Redis. But for this to happen a given object should never be touched
849 * for 1.5 years.
850 *
851 * Note that you can change the resolution altering the
852 * REDIS_LRU_CLOCK_RESOLUTION define.
853 */
854 updateLRUClock();
855
856 /* Record the max memory used since the server was started. */
857 if (zmalloc_used_memory() > server.stat_peak_memory)
858 server.stat_peak_memory = zmalloc_used_memory();
859
860 /* We received a SIGTERM, shutting down here in a safe way, as it is
861 * not ok doing so inside the signal handler. */
862 if (server.shutdown_asap) {
863 if (prepareForShutdown(0) == REDIS_OK) exit(0);
864 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
865 }
866
867 /* Show some info about non-empty databases */
868 run_with_period(5000) {
869 for (j = 0; j < server.dbnum; j++) {
870 long long size, used, vkeys;
871
872 size = dictSlots(server.db[j].dict);
873 used = dictSize(server.db[j].dict);
874 vkeys = dictSize(server.db[j].expires);
875 if (used || vkeys) {
876 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
877 /* dictPrintStats(server.dict); */
878 }
879 }
880 }
881
882 /* We don't want to resize the hash tables while a bacground saving
883 * is in progress: the saving child is created using fork() that is
884 * implemented with a copy-on-write semantic in most modern systems, so
885 * if we resize the HT while there is the saving child at work actually
886 * a lot of memory movements in the parent will cause a lot of pages
887 * copied. */
888 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
889 tryResizeHashTables();
890 if (server.activerehashing) incrementallyRehash();
891 }
892
893 /* Show information about connected clients */
894 if (!server.sentinel_mode) {
895 run_with_period(5000) {
896 redisLog(REDIS_VERBOSE,
897 "%d clients connected (%d slaves), %zu bytes in use",
898 listLength(server.clients)-listLength(server.slaves),
899 listLength(server.slaves),
900 zmalloc_used_memory());
901 }
902 }
903
904 /* We need to do a few operations on clients asynchronously. */
905 clientsCron();
906
907 /* Start a scheduled AOF rewrite if this was requested by the user while
908 * a BGSAVE was in progress. */
909 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
910 server.aof_rewrite_scheduled)
911 {
912 rewriteAppendOnlyFileBackground();
913 }
914
915 /* Check if a background saving or AOF rewrite in progress terminated. */
916 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
917 int statloc;
918 pid_t pid;
919
920 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
921 int exitcode = WEXITSTATUS(statloc);
922 int bysignal = 0;
923
924 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
925
926 if (pid == server.rdb_child_pid) {
927 backgroundSaveDoneHandler(exitcode,bysignal);
928 } else if (pid == server.aof_child_pid) {
929 backgroundRewriteDoneHandler(exitcode,bysignal);
930 } else {
931 redisLog(REDIS_WARNING,
932 "Warning, detected child with unmatched pid: %ld",
933 (long)pid);
934 }
935 updateDictResizePolicy();
936 }
937 } else {
938 /* If there is not a background saving/rewrite in progress check if
939 * we have to save/rewrite now */
940 for (j = 0; j < server.saveparamslen; j++) {
941 struct saveparam *sp = server.saveparams+j;
942
943 if (server.dirty >= sp->changes &&
944 server.unixtime-server.lastsave > sp->seconds) {
945 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
946 sp->changes, sp->seconds);
947 rdbSaveBackground(server.rdb_filename);
948 break;
949 }
950 }
951
952 /* Trigger an AOF rewrite if needed */
953 if (server.rdb_child_pid == -1 &&
954 server.aof_child_pid == -1 &&
955 server.aof_rewrite_perc &&
956 server.aof_current_size > server.aof_rewrite_min_size)
957 {
958 long long base = server.aof_rewrite_base_size ?
959 server.aof_rewrite_base_size : 1;
960 long long growth = (server.aof_current_size*100/base) - 100;
961 if (growth >= server.aof_rewrite_perc) {
962 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
963 rewriteAppendOnlyFileBackground();
964 }
965 }
966 }
967
968
969 /* If we postponed an AOF buffer flush, let's try to do it every time the
970 * cron function is called. */
971 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
972
973 /* Expire a few keys per cycle, only if this is a master.
974 * On slaves we wait for DEL operations synthesized by the master
975 * in order to guarantee a strict consistency. */
976 if (server.masterhost == NULL) activeExpireCycle();
977
978 /* Close clients that need to be closed asynchronous */
979 freeClientsInAsyncFreeQueue();
980
981 /* Replication cron function -- used to reconnect to master and
982 * to detect transfer failures. */
983 run_with_period(1000) replicationCron();
984
985 /* Run the Redis Cluster cron. */
986 run_with_period(1000) {
987 if (server.cluster_enabled) clusterCron();
988 }
989
990 /* Run the Sentinel timer if we are in sentinel mode. */
991 run_with_period(100) {
992 if (server.sentinel_mode) sentinelTimer();
993 }
994
995 /* Cleanup expired MIGRATE cached sockets. */
996 run_with_period(1000) {
997 migrateCloseTimedoutSockets();
998 }
999
1000 server.cronloops++;
1001 return 1000/REDIS_HZ;
1002 }
1003
1004 /* This function gets called every time Redis is entering the
1005 * main loop of the event driven library, that is, before to sleep
1006 * for ready file descriptors. */
1007 void beforeSleep(struct aeEventLoop *eventLoop) {
1008 REDIS_NOTUSED(eventLoop);
1009 listNode *ln;
1010 redisClient *c;
1011
1012 /* Try to process pending commands for clients that were just unblocked. */
1013 while (listLength(server.unblocked_clients)) {
1014 ln = listFirst(server.unblocked_clients);
1015 redisAssert(ln != NULL);
1016 c = ln->value;
1017 listDelNode(server.unblocked_clients,ln);
1018 c->flags &= ~REDIS_UNBLOCKED;
1019
1020 /* Process remaining data in the input buffer. */
1021 if (c->querybuf && sdslen(c->querybuf) > 0) {
1022 server.current_client = c;
1023 processInputBuffer(c);
1024 server.current_client = NULL;
1025 }
1026 }
1027
1028 /* Write the AOF buffer on disk */
1029 flushAppendOnlyFile(0);
1030 }
1031
1032 /* =========================== Server initialization ======================== */
1033
1034 void createSharedObjects(void) {
1035 int j;
1036
1037 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
1038 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
1039 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
1040 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
1041 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
1042 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
1043 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
1044 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
1045 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
1046 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
1047 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
1048 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
1049 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
1050 "-WRONGTYPE Operation against a key holding the wrong kind of value\r\n"));
1051 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
1052 "-ERR no such key\r\n"));
1053 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
1054 "-ERR syntax error\r\n"));
1055 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
1056 "-ERR source and destination objects are the same\r\n"));
1057 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
1058 "-ERR index out of range\r\n"));
1059 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
1060 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
1061 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
1062 "-LOADING Redis is loading the dataset in memory\r\n"));
1063 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
1064 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
1065 shared.masterdownerr = createObject(REDIS_STRING,sdsnew(
1066 "-MASTERDOWN Link with MASTER is down and slave-serve-stale-data is set to 'no'.\r\n"));
1067 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
1068 "-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"));
1069 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
1070 "-READONLY You can't write against a read only slave.\r\n"));
1071 shared.oomerr = createObject(REDIS_STRING,sdsnew(
1072 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
1073 shared.space = createObject(REDIS_STRING,sdsnew(" "));
1074 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1075 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1076
1077 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1078 shared.select[j] = createObject(REDIS_STRING,
1079 sdscatprintf(sdsempty(),"select %d\r\n", j));
1080 }
1081 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1082 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1083 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1084 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1085 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1086 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1087 shared.del = createStringObject("DEL",3);
1088 shared.rpop = createStringObject("RPOP",4);
1089 shared.lpop = createStringObject("LPOP",4);
1090 shared.lpush = createStringObject("LPUSH",5);
1091 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1092 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1093 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1094 }
1095 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1096 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1097 sdscatprintf(sdsempty(),"*%d\r\n",j));
1098 shared.bulkhdr[j] = createObject(REDIS_STRING,
1099 sdscatprintf(sdsempty(),"$%d\r\n",j));
1100 }
1101 }
1102
1103 void initServerConfig() {
1104 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1105 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1106 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1107 server.port = REDIS_SERVERPORT;
1108 server.bindaddr = NULL;
1109 server.unixsocket = NULL;
1110 server.unixsocketperm = 0;
1111 server.ipfd = -1;
1112 server.sofd = -1;
1113 server.dbnum = REDIS_DEFAULT_DBNUM;
1114 server.verbosity = REDIS_NOTICE;
1115 server.maxidletime = REDIS_MAXIDLETIME;
1116 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1117 server.saveparams = NULL;
1118 server.loading = 0;
1119 server.logfile = NULL; /* NULL = log on standard output */
1120 server.syslog_enabled = 0;
1121 server.syslog_ident = zstrdup("redis");
1122 server.syslog_facility = LOG_LOCAL0;
1123 server.daemonize = 0;
1124 server.aof_state = REDIS_AOF_OFF;
1125 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1126 server.aof_no_fsync_on_rewrite = 0;
1127 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1128 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1129 server.aof_rewrite_base_size = 0;
1130 server.aof_rewrite_scheduled = 0;
1131 server.aof_last_fsync = time(NULL);
1132 server.aof_rewrite_time_last = -1;
1133 server.aof_rewrite_time_start = -1;
1134 server.aof_lastbgrewrite_status = REDIS_OK;
1135 server.aof_delayed_fsync = 0;
1136 server.aof_fd = -1;
1137 server.aof_selected_db = -1; /* Make sure the first time will not match */
1138 server.aof_flush_postponed_start = 0;
1139 server.pidfile = zstrdup("/var/run/redis.pid");
1140 server.rdb_filename = zstrdup("dump.rdb");
1141 server.aof_filename = zstrdup("appendonly.aof");
1142 server.requirepass = NULL;
1143 server.rdb_compression = 1;
1144 server.rdb_checksum = 1;
1145 server.activerehashing = 1;
1146 server.maxclients = REDIS_MAX_CLIENTS;
1147 server.bpop_blocked_clients = 0;
1148 server.maxmemory = 0;
1149 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1150 server.maxmemory_samples = 3;
1151 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1152 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1153 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1154 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1155 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1156 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1157 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1158 server.shutdown_asap = 0;
1159 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1160 server.repl_timeout = REDIS_REPL_TIMEOUT;
1161 server.cluster_enabled = 0;
1162 server.cluster.configfile = zstrdup("nodes.conf");
1163 server.lua_caller = NULL;
1164 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1165 server.lua_client = NULL;
1166 server.lua_timedout = 0;
1167 server.migrate_cached_sockets = dictCreate(&migrateCacheDictType,NULL);
1168
1169 updateLRUClock();
1170 resetServerSaveParams();
1171
1172 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1173 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1174 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1175 /* Replication related */
1176 server.masterauth = NULL;
1177 server.masterhost = NULL;
1178 server.masterport = 6379;
1179 server.master = NULL;
1180 server.repl_state = REDIS_REPL_NONE;
1181 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1182 server.repl_serve_stale_data = 1;
1183 server.repl_slave_ro = 1;
1184 server.repl_down_since = time(NULL);
1185 server.slave_priority = REDIS_DEFAULT_SLAVE_PRIORITY;
1186
1187 /* Client output buffer limits */
1188 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1189 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1190 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1191 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1192 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1193 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1194 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1195 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1196 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1197
1198 /* Double constants initialization */
1199 R_Zero = 0.0;
1200 R_PosInf = 1.0/R_Zero;
1201 R_NegInf = -1.0/R_Zero;
1202 R_Nan = R_Zero/R_Zero;
1203
1204 /* Command table -- we intiialize it here as it is part of the
1205 * initial configuration, since command names may be changed via
1206 * redis.conf using the rename-command directive. */
1207 server.commands = dictCreate(&commandTableDictType,NULL);
1208 populateCommandTable();
1209 server.delCommand = lookupCommandByCString("del");
1210 server.multiCommand = lookupCommandByCString("multi");
1211 server.lpushCommand = lookupCommandByCString("lpush");
1212 server.lpopCommand = lookupCommandByCString("lpop");
1213 server.rpopCommand = lookupCommandByCString("rpop");
1214
1215 /* Slow log */
1216 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1217 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1218
1219 /* Debugging */
1220 server.assert_failed = "<no assertion failed>";
1221 server.assert_file = "<no file>";
1222 server.assert_line = 0;
1223 server.bug_report_start = 0;
1224 server.watchdog_period = 0;
1225 }
1226
1227 /* This function will try to raise the max number of open files accordingly to
1228 * the configured max number of clients. It will also account for 32 additional
1229 * file descriptors as we need a few more for persistence, listening
1230 * sockets, log files and so forth.
1231 *
1232 * If it will not be possible to set the limit accordingly to the configured
1233 * max number of clients, the function will do the reverse setting
1234 * server.maxclients to the value that we can actually handle. */
1235 void adjustOpenFilesLimit(void) {
1236 rlim_t maxfiles = server.maxclients+32;
1237 struct rlimit limit;
1238
1239 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1240 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1241 strerror(errno));
1242 server.maxclients = 1024-32;
1243 } else {
1244 rlim_t oldlimit = limit.rlim_cur;
1245
1246 /* Set the max number of files if the current limit is not enough
1247 * for our needs. */
1248 if (oldlimit < maxfiles) {
1249 rlim_t f;
1250
1251 f = maxfiles;
1252 while(f > oldlimit) {
1253 limit.rlim_cur = f;
1254 limit.rlim_max = f;
1255 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1256 f -= 128;
1257 }
1258 if (f < oldlimit) f = oldlimit;
1259 if (f != maxfiles) {
1260 server.maxclients = f-32;
1261 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1262 (int) maxfiles, strerror(errno), (int) server.maxclients);
1263 } else {
1264 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1265 (int) maxfiles);
1266 }
1267 }
1268 }
1269 }
1270
1271 void initServer() {
1272 int j;
1273
1274 signal(SIGHUP, SIG_IGN);
1275 signal(SIGPIPE, SIG_IGN);
1276 setupSignalHandlers();
1277
1278 if (server.syslog_enabled) {
1279 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1280 server.syslog_facility);
1281 }
1282
1283 server.current_client = NULL;
1284 server.clients = listCreate();
1285 server.clients_to_close = listCreate();
1286 server.slaves = listCreate();
1287 server.monitors = listCreate();
1288 server.unblocked_clients = listCreate();
1289 server.ready_keys = listCreate();
1290
1291 createSharedObjects();
1292 adjustOpenFilesLimit();
1293 server.el = aeCreateEventLoop(server.maxclients+1024);
1294 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1295
1296 if (server.port != 0) {
1297 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1298 if (server.ipfd == ANET_ERR) {
1299 redisLog(REDIS_WARNING, "Opening port %d: %s",
1300 server.port, server.neterr);
1301 exit(1);
1302 }
1303 }
1304 if (server.unixsocket != NULL) {
1305 unlink(server.unixsocket); /* don't care if this fails */
1306 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1307 if (server.sofd == ANET_ERR) {
1308 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1309 exit(1);
1310 }
1311 }
1312 if (server.ipfd < 0 && server.sofd < 0) {
1313 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1314 exit(1);
1315 }
1316 for (j = 0; j < server.dbnum; j++) {
1317 server.db[j].dict = dictCreate(&dbDictType,NULL);
1318 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1319 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1320 server.db[j].ready_keys = dictCreate(&setDictType,NULL);
1321 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1322 server.db[j].id = j;
1323 }
1324 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1325 server.pubsub_patterns = listCreate();
1326 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1327 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1328 server.cronloops = 0;
1329 server.rdb_child_pid = -1;
1330 server.aof_child_pid = -1;
1331 aofRewriteBufferReset();
1332 server.aof_buf = sdsempty();
1333 server.lastsave = time(NULL);
1334 server.rdb_save_time_last = -1;
1335 server.rdb_save_time_start = -1;
1336 server.dirty = 0;
1337 server.stat_numcommands = 0;
1338 server.stat_numconnections = 0;
1339 server.stat_expiredkeys = 0;
1340 server.stat_evictedkeys = 0;
1341 server.stat_starttime = time(NULL);
1342 server.stat_keyspace_misses = 0;
1343 server.stat_keyspace_hits = 0;
1344 server.stat_peak_memory = 0;
1345 server.stat_fork_time = 0;
1346 server.stat_rejected_conn = 0;
1347 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1348 server.ops_sec_idx = 0;
1349 server.ops_sec_last_sample_time = mstime();
1350 server.ops_sec_last_sample_ops = 0;
1351 server.unixtime = time(NULL);
1352 server.lastbgsave_status = REDIS_OK;
1353 server.stop_writes_on_bgsave_err = 1;
1354 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1355 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1356 acceptTcpHandler,NULL) == AE_ERR) redisPanic("Unrecoverable error creating server.ipfd file event.");
1357 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1358 acceptUnixHandler,NULL) == AE_ERR) redisPanic("Unrecoverable error creating server.sofd file event.");
1359
1360 if (server.aof_state == REDIS_AOF_ON) {
1361 server.aof_fd = open(server.aof_filename,
1362 O_WRONLY|O_APPEND|O_CREAT,0644);
1363 if (server.aof_fd == -1) {
1364 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1365 strerror(errno));
1366 exit(1);
1367 }
1368 }
1369
1370 /* 32 bit instances are limited to 4GB of address space, so if there is
1371 * no explicit limit in the user provided configuration we set a limit
1372 * at 3 GB using maxmemory with 'noeviction' policy'. This avoids
1373 * useless crashes of the Redis instance for out of memory. */
1374 if (server.arch_bits == 32 && server.maxmemory == 0) {
1375 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3 GB maxmemory limit with 'noeviction' policy now.");
1376 server.maxmemory = 3072LL*(1024*1024); /* 3 GB */
1377 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1378 }
1379
1380 if (server.cluster_enabled) clusterInit();
1381 scriptingInit();
1382 slowlogInit();
1383 bioInit();
1384 }
1385
1386 /* Populates the Redis Command Table starting from the hard coded list
1387 * we have on top of redis.c file. */
1388 void populateCommandTable(void) {
1389 int j;
1390 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1391
1392 for (j = 0; j < numcommands; j++) {
1393 struct redisCommand *c = redisCommandTable+j;
1394 char *f = c->sflags;
1395 int retval;
1396
1397 while(*f != '\0') {
1398 switch(*f) {
1399 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1400 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1401 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1402 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1403 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1404 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1405 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1406 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1407 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1408 case 'l': c->flags |= REDIS_CMD_LOADING; break;
1409 case 't': c->flags |= REDIS_CMD_STALE; break;
1410 case 'M': c->flags |= REDIS_CMD_SKIP_MONITOR; break;
1411 default: redisPanic("Unsupported command flag"); break;
1412 }
1413 f++;
1414 }
1415
1416 retval = dictAdd(server.commands, sdsnew(c->name), c);
1417 assert(retval == DICT_OK);
1418 }
1419 }
1420
1421 void resetCommandTableStats(void) {
1422 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1423 int j;
1424
1425 for (j = 0; j < numcommands; j++) {
1426 struct redisCommand *c = redisCommandTable+j;
1427
1428 c->microseconds = 0;
1429 c->calls = 0;
1430 }
1431 }
1432
1433 /* ========================== Redis OP Array API ============================ */
1434
1435 void redisOpArrayInit(redisOpArray *oa) {
1436 oa->ops = NULL;
1437 oa->numops = 0;
1438 }
1439
1440 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1441 robj **argv, int argc, int target)
1442 {
1443 redisOp *op;
1444
1445 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1446 op = oa->ops+oa->numops;
1447 op->cmd = cmd;
1448 op->dbid = dbid;
1449 op->argv = argv;
1450 op->argc = argc;
1451 op->target = target;
1452 oa->numops++;
1453 return oa->numops;
1454 }
1455
1456 void redisOpArrayFree(redisOpArray *oa) {
1457 while(oa->numops) {
1458 int j;
1459 redisOp *op;
1460
1461 oa->numops--;
1462 op = oa->ops+oa->numops;
1463 for (j = 0; j < op->argc; j++)
1464 decrRefCount(op->argv[j]);
1465 zfree(op->argv);
1466 }
1467 zfree(oa->ops);
1468 }
1469
1470 /* ====================== Commands lookup and execution ===================== */
1471
1472 struct redisCommand *lookupCommand(sds name) {
1473 return dictFetchValue(server.commands, name);
1474 }
1475
1476 struct redisCommand *lookupCommandByCString(char *s) {
1477 struct redisCommand *cmd;
1478 sds name = sdsnew(s);
1479
1480 cmd = dictFetchValue(server.commands, name);
1481 sdsfree(name);
1482 return cmd;
1483 }
1484
1485 /* Propagate the specified command (in the context of the specified database id)
1486 * to AOF and Slaves.
1487 *
1488 * flags are an xor between:
1489 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1490 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1491 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1492 */
1493 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1494 int flags)
1495 {
1496 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1497 feedAppendOnlyFile(cmd,dbid,argv,argc);
1498 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1499 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1500 }
1501
1502 /* Used inside commands to schedule the propagation of additional commands
1503 * after the current command is propagated to AOF / Replication. */
1504 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1505 int target)
1506 {
1507 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1508 }
1509
1510 /* Call() is the core of Redis execution of a command */
1511 void call(redisClient *c, int flags) {
1512 long long dirty, start = ustime(), duration;
1513
1514 /* Sent the command to clients in MONITOR mode, only if the commands are
1515 * not geneated from reading an AOF. */
1516 if (listLength(server.monitors) &&
1517 !server.loading &&
1518 !(c->cmd->flags & REDIS_CMD_SKIP_MONITOR))
1519 {
1520 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1521 }
1522
1523 /* Call the command. */
1524 redisOpArrayInit(&server.also_propagate);
1525 dirty = server.dirty;
1526 c->cmd->proc(c);
1527 dirty = server.dirty-dirty;
1528 duration = ustime()-start;
1529
1530 /* When EVAL is called loading the AOF we don't want commands called
1531 * from Lua to go into the slowlog or to populate statistics. */
1532 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1533 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1534
1535 /* Log the command into the Slow log if needed, and populate the
1536 * per-command statistics that we show in INFO commandstats. */
1537 if (flags & REDIS_CALL_SLOWLOG)
1538 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1539 if (flags & REDIS_CALL_STATS) {
1540 c->cmd->microseconds += duration;
1541 c->cmd->calls++;
1542 }
1543
1544 /* Propagate the command into the AOF and replication link */
1545 if (flags & REDIS_CALL_PROPAGATE) {
1546 int flags = REDIS_PROPAGATE_NONE;
1547
1548 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1549 flags |= REDIS_PROPAGATE_REPL;
1550 if (dirty)
1551 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1552 if (flags != REDIS_PROPAGATE_NONE)
1553 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1554 }
1555 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1556 * PUSH command. */
1557 if (server.also_propagate.numops) {
1558 int j;
1559 redisOp *rop;
1560
1561 for (j = 0; j < server.also_propagate.numops; j++) {
1562 rop = &server.also_propagate.ops[j];
1563 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1564 }
1565 redisOpArrayFree(&server.also_propagate);
1566 }
1567 server.stat_numcommands++;
1568 }
1569
1570 /* If this function gets called we already read a whole
1571 * command, arguments are in the client argv/argc fields.
1572 * processCommand() execute the command or prepare the
1573 * server for a bulk read from the client.
1574 *
1575 * If 1 is returned the client is still alive and valid and
1576 * and other operations can be performed by the caller. Otherwise
1577 * if 0 is returned the client was destroied (i.e. after QUIT). */
1578 int processCommand(redisClient *c) {
1579 /* The QUIT command is handled separately. Normal command procs will
1580 * go through checking for replication and QUIT will cause trouble
1581 * when FORCE_REPLICATION is enabled and would be implemented in
1582 * a regular command proc. */
1583 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1584 addReply(c,shared.ok);
1585 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1586 return REDIS_ERR;
1587 }
1588
1589 /* Now lookup the command and check ASAP about trivial error conditions
1590 * such as wrong arity, bad command name and so forth. */
1591 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1592 if (!c->cmd) {
1593 addReplyErrorFormat(c,"unknown command '%s'",
1594 (char*)c->argv[0]->ptr);
1595 return REDIS_OK;
1596 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1597 (c->argc < -c->cmd->arity)) {
1598 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1599 c->cmd->name);
1600 return REDIS_OK;
1601 }
1602
1603 /* Check if the user is authenticated */
1604 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1605 {
1606 addReplyError(c,"operation not permitted");
1607 return REDIS_OK;
1608 }
1609
1610 /* If cluster is enabled, redirect here */
1611 if (server.cluster_enabled &&
1612 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1613 int hashslot;
1614
1615 if (server.cluster.state != REDIS_CLUSTER_OK) {
1616 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1617 return REDIS_OK;
1618 } else {
1619 int ask;
1620 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1621 if (n == NULL) {
1622 addReplyError(c,"Multi keys request invalid in cluster");
1623 return REDIS_OK;
1624 } else if (n != server.cluster.myself) {
1625 addReplySds(c,sdscatprintf(sdsempty(),
1626 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1627 hashslot,n->ip,n->port));
1628 return REDIS_OK;
1629 }
1630 }
1631 }
1632
1633 /* Handle the maxmemory directive.
1634 *
1635 * First we try to free some memory if possible (if there are volatile
1636 * keys in the dataset). If there are not the only thing we can do
1637 * is returning an error. */
1638 if (server.maxmemory) {
1639 int retval = freeMemoryIfNeeded();
1640 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1641 addReply(c, shared.oomerr);
1642 return REDIS_OK;
1643 }
1644 }
1645
1646 /* Don't accept write commands if there are problems persisting on disk. */
1647 if (server.stop_writes_on_bgsave_err &&
1648 server.saveparamslen > 0
1649 && server.lastbgsave_status == REDIS_ERR &&
1650 c->cmd->flags & REDIS_CMD_WRITE)
1651 {
1652 addReply(c, shared.bgsaveerr);
1653 return REDIS_OK;
1654 }
1655
1656 /* Don't accept write commands if this is a read only slave. But
1657 * accept write commands if this is our master. */
1658 if (server.masterhost && server.repl_slave_ro &&
1659 !(c->flags & REDIS_MASTER) &&
1660 c->cmd->flags & REDIS_CMD_WRITE)
1661 {
1662 addReply(c, shared.roslaveerr);
1663 return REDIS_OK;
1664 }
1665
1666 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1667 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1668 &&
1669 c->cmd->proc != subscribeCommand &&
1670 c->cmd->proc != unsubscribeCommand &&
1671 c->cmd->proc != psubscribeCommand &&
1672 c->cmd->proc != punsubscribeCommand) {
1673 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1674 return REDIS_OK;
1675 }
1676
1677 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1678 * we are a slave with a broken link with master. */
1679 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1680 server.repl_serve_stale_data == 0 &&
1681 !(c->cmd->flags & REDIS_CMD_STALE))
1682 {
1683 addReply(c, shared.masterdownerr);
1684 return REDIS_OK;
1685 }
1686
1687 /* Loading DB? Return an error if the command has not the
1688 * REDIS_CMD_LOADING flag. */
1689 if (server.loading && !(c->cmd->flags & REDIS_CMD_LOADING)) {
1690 addReply(c, shared.loadingerr);
1691 return REDIS_OK;
1692 }
1693
1694 /* Lua script too slow? Only allow commands with REDIS_CMD_STALE flag. */
1695 if (server.lua_timedout &&
1696 c->cmd->proc != authCommand &&
1697 !(c->cmd->proc == shutdownCommand &&
1698 c->argc == 2 &&
1699 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1700 !(c->cmd->proc == scriptCommand &&
1701 c->argc == 2 &&
1702 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1703 {
1704 addReply(c, shared.slowscripterr);
1705 return REDIS_OK;
1706 }
1707
1708 /* Exec the command */
1709 if (c->flags & REDIS_MULTI &&
1710 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1711 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1712 {
1713 queueMultiCommand(c);
1714 addReply(c,shared.queued);
1715 } else {
1716 call(c,REDIS_CALL_FULL);
1717 if (listLength(server.ready_keys))
1718 handleClientsBlockedOnLists();
1719 }
1720 return REDIS_OK;
1721 }
1722
1723 /*================================== Shutdown =============================== */
1724
1725 int prepareForShutdown(int flags) {
1726 int save = flags & REDIS_SHUTDOWN_SAVE;
1727 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1728
1729 redisLog(REDIS_WARNING,"User requested shutdown...");
1730 /* Kill the saving child if there is a background saving in progress.
1731 We want to avoid race conditions, for instance our saving child may
1732 overwrite the synchronous saving did by SHUTDOWN. */
1733 if (server.rdb_child_pid != -1) {
1734 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1735 kill(server.rdb_child_pid,SIGKILL);
1736 rdbRemoveTempFile(server.rdb_child_pid);
1737 }
1738 if (server.aof_state != REDIS_AOF_OFF) {
1739 /* Kill the AOF saving child as the AOF we already have may be longer
1740 * but contains the full dataset anyway. */
1741 if (server.aof_child_pid != -1) {
1742 redisLog(REDIS_WARNING,
1743 "There is a child rewriting the AOF. Killing it!");
1744 kill(server.aof_child_pid,SIGKILL);
1745 }
1746 /* Append only file: fsync() the AOF and exit */
1747 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1748 aof_fsync(server.aof_fd);
1749 }
1750 if ((server.saveparamslen > 0 && !nosave) || save) {
1751 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1752 /* Snapshotting. Perform a SYNC SAVE and exit */
1753 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1754 /* Ooops.. error saving! The best we can do is to continue
1755 * operating. Note that if there was a background saving process,
1756 * in the next cron() Redis will be notified that the background
1757 * saving aborted, handling special stuff like slaves pending for
1758 * synchronization... */
1759 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1760 return REDIS_ERR;
1761 }
1762 }
1763 if (server.daemonize) {
1764 redisLog(REDIS_NOTICE,"Removing the pid file.");
1765 unlink(server.pidfile);
1766 }
1767 /* Close the listening sockets. Apparently this allows faster restarts. */
1768 if (server.ipfd != -1) close(server.ipfd);
1769 if (server.sofd != -1) close(server.sofd);
1770 if (server.unixsocket) {
1771 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1772 unlink(server.unixsocket); /* don't care if this fails */
1773 }
1774
1775 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1776 return REDIS_OK;
1777 }
1778
1779 /*================================== Commands =============================== */
1780
1781 /* Return zero if strings are the same, non-zero if they are not.
1782 * The comparison is performed in a way that prevents an attacker to obtain
1783 * information about the nature of the strings just monitoring the execution
1784 * time of the function.
1785 *
1786 * Note that limiting the comparison length to strings up to 512 bytes we
1787 * can avoid leaking any information about the password length and any
1788 * possible branch misprediction related leak.
1789 */
1790 int time_independent_strcmp(char *a, char *b) {
1791 char bufa[REDIS_AUTHPASS_MAX_LEN], bufb[REDIS_AUTHPASS_MAX_LEN];
1792 /* The above two strlen perform len(a) + len(b) operations where either
1793 * a or b are fixed (our password) length, and the difference is only
1794 * relative to the length of the user provided string, so no information
1795 * leak is possible in the following two lines of code. */
1796 int alen = strlen(a);
1797 int blen = strlen(b);
1798 int j;
1799 int diff = 0;
1800
1801 /* We can't compare strings longer than our static buffers.
1802 * Note that this will never pass the first test in practical circumstances
1803 * so there is no info leak. */
1804 if (alen > sizeof(bufa) || blen > sizeof(bufb)) return 1;
1805
1806 memset(bufa,0,sizeof(bufa)); /* Constant time. */
1807 memset(bufb,0,sizeof(bufb)); /* Constant time. */
1808 /* Again the time of the following two copies is proportional to
1809 * len(a) + len(b) so no info is leaked. */
1810 memcpy(bufa,a,alen);
1811 memcpy(bufb,b,blen);
1812
1813 /* Always compare all the chars in the two buffers without
1814 * conditional expressions. */
1815 for (j = 0; j < sizeof(bufa); j++) {
1816 diff |= (bufa[j] ^ bufb[j]);
1817 }
1818 /* Length must be equal as well. */
1819 diff |= alen ^ blen;
1820 return diff; /* If zero strings are the same. */
1821 }
1822
1823 void authCommand(redisClient *c) {
1824 if (!server.requirepass) {
1825 addReplyError(c,"Client sent AUTH, but no password is set");
1826 } else if (!time_independent_strcmp(c->argv[1]->ptr, server.requirepass)) {
1827 c->authenticated = 1;
1828 addReply(c,shared.ok);
1829 } else {
1830 c->authenticated = 0;
1831 addReplyError(c,"invalid password");
1832 }
1833 }
1834
1835 void pingCommand(redisClient *c) {
1836 addReply(c,shared.pong);
1837 }
1838
1839 void echoCommand(redisClient *c) {
1840 addReplyBulk(c,c->argv[1]);
1841 }
1842
1843 void timeCommand(redisClient *c) {
1844 struct timeval tv;
1845
1846 /* gettimeofday() can only fail if &tv is a bad addresss so we
1847 * don't check for errors. */
1848 gettimeofday(&tv,NULL);
1849 addReplyMultiBulkLen(c,2);
1850 addReplyBulkLongLong(c,tv.tv_sec);
1851 addReplyBulkLongLong(c,tv.tv_usec);
1852 }
1853
1854 /* Convert an amount of bytes into a human readable string in the form
1855 * of 100B, 2G, 100M, 4K, and so forth. */
1856 void bytesToHuman(char *s, unsigned long long n) {
1857 double d;
1858
1859 if (n < 1024) {
1860 /* Bytes */
1861 sprintf(s,"%lluB",n);
1862 return;
1863 } else if (n < (1024*1024)) {
1864 d = (double)n/(1024);
1865 sprintf(s,"%.2fK",d);
1866 } else if (n < (1024LL*1024*1024)) {
1867 d = (double)n/(1024*1024);
1868 sprintf(s,"%.2fM",d);
1869 } else if (n < (1024LL*1024*1024*1024)) {
1870 d = (double)n/(1024LL*1024*1024);
1871 sprintf(s,"%.2fG",d);
1872 }
1873 }
1874
1875 /* Create the string returned by the INFO command. This is decoupled
1876 * by the INFO command itself as we need to report the same information
1877 * on memory corruption problems. */
1878 sds genRedisInfoString(char *section) {
1879 sds info = sdsempty();
1880 time_t uptime = server.unixtime-server.stat_starttime;
1881 int j, numcommands;
1882 struct rusage self_ru, c_ru;
1883 unsigned long lol, bib;
1884 int allsections = 0, defsections = 0;
1885 int sections = 0;
1886
1887 if (section) {
1888 allsections = strcasecmp(section,"all") == 0;
1889 defsections = strcasecmp(section,"default") == 0;
1890 }
1891
1892 getrusage(RUSAGE_SELF, &self_ru);
1893 getrusage(RUSAGE_CHILDREN, &c_ru);
1894 getClientsMaxBuffers(&lol,&bib);
1895
1896 /* Server */
1897 if (allsections || defsections || !strcasecmp(section,"server")) {
1898 struct utsname name;
1899 char *mode;
1900
1901 if (server.cluster_enabled) mode = "cluster";
1902 else if (server.sentinel_mode) mode = "sentinel";
1903 else mode = "standalone";
1904
1905 if (sections++) info = sdscat(info,"\r\n");
1906 uname(&name);
1907 info = sdscatprintf(info,
1908 "# Server\r\n"
1909 "redis_version:%s\r\n"
1910 "redis_git_sha1:%s\r\n"
1911 "redis_git_dirty:%d\r\n"
1912 "redis_mode:%s\r\n"
1913 "os:%s %s %s\r\n"
1914 "arch_bits:%d\r\n"
1915 "multiplexing_api:%s\r\n"
1916 "gcc_version:%d.%d.%d\r\n"
1917 "process_id:%ld\r\n"
1918 "run_id:%s\r\n"
1919 "tcp_port:%d\r\n"
1920 "uptime_in_seconds:%ld\r\n"
1921 "uptime_in_days:%ld\r\n"
1922 "lru_clock:%ld\r\n",
1923 REDIS_VERSION,
1924 redisGitSHA1(),
1925 strtol(redisGitDirty(),NULL,10) > 0,
1926 mode,
1927 name.sysname, name.release, name.machine,
1928 server.arch_bits,
1929 aeGetApiName(),
1930 #ifdef __GNUC__
1931 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1932 #else
1933 0,0,0,
1934 #endif
1935 (long) getpid(),
1936 server.runid,
1937 server.port,
1938 uptime,
1939 uptime/(3600*24),
1940 (unsigned long) server.lruclock);
1941 }
1942
1943 /* Clients */
1944 if (allsections || defsections || !strcasecmp(section,"clients")) {
1945 if (sections++) info = sdscat(info,"\r\n");
1946 info = sdscatprintf(info,
1947 "# Clients\r\n"
1948 "connected_clients:%lu\r\n"
1949 "client_longest_output_list:%lu\r\n"
1950 "client_biggest_input_buf:%lu\r\n"
1951 "blocked_clients:%d\r\n",
1952 listLength(server.clients)-listLength(server.slaves),
1953 lol, bib,
1954 server.bpop_blocked_clients);
1955 }
1956
1957 /* Memory */
1958 if (allsections || defsections || !strcasecmp(section,"memory")) {
1959 char hmem[64];
1960 char peak_hmem[64];
1961
1962 bytesToHuman(hmem,zmalloc_used_memory());
1963 bytesToHuman(peak_hmem,server.stat_peak_memory);
1964 if (sections++) info = sdscat(info,"\r\n");
1965 info = sdscatprintf(info,
1966 "# Memory\r\n"
1967 "used_memory:%zu\r\n"
1968 "used_memory_human:%s\r\n"
1969 "used_memory_rss:%zu\r\n"
1970 "used_memory_peak:%zu\r\n"
1971 "used_memory_peak_human:%s\r\n"
1972 "used_memory_lua:%lld\r\n"
1973 "mem_fragmentation_ratio:%.2f\r\n"
1974 "mem_allocator:%s\r\n",
1975 zmalloc_used_memory(),
1976 hmem,
1977 zmalloc_get_rss(),
1978 server.stat_peak_memory,
1979 peak_hmem,
1980 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1981 zmalloc_get_fragmentation_ratio(),
1982 ZMALLOC_LIB
1983 );
1984 }
1985
1986 /* Persistence */
1987 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1988 if (sections++) info = sdscat(info,"\r\n");
1989 info = sdscatprintf(info,
1990 "# Persistence\r\n"
1991 "loading:%d\r\n"
1992 "rdb_changes_since_last_save:%lld\r\n"
1993 "rdb_bgsave_in_progress:%d\r\n"
1994 "rdb_last_save_time:%ld\r\n"
1995 "rdb_last_bgsave_status:%s\r\n"
1996 "rdb_last_bgsave_time_sec:%ld\r\n"
1997 "rdb_current_bgsave_time_sec:%ld\r\n"
1998 "aof_enabled:%d\r\n"
1999 "aof_rewrite_in_progress:%d\r\n"
2000 "aof_rewrite_scheduled:%d\r\n"
2001 "aof_last_rewrite_time_sec:%ld\r\n"
2002 "aof_current_rewrite_time_sec:%ld\r\n"
2003 "aof_last_bgrewrite_status:%s\r\n",
2004 server.loading,
2005 server.dirty,
2006 server.rdb_child_pid != -1,
2007 server.lastsave,
2008 (server.lastbgsave_status == REDIS_OK) ? "ok" : "err",
2009 server.rdb_save_time_last,
2010 (server.rdb_child_pid == -1) ?
2011 -1 : time(NULL)-server.rdb_save_time_start,
2012 server.aof_state != REDIS_AOF_OFF,
2013 server.aof_child_pid != -1,
2014 server.aof_rewrite_scheduled,
2015 server.aof_rewrite_time_last,
2016 (server.aof_child_pid == -1) ?
2017 -1 : time(NULL)-server.aof_rewrite_time_start,
2018 (server.aof_lastbgrewrite_status == REDIS_OK) ? "ok" : "err");
2019
2020 if (server.aof_state != REDIS_AOF_OFF) {
2021 info = sdscatprintf(info,
2022 "aof_current_size:%lld\r\n"
2023 "aof_base_size:%lld\r\n"
2024 "aof_pending_rewrite:%d\r\n"
2025 "aof_buffer_length:%zu\r\n"
2026 "aof_rewrite_buffer_length:%lu\r\n"
2027 "aof_pending_bio_fsync:%llu\r\n"
2028 "aof_delayed_fsync:%lu\r\n",
2029 (long long) server.aof_current_size,
2030 (long long) server.aof_rewrite_base_size,
2031 server.aof_rewrite_scheduled,
2032 sdslen(server.aof_buf),
2033 aofRewriteBufferSize(),
2034 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
2035 server.aof_delayed_fsync);
2036 }
2037
2038 if (server.loading) {
2039 double perc;
2040 time_t eta, elapsed;
2041 off_t remaining_bytes = server.loading_total_bytes-
2042 server.loading_loaded_bytes;
2043
2044 perc = ((double)server.loading_loaded_bytes /
2045 server.loading_total_bytes) * 100;
2046
2047 elapsed = server.unixtime-server.loading_start_time;
2048 if (elapsed == 0) {
2049 eta = 1; /* A fake 1 second figure if we don't have
2050 enough info */
2051 } else {
2052 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
2053 }
2054
2055 info = sdscatprintf(info,
2056 "loading_start_time:%ld\r\n"
2057 "loading_total_bytes:%llu\r\n"
2058 "loading_loaded_bytes:%llu\r\n"
2059 "loading_loaded_perc:%.2f\r\n"
2060 "loading_eta_seconds:%ld\r\n"
2061 ,(unsigned long) server.loading_start_time,
2062 (unsigned long long) server.loading_total_bytes,
2063 (unsigned long long) server.loading_loaded_bytes,
2064 perc,
2065 eta
2066 );
2067 }
2068 }
2069
2070 /* Stats */
2071 if (allsections || defsections || !strcasecmp(section,"stats")) {
2072 if (sections++) info = sdscat(info,"\r\n");
2073 info = sdscatprintf(info,
2074 "# Stats\r\n"
2075 "total_connections_received:%lld\r\n"
2076 "total_commands_processed:%lld\r\n"
2077 "instantaneous_ops_per_sec:%lld\r\n"
2078 "rejected_connections:%lld\r\n"
2079 "expired_keys:%lld\r\n"
2080 "evicted_keys:%lld\r\n"
2081 "keyspace_hits:%lld\r\n"
2082 "keyspace_misses:%lld\r\n"
2083 "pubsub_channels:%ld\r\n"
2084 "pubsub_patterns:%lu\r\n"
2085 "latest_fork_usec:%lld\r\n"
2086 "migrate_cached_sockets:%ld\r\n",
2087 server.stat_numconnections,
2088 server.stat_numcommands,
2089 getOperationsPerSecond(),
2090 server.stat_rejected_conn,
2091 server.stat_expiredkeys,
2092 server.stat_evictedkeys,
2093 server.stat_keyspace_hits,
2094 server.stat_keyspace_misses,
2095 dictSize(server.pubsub_channels),
2096 listLength(server.pubsub_patterns),
2097 server.stat_fork_time,
2098 dictSize(server.migrate_cached_sockets));
2099 }
2100
2101 /* Replication */
2102 if (allsections || defsections || !strcasecmp(section,"replication")) {
2103 if (sections++) info = sdscat(info,"\r\n");
2104 info = sdscatprintf(info,
2105 "# Replication\r\n"
2106 "role:%s\r\n",
2107 server.masterhost == NULL ? "master" : "slave");
2108 if (server.masterhost) {
2109 info = sdscatprintf(info,
2110 "master_host:%s\r\n"
2111 "master_port:%d\r\n"
2112 "master_link_status:%s\r\n"
2113 "master_last_io_seconds_ago:%d\r\n"
2114 "master_sync_in_progress:%d\r\n"
2115 ,server.masterhost,
2116 server.masterport,
2117 (server.repl_state == REDIS_REPL_CONNECTED) ?
2118 "up" : "down",
2119 server.master ?
2120 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
2121 server.repl_state == REDIS_REPL_TRANSFER
2122 );
2123
2124 if (server.repl_state == REDIS_REPL_TRANSFER) {
2125 info = sdscatprintf(info,
2126 "master_sync_left_bytes:%lld\r\n"
2127 "master_sync_last_io_seconds_ago:%d\r\n"
2128 , (long long)
2129 (server.repl_transfer_size - server.repl_transfer_read),
2130 (int)(server.unixtime-server.repl_transfer_lastio)
2131 );
2132 }
2133
2134 if (server.repl_state != REDIS_REPL_CONNECTED) {
2135 info = sdscatprintf(info,
2136 "master_link_down_since_seconds:%ld\r\n",
2137 (long)server.unixtime-server.repl_down_since);
2138 }
2139 info = sdscatprintf(info,
2140 "slave_priority:%d\r\n"
2141 "slave_read_only:%d\r\n",
2142 server.slave_priority,
2143 server.repl_slave_ro);
2144 }
2145 info = sdscatprintf(info,
2146 "connected_slaves:%lu\r\n",
2147 listLength(server.slaves));
2148 if (listLength(server.slaves)) {
2149 int slaveid = 0;
2150 listNode *ln;
2151 listIter li;
2152
2153 listRewind(server.slaves,&li);
2154 while((ln = listNext(&li))) {
2155 redisClient *slave = listNodeValue(ln);
2156 char *state = NULL;
2157 char ip[32];
2158 int port;
2159
2160 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
2161 switch(slave->replstate) {
2162 case REDIS_REPL_WAIT_BGSAVE_START:
2163 case REDIS_REPL_WAIT_BGSAVE_END:
2164 state = "wait_bgsave";
2165 break;
2166 case REDIS_REPL_SEND_BULK:
2167 state = "send_bulk";
2168 break;
2169 case REDIS_REPL_ONLINE:
2170 state = "online";
2171 break;
2172 }
2173 if (state == NULL) continue;
2174 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2175 slaveid,ip,slave->slave_listening_port,state);
2176 slaveid++;
2177 }
2178 }
2179 }
2180
2181 /* CPU */
2182 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2183 if (sections++) info = sdscat(info,"\r\n");
2184 info = sdscatprintf(info,
2185 "# CPU\r\n"
2186 "used_cpu_sys:%.2f\r\n"
2187 "used_cpu_user:%.2f\r\n"
2188 "used_cpu_sys_children:%.2f\r\n"
2189 "used_cpu_user_children:%.2f\r\n",
2190 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2191 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2192 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2193 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2194 }
2195
2196 /* cmdtime */
2197 if (allsections || !strcasecmp(section,"commandstats")) {
2198 if (sections++) info = sdscat(info,"\r\n");
2199 info = sdscatprintf(info, "# Commandstats\r\n");
2200 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2201 for (j = 0; j < numcommands; j++) {
2202 struct redisCommand *c = redisCommandTable+j;
2203
2204 if (!c->calls) continue;
2205 info = sdscatprintf(info,
2206 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2207 c->name, c->calls, c->microseconds,
2208 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2209 }
2210 }
2211
2212 /* Cluster */
2213 if (allsections || defsections || !strcasecmp(section,"cluster")) {
2214 if (sections++) info = sdscat(info,"\r\n");
2215 info = sdscatprintf(info,
2216 "# Cluster\r\n"
2217 "cluster_enabled:%d\r\n",
2218 server.cluster_enabled);
2219 }
2220
2221 /* Key space */
2222 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2223 if (sections++) info = sdscat(info,"\r\n");
2224 info = sdscatprintf(info, "# Keyspace\r\n");
2225 for (j = 0; j < server.dbnum; j++) {
2226 long long keys, vkeys;
2227
2228 keys = dictSize(server.db[j].dict);
2229 vkeys = dictSize(server.db[j].expires);
2230 if (keys || vkeys) {
2231 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2232 j, keys, vkeys);
2233 }
2234 }
2235 }
2236 return info;
2237 }
2238
2239 void infoCommand(redisClient *c) {
2240 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2241
2242 if (c->argc > 2) {
2243 addReply(c,shared.syntaxerr);
2244 return;
2245 }
2246 sds info = genRedisInfoString(section);
2247 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2248 (unsigned long)sdslen(info)));
2249 addReplySds(c,info);
2250 addReply(c,shared.crlf);
2251 }
2252
2253 void monitorCommand(redisClient *c) {
2254 /* ignore MONITOR if already slave or in monitor mode */
2255 if (c->flags & REDIS_SLAVE) return;
2256
2257 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2258 c->slaveseldb = 0;
2259 listAddNodeTail(server.monitors,c);
2260 addReply(c,shared.ok);
2261 }
2262
2263 /* ============================ Maxmemory directive ======================== */
2264
2265 /* This function gets called when 'maxmemory' is set on the config file to limit
2266 * the max memory used by the server, before processing a command.
2267 *
2268 * The goal of the function is to free enough memory to keep Redis under the
2269 * configured memory limit.
2270 *
2271 * The function starts calculating how many bytes should be freed to keep
2272 * Redis under the limit, and enters a loop selecting the best keys to
2273 * evict accordingly to the configured policy.
2274 *
2275 * If all the bytes needed to return back under the limit were freed the
2276 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2277 * should block the execution of commands that will result in more memory
2278 * used by the server.
2279 */
2280 int freeMemoryIfNeeded(void) {
2281 size_t mem_used, mem_tofree, mem_freed;
2282 int slaves = listLength(server.slaves);
2283
2284 /* Remove the size of slaves output buffers and AOF buffer from the
2285 * count of used memory. */
2286 mem_used = zmalloc_used_memory();
2287 if (slaves) {
2288 listIter li;
2289 listNode *ln;
2290
2291 listRewind(server.slaves,&li);
2292 while((ln = listNext(&li))) {
2293 redisClient *slave = listNodeValue(ln);
2294 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2295 if (obuf_bytes > mem_used)
2296 mem_used = 0;
2297 else
2298 mem_used -= obuf_bytes;
2299 }
2300 }
2301 if (server.aof_state != REDIS_AOF_OFF) {
2302 mem_used -= sdslen(server.aof_buf);
2303 mem_used -= aofRewriteBufferSize();
2304 }
2305
2306 /* Check if we are over the memory limit. */
2307 if (mem_used <= server.maxmemory) return REDIS_OK;
2308
2309 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2310 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2311
2312 /* Compute how much memory we need to free. */
2313 mem_tofree = mem_used - server.maxmemory;
2314 mem_freed = 0;
2315 while (mem_freed < mem_tofree) {
2316 int j, k, keys_freed = 0;
2317
2318 for (j = 0; j < server.dbnum; j++) {
2319 long bestval = 0; /* just to prevent warning */
2320 sds bestkey = NULL;
2321 struct dictEntry *de;
2322 redisDb *db = server.db+j;
2323 dict *dict;
2324
2325 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2326 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2327 {
2328 dict = server.db[j].dict;
2329 } else {
2330 dict = server.db[j].expires;
2331 }
2332 if (dictSize(dict) == 0) continue;
2333
2334 /* volatile-random and allkeys-random policy */
2335 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2336 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2337 {
2338 de = dictGetRandomKey(dict);
2339 bestkey = dictGetKey(de);
2340 }
2341
2342 /* volatile-lru and allkeys-lru policy */
2343 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2344 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2345 {
2346 for (k = 0; k < server.maxmemory_samples; k++) {
2347 sds thiskey;
2348 long thisval;
2349 robj *o;
2350
2351 de = dictGetRandomKey(dict);
2352 thiskey = dictGetKey(de);
2353 /* When policy is volatile-lru we need an additional lookup
2354 * to locate the real key, as dict is set to db->expires. */
2355 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2356 de = dictFind(db->dict, thiskey);
2357 o = dictGetVal(de);
2358 thisval = estimateObjectIdleTime(o);
2359
2360 /* Higher idle time is better candidate for deletion */
2361 if (bestkey == NULL || thisval > bestval) {
2362 bestkey = thiskey;
2363 bestval = thisval;
2364 }
2365 }
2366 }
2367
2368 /* volatile-ttl */
2369 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2370 for (k = 0; k < server.maxmemory_samples; k++) {
2371 sds thiskey;
2372 long thisval;
2373
2374 de = dictGetRandomKey(dict);
2375 thiskey = dictGetKey(de);
2376 thisval = (long) dictGetVal(de);
2377
2378 /* Expire sooner (minor expire unix timestamp) is better
2379 * candidate for deletion */
2380 if (bestkey == NULL || thisval < bestval) {
2381 bestkey = thiskey;
2382 bestval = thisval;
2383 }
2384 }
2385 }
2386
2387 /* Finally remove the selected key. */
2388 if (bestkey) {
2389 long long delta;
2390
2391 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2392 propagateExpire(db,keyobj);
2393 /* We compute the amount of memory freed by dbDelete() alone.
2394 * It is possible that actually the memory needed to propagate
2395 * the DEL in AOF and replication link is greater than the one
2396 * we are freeing removing the key, but we can't account for
2397 * that otherwise we would never exit the loop.
2398 *
2399 * AOF and Output buffer memory will be freed eventually so
2400 * we only care about memory used by the key space. */
2401 delta = (long long) zmalloc_used_memory();
2402 dbDelete(db,keyobj);
2403 delta -= (long long) zmalloc_used_memory();
2404 mem_freed += delta;
2405 server.stat_evictedkeys++;
2406 decrRefCount(keyobj);
2407 keys_freed++;
2408
2409 /* When the memory to free starts to be big enough, we may
2410 * start spending so much time here that is impossible to
2411 * deliver data to the slaves fast enough, so we force the
2412 * transmission here inside the loop. */
2413 if (slaves) flushSlavesOutputBuffers();
2414 }
2415 }
2416 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2417 }
2418 return REDIS_OK;
2419 }
2420
2421 /* =================================== Main! ================================ */
2422
2423 #ifdef __linux__
2424 int linuxOvercommitMemoryValue(void) {
2425 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2426 char buf[64];
2427
2428 if (!fp) return -1;
2429 if (fgets(buf,64,fp) == NULL) {
2430 fclose(fp);
2431 return -1;
2432 }
2433 fclose(fp);
2434
2435 return atoi(buf);
2436 }
2437
2438 void linuxOvercommitMemoryWarning(void) {
2439 if (linuxOvercommitMemoryValue() == 0) {
2440 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.");
2441 }
2442 }
2443 #endif /* __linux__ */
2444
2445 void createPidFile(void) {
2446 /* Try to write the pid file in a best-effort way. */
2447 FILE *fp = fopen(server.pidfile,"w");
2448 if (fp) {
2449 fprintf(fp,"%d\n",(int)getpid());
2450 fclose(fp);
2451 }
2452 }
2453
2454 void daemonize(void) {
2455 int fd;
2456
2457 if (fork() != 0) exit(0); /* parent exits */
2458 setsid(); /* create a new session */
2459
2460 /* Every output goes to /dev/null. If Redis is daemonized but
2461 * the 'logfile' is set to 'stdout' in the configuration file
2462 * it will not log at all. */
2463 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2464 dup2(fd, STDIN_FILENO);
2465 dup2(fd, STDOUT_FILENO);
2466 dup2(fd, STDERR_FILENO);
2467 if (fd > STDERR_FILENO) close(fd);
2468 }
2469 }
2470
2471 void version() {
2472 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2473 REDIS_VERSION,
2474 redisGitSHA1(),
2475 atoi(redisGitDirty()) > 0,
2476 ZMALLOC_LIB,
2477 sizeof(long) == 4 ? 32 : 64);
2478 exit(0);
2479 }
2480
2481 void usage() {
2482 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2483 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2484 fprintf(stderr," ./redis-server -v or --version\n");
2485 fprintf(stderr," ./redis-server -h or --help\n");
2486 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2487 fprintf(stderr,"Examples:\n");
2488 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2489 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2490 fprintf(stderr," ./redis-server --port 7777\n");
2491 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2492 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n\n");
2493 fprintf(stderr,"Sentinel mode:\n");
2494 fprintf(stderr," ./redis-server /etc/sentinel.conf --sentinel\n");
2495 exit(1);
2496 }
2497
2498 void redisAsciiArt(void) {
2499 #include "asciilogo.h"
2500 char *buf = zmalloc(1024*16);
2501 char *mode = "stand alone";
2502
2503 if (server.cluster_enabled) mode = "cluster";
2504 else if (server.sentinel_mode) mode = "sentinel";
2505
2506 snprintf(buf,1024*16,ascii_logo,
2507 REDIS_VERSION,
2508 redisGitSHA1(),
2509 strtol(redisGitDirty(),NULL,10) > 0,
2510 (sizeof(long) == 8) ? "64" : "32",
2511 mode, server.port,
2512 (long) getpid()
2513 );
2514 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2515 zfree(buf);
2516 }
2517
2518 static void sigtermHandler(int sig) {
2519 REDIS_NOTUSED(sig);
2520
2521 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2522 server.shutdown_asap = 1;
2523 }
2524
2525 void setupSignalHandlers(void) {
2526 struct sigaction act;
2527
2528 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2529 * Otherwise, sa_handler is used. */
2530 sigemptyset(&act.sa_mask);
2531 act.sa_flags = 0;
2532 act.sa_handler = sigtermHandler;
2533 sigaction(SIGTERM, &act, NULL);
2534
2535 #ifdef HAVE_BACKTRACE
2536 sigemptyset(&act.sa_mask);
2537 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2538 act.sa_sigaction = sigsegvHandler;
2539 sigaction(SIGSEGV, &act, NULL);
2540 sigaction(SIGBUS, &act, NULL);
2541 sigaction(SIGFPE, &act, NULL);
2542 sigaction(SIGILL, &act, NULL);
2543 #endif
2544 return;
2545 }
2546
2547 void memtest(size_t megabytes, int passes);
2548
2549 /* Returns 1 if there is --sentinel among the arguments or if
2550 * argv[0] is exactly "redis-sentinel". */
2551 int checkForSentinelMode(int argc, char **argv) {
2552 int j;
2553
2554 if (strstr(argv[0],"redis-sentinel") != NULL) return 1;
2555 for (j = 1; j < argc; j++)
2556 if (!strcmp(argv[j],"--sentinel")) return 1;
2557 return 0;
2558 }
2559
2560 /* Function called at startup to load RDB or AOF file in memory. */
2561 void loadDataFromDisk(void) {
2562 long long start = ustime();
2563 if (server.aof_state == REDIS_AOF_ON) {
2564 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2565 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2566 } else {
2567 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2568 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2569 (float)(ustime()-start)/1000000);
2570 } else if (errno != ENOENT) {
2571 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2572 exit(1);
2573 }
2574 }
2575 }
2576
2577 void redisOutOfMemoryHandler(size_t allocation_size) {
2578 redisLog(REDIS_WARNING,"Out Of Memory allocating %zu bytes!",
2579 allocation_size);
2580 redisPanic("OOM");
2581 }
2582
2583 int main(int argc, char **argv) {
2584 struct timeval tv;
2585
2586 /* We need to initialize our libraries, and the server configuration. */
2587 zmalloc_enable_thread_safeness();
2588 zmalloc_set_oom_handler(redisOutOfMemoryHandler);
2589 srand(time(NULL)^getpid());
2590 gettimeofday(&tv,NULL);
2591 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2592 server.sentinel_mode = checkForSentinelMode(argc,argv);
2593 initServerConfig();
2594
2595 /* We need to init sentinel right now as parsing the configuration file
2596 * in sentinel mode will have the effect of populating the sentinel
2597 * data structures with master nodes to monitor. */
2598 if (server.sentinel_mode) {
2599 initSentinelConfig();
2600 initSentinel();
2601 }
2602
2603 if (argc >= 2) {
2604 int j = 1; /* First option to parse in argv[] */
2605 sds options = sdsempty();
2606 char *configfile = NULL;
2607
2608 /* Handle special options --help and --version */
2609 if (strcmp(argv[1], "-v") == 0 ||
2610 strcmp(argv[1], "--version") == 0) version();
2611 if (strcmp(argv[1], "--help") == 0 ||
2612 strcmp(argv[1], "-h") == 0) usage();
2613 if (strcmp(argv[1], "--test-memory") == 0) {
2614 if (argc == 3) {
2615 memtest(atoi(argv[2]),50);
2616 exit(0);
2617 } else {
2618 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2619 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2620 exit(1);
2621 }
2622 }
2623
2624 /* First argument is the config file name? */
2625 if (argv[j][0] != '-' || argv[j][1] != '-')
2626 configfile = argv[j++];
2627 /* All the other options are parsed and conceptually appended to the
2628 * configuration file. For instance --port 6380 will generate the
2629 * string "port 6380\n" to be parsed after the actual file name
2630 * is parsed, if any. */
2631 while(j != argc) {
2632 if (argv[j][0] == '-' && argv[j][1] == '-') {
2633 /* Option name */
2634 if (sdslen(options)) options = sdscat(options,"\n");
2635 options = sdscat(options,argv[j]+2);
2636 options = sdscat(options," ");
2637 } else {
2638 /* Option argument */
2639 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2640 options = sdscat(options," ");
2641 }
2642 j++;
2643 }
2644 resetServerSaveParams();
2645 loadServerConfig(configfile,options);
2646 sdsfree(options);
2647 } else {
2648 redisLog(REDIS_WARNING, "Warning: no config file specified, using the default config. In order to specify a config file use %s /path/to/%s.conf", argv[0], server.sentinel_mode ? "sentinel" : "redis");
2649 }
2650 if (server.daemonize) daemonize();
2651 initServer();
2652 if (server.daemonize) createPidFile();
2653 redisAsciiArt();
2654
2655 if (!server.sentinel_mode) {
2656 /* Things only needed when not running in Sentinel mode. */
2657 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2658 #ifdef __linux__
2659 linuxOvercommitMemoryWarning();
2660 #endif
2661 loadDataFromDisk();
2662 if (server.ipfd > 0)
2663 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2664 if (server.sofd > 0)
2665 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2666 }
2667
2668 /* Warning the user about suspicious maxmemory setting. */
2669 if (server.maxmemory > 0 && server.maxmemory < 1024*1024) {
2670 redisLog(REDIS_WARNING,"WARNING: You specified a maxmemory value that is less than 1MB (current value is %llu bytes). Are you sure this is what you really want?", server.maxmemory);
2671 }
2672
2673 aeSetBeforeSleepProc(server.el,beforeSleep);
2674 aeMain(server.el);
2675 aeDeleteEventLoop(server.el);
2676 return 0;
2677 }
2678
2679 /* The End */