]> git.saurik.com Git - redis.git/blob - src/redis.c
First implementation of Redis Sentinel.
[redis.git] / src / redis.c
1 /*
2 * Copyright (c) 2009-2010, Salvatore Sanfilippo <antirez at gmail dot com>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are met:
7 *
8 * * Redistributions of source code must retain the above copyright notice,
9 * this list of conditions and the following disclaimer.
10 * * Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * * Neither the name of Redis nor the names of its contributors may be used
14 * to endorse or promote products derived from this software without
15 * specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27 * POSSIBILITY OF SUCH DAMAGE.
28 */
29
30 #include "redis.h"
31 #include "slowlog.h"
32 #include "bio.h"
33
34 #include <time.h>
35 #include <signal.h>
36 #include <sys/wait.h>
37 #include <errno.h>
38 #include <assert.h>
39 #include <ctype.h>
40 #include <stdarg.h>
41 #include <arpa/inet.h>
42 #include <sys/stat.h>
43 #include <fcntl.h>
44 #include <sys/time.h>
45 #include <sys/resource.h>
46 #include <sys/uio.h>
47 #include <limits.h>
48 #include <float.h>
49 #include <math.h>
50 #include <sys/resource.h>
51 #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 */
114 struct redisCommand redisCommandTable[] = {
115 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
116 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
117 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
118 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
119 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
120 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
121 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
122 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
123 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
124 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
125 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
126 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
127 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
128 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
129 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
130 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
131 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
132 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
133 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
134 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
135 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
136 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
137 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
138 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
139 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
140 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
141 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
142 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
143 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
144 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
145 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
146 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
147 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
148 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
149 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
150 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
151 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
152 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
153 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
154 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
155 {"srandmember",srandmemberCommand,-2,"rR",0,NULL,1,1,1,0,0},
156 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
157 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
158 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
159 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
160 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
161 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
162 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
163 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
164 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
165 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
166 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
167 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
168 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
169 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
170 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
171 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
172 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
173 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
174 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
175 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
176 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
177 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
178 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
179 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
180 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
182 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
183 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
184 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
185 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
186 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
187 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
188 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
189 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
190 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
191 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
192 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
193 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
194 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
195 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
196 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
197 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
198 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
199 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
200 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
201 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
202 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
203 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
204 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
205 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
206 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
207 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
208 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
209 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
210 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
211 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
212 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
213 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
214 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
215 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
216 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
217 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
218 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
219 {"exec",execCommand,1,"s",0,NULL,0,0,0,0,0},
220 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
221 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
222 {"replconf",replconfCommand,-1,"ars",0,NULL,0,0,0,0,0},
223 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
224 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
225 {"sort",sortCommand,-2,"wm",0,NULL,1,1,1,0,0},
226 {"info",infoCommand,-1,"rlt",0,NULL,0,0,0,0,0},
227 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
228 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
229 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
230 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
231 {"slaveof",slaveofCommand,3,"ast",0,NULL,0,0,0,0,0},
232 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
233 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
234 {"subscribe",subscribeCommand,-2,"rpslt",0,NULL,0,0,0,0,0},
235 {"unsubscribe",unsubscribeCommand,-1,"rpslt",0,NULL,0,0,0,0,0},
236 {"psubscribe",psubscribeCommand,-2,"rpslt",0,NULL,0,0,0,0,0},
237 {"punsubscribe",punsubscribeCommand,-1,"rpslt",0,NULL,0,0,0,0,0},
238 {"publish",publishCommand,3,"pflt",0,NULL,0,0,0,0,0},
239 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
240 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
241 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
242 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
243 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
244 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
245 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
246 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
247 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
248 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
249 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
250 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0},
251 {"bitop",bitopCommand,-4,"wm",0,NULL,2,-1,1,0,0},
252 {"bitcount",bitcountCommand,-2,"r",0,NULL,1,1,1,0,0}
253 };
254
255 /*============================ Utility functions ============================ */
256
257 /* Low level logging. To use only for very big messages, otherwise
258 * redisLog() is to prefer. */
259 void redisLogRaw(int level, const char *msg) {
260 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
261 const char *c = ".-*#";
262 FILE *fp;
263 char buf[64];
264 int rawmode = (level & REDIS_LOG_RAW);
265
266 level &= 0xff; /* clear flags */
267 if (level < server.verbosity) return;
268
269 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
270 if (!fp) return;
271
272 if (rawmode) {
273 fprintf(fp,"%s",msg);
274 } else {
275 int off;
276 struct timeval tv;
277
278 gettimeofday(&tv,NULL);
279 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
280 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
281 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
282 }
283 fflush(fp);
284
285 if (server.logfile) fclose(fp);
286
287 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
288 }
289
290 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
291 * is used across the code. The raw version is only used in order to dump
292 * the INFO output on crash. */
293 void redisLog(int level, const char *fmt, ...) {
294 va_list ap;
295 char msg[REDIS_MAX_LOGMSG_LEN];
296
297 if ((level&0xff) < server.verbosity) return;
298
299 va_start(ap, fmt);
300 vsnprintf(msg, sizeof(msg), fmt, ap);
301 va_end(ap);
302
303 redisLogRaw(level,msg);
304 }
305
306 /* Log a fixed message without printf-alike capabilities, in a way that is
307 * safe to call from a signal handler.
308 *
309 * We actually use this only for signals that are not fatal from the point
310 * of view of Redis. Signals that are going to kill the server anyway and
311 * where we need printf-alike features are served by redisLog(). */
312 void redisLogFromHandler(int level, const char *msg) {
313 int fd;
314 char buf[64];
315
316 if ((level&0xff) < server.verbosity ||
317 (server.logfile == NULL && server.daemonize)) return;
318 fd = server.logfile ?
319 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
320 STDOUT_FILENO;
321 if (fd == -1) return;
322 ll2string(buf,sizeof(buf),getpid());
323 if (write(fd,"[",1) == -1) goto err;
324 if (write(fd,buf,strlen(buf)) == -1) goto err;
325 if (write(fd," | signal handler] (",20) == -1) goto err;
326 ll2string(buf,sizeof(buf),time(NULL));
327 if (write(fd,buf,strlen(buf)) == -1) goto err;
328 if (write(fd,") ",2) == -1) goto err;
329 if (write(fd,msg,strlen(msg)) == -1) goto err;
330 if (write(fd,"\n",1) == -1) goto err;
331 err:
332 if (server.logfile) close(fd);
333 }
334
335 /* Return the UNIX time in microseconds */
336 long long ustime(void) {
337 struct timeval tv;
338 long long ust;
339
340 gettimeofday(&tv, NULL);
341 ust = ((long long)tv.tv_sec)*1000000;
342 ust += tv.tv_usec;
343 return ust;
344 }
345
346 /* Return the UNIX time in milliseconds */
347 long long mstime(void) {
348 return ustime()/1000;
349 }
350
351 /* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
352 * exit(), because the latter may interact with the same file objects used by
353 * the parent process. However if we are testing the coverage normal exit() is
354 * used in order to obtain the right coverage information. */
355 void exitFromChild(int retcode) {
356 #ifdef COVERAGE_TEST
357 exit(retcode);
358 #else
359 _exit(retcode);
360 #endif
361 }
362
363 /*====================== Hash table type implementation ==================== */
364
365 /* This is an hash table type that uses the SDS dynamic strings libary as
366 * keys and radis objects as values (objects can hold SDS strings,
367 * lists, sets). */
368
369 void dictVanillaFree(void *privdata, void *val)
370 {
371 DICT_NOTUSED(privdata);
372 zfree(val);
373 }
374
375 void dictListDestructor(void *privdata, void *val)
376 {
377 DICT_NOTUSED(privdata);
378 listRelease((list*)val);
379 }
380
381 int dictSdsKeyCompare(void *privdata, const void *key1,
382 const void *key2)
383 {
384 int l1,l2;
385 DICT_NOTUSED(privdata);
386
387 l1 = sdslen((sds)key1);
388 l2 = sdslen((sds)key2);
389 if (l1 != l2) return 0;
390 return memcmp(key1, key2, l1) == 0;
391 }
392
393 /* A case insensitive version used for the command lookup table. */
394 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
395 const void *key2)
396 {
397 DICT_NOTUSED(privdata);
398
399 return strcasecmp(key1, key2) == 0;
400 }
401
402 void dictRedisObjectDestructor(void *privdata, void *val)
403 {
404 DICT_NOTUSED(privdata);
405
406 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
407 decrRefCount(val);
408 }
409
410 void dictSdsDestructor(void *privdata, void *val)
411 {
412 DICT_NOTUSED(privdata);
413
414 sdsfree(val);
415 }
416
417 int dictObjKeyCompare(void *privdata, const void *key1,
418 const void *key2)
419 {
420 const robj *o1 = key1, *o2 = key2;
421 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
422 }
423
424 unsigned int dictObjHash(const void *key) {
425 const robj *o = key;
426 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
427 }
428
429 unsigned int dictSdsHash(const void *key) {
430 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
431 }
432
433 unsigned int dictSdsCaseHash(const void *key) {
434 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
435 }
436
437 int dictEncObjKeyCompare(void *privdata, const void *key1,
438 const void *key2)
439 {
440 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
441 int cmp;
442
443 if (o1->encoding == REDIS_ENCODING_INT &&
444 o2->encoding == REDIS_ENCODING_INT)
445 return o1->ptr == o2->ptr;
446
447 o1 = getDecodedObject(o1);
448 o2 = getDecodedObject(o2);
449 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
450 decrRefCount(o1);
451 decrRefCount(o2);
452 return cmp;
453 }
454
455 unsigned int dictEncObjHash(const void *key) {
456 robj *o = (robj*) key;
457
458 if (o->encoding == REDIS_ENCODING_RAW) {
459 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
460 } else {
461 if (o->encoding == REDIS_ENCODING_INT) {
462 char buf[32];
463 int len;
464
465 len = ll2string(buf,32,(long)o->ptr);
466 return dictGenHashFunction((unsigned char*)buf, len);
467 } else {
468 unsigned int hash;
469
470 o = getDecodedObject(o);
471 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
472 decrRefCount(o);
473 return hash;
474 }
475 }
476 }
477
478 /* Sets type hash table */
479 dictType setDictType = {
480 dictEncObjHash, /* hash function */
481 NULL, /* key dup */
482 NULL, /* val dup */
483 dictEncObjKeyCompare, /* key compare */
484 dictRedisObjectDestructor, /* key destructor */
485 NULL /* val destructor */
486 };
487
488 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
489 dictType zsetDictType = {
490 dictEncObjHash, /* hash function */
491 NULL, /* key dup */
492 NULL, /* val dup */
493 dictEncObjKeyCompare, /* key compare */
494 dictRedisObjectDestructor, /* key destructor */
495 NULL /* val destructor */
496 };
497
498 /* Db->dict, keys are sds strings, vals are Redis objects. */
499 dictType dbDictType = {
500 dictSdsHash, /* hash function */
501 NULL, /* key dup */
502 NULL, /* val dup */
503 dictSdsKeyCompare, /* key compare */
504 dictSdsDestructor, /* key destructor */
505 dictRedisObjectDestructor /* val destructor */
506 };
507
508 /* Db->expires */
509 dictType keyptrDictType = {
510 dictSdsHash, /* hash function */
511 NULL, /* key dup */
512 NULL, /* val dup */
513 dictSdsKeyCompare, /* key compare */
514 NULL, /* key destructor */
515 NULL /* val destructor */
516 };
517
518 /* Command table. sds string -> command struct pointer. */
519 dictType commandTableDictType = {
520 dictSdsCaseHash, /* hash function */
521 NULL, /* key dup */
522 NULL, /* val dup */
523 dictSdsKeyCaseCompare, /* key compare */
524 dictSdsDestructor, /* key destructor */
525 NULL /* val destructor */
526 };
527
528 /* Hash type hash table (note that small hashes are represented with zimpaps) */
529 dictType hashDictType = {
530 dictEncObjHash, /* hash function */
531 NULL, /* key dup */
532 NULL, /* val dup */
533 dictEncObjKeyCompare, /* key compare */
534 dictRedisObjectDestructor, /* key destructor */
535 dictRedisObjectDestructor /* val destructor */
536 };
537
538 /* Keylist hash table type has unencoded redis objects as keys and
539 * lists as values. It's used for blocking operations (BLPOP) and to
540 * map swapped keys to a list of clients waiting for this keys to be loaded. */
541 dictType keylistDictType = {
542 dictObjHash, /* hash function */
543 NULL, /* key dup */
544 NULL, /* val dup */
545 dictObjKeyCompare, /* key compare */
546 dictRedisObjectDestructor, /* key destructor */
547 dictListDestructor /* val destructor */
548 };
549
550 int htNeedsResize(dict *dict) {
551 long long size, used;
552
553 size = dictSlots(dict);
554 used = dictSize(dict);
555 return (size && used && size > DICT_HT_INITIAL_SIZE &&
556 (used*100/size < REDIS_HT_MINFILL));
557 }
558
559 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
560 * we resize the hash table to save memory */
561 void tryResizeHashTables(void) {
562 int j;
563
564 for (j = 0; j < server.dbnum; j++) {
565 if (htNeedsResize(server.db[j].dict))
566 dictResize(server.db[j].dict);
567 if (htNeedsResize(server.db[j].expires))
568 dictResize(server.db[j].expires);
569 }
570 }
571
572 /* Our hash table implementation performs rehashing incrementally while
573 * we write/read from the hash table. Still if the server is idle, the hash
574 * table will use two tables for a long time. So we try to use 1 millisecond
575 * of CPU time at every serverCron() loop in order to rehash some key. */
576 void incrementallyRehash(void) {
577 int j;
578
579 for (j = 0; j < server.dbnum; j++) {
580 /* Keys dictionary */
581 if (dictIsRehashing(server.db[j].dict)) {
582 dictRehashMilliseconds(server.db[j].dict,1);
583 break; /* already used our millisecond for this loop... */
584 }
585 /* Expires */
586 if (dictIsRehashing(server.db[j].expires)) {
587 dictRehashMilliseconds(server.db[j].expires,1);
588 break; /* already used our millisecond for this loop... */
589 }
590 }
591 }
592
593 /* This function is called once a background process of some kind terminates,
594 * as we want to avoid resizing the hash tables when there is a child in order
595 * to play well with copy-on-write (otherwise when a resize happens lots of
596 * memory pages are copied). The goal of this function is to update the ability
597 * for dict.c to resize the hash tables accordingly to the fact we have o not
598 * running childs. */
599 void updateDictResizePolicy(void) {
600 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
601 dictEnableResize();
602 else
603 dictDisableResize();
604 }
605
606 /* ======================= Cron: called every 100 ms ======================== */
607
608 /* Try to expire a few timed out keys. The algorithm used is adaptive and
609 * will use few CPU cycles if there are few expiring keys, otherwise
610 * it will get more aggressive to avoid that too much memory is used by
611 * keys that can be removed from the keyspace. */
612 void activeExpireCycle(void) {
613 int j, iteration = 0;
614 long long start = ustime(), timelimit;
615
616 /* We can use at max REDIS_EXPIRELOOKUPS_TIME_PERC percentage of CPU time
617 * per iteration. Since this function gets called with a frequency of
618 * REDIS_HZ times per second, the following is the max amount of
619 * microseconds we can spend in this function. */
620 timelimit = 1000000*REDIS_EXPIRELOOKUPS_TIME_PERC/REDIS_HZ/100;
621 if (timelimit <= 0) timelimit = 1;
622
623 for (j = 0; j < server.dbnum; j++) {
624 int expired;
625 redisDb *db = server.db+j;
626
627 /* Continue to expire if at the end of the cycle more than 25%
628 * of the keys were expired. */
629 do {
630 unsigned long num = dictSize(db->expires);
631 unsigned long slots = dictSlots(db->expires);
632 long long now = mstime();
633
634 /* When there are less than 1% filled slots getting random
635 * keys is expensive, so stop here waiting for better times...
636 * The dictionary will be resized asap. */
637 if (num && slots > DICT_HT_INITIAL_SIZE &&
638 (num*100/slots < 1)) break;
639
640 /* The main collection cycle. Sample random keys among keys
641 * with an expire set, checking for expired ones. */
642 expired = 0;
643 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
644 num = REDIS_EXPIRELOOKUPS_PER_CRON;
645 while (num--) {
646 dictEntry *de;
647 long long t;
648
649 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
650 t = dictGetSignedIntegerVal(de);
651 if (now > t) {
652 sds key = dictGetKey(de);
653 robj *keyobj = createStringObject(key,sdslen(key));
654
655 propagateExpire(db,keyobj);
656 dbDelete(db,keyobj);
657 decrRefCount(keyobj);
658 expired++;
659 server.stat_expiredkeys++;
660 }
661 }
662 /* We can't block forever here even if there are many keys to
663 * expire. So after a given amount of milliseconds return to the
664 * caller waiting for the other active expire cycle. */
665 iteration++;
666 if ((iteration & 0xf) == 0 && /* check once every 16 cycles. */
667 (ustime()-start) > timelimit) return;
668 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
669 }
670 }
671
672 void updateLRUClock(void) {
673 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
674 REDIS_LRU_CLOCK_MAX;
675 }
676
677
678 /* Add a sample to the operations per second array of samples. */
679 void trackOperationsPerSecond(void) {
680 long long t = mstime() - server.ops_sec_last_sample_time;
681 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
682 long long ops_sec;
683
684 ops_sec = t > 0 ? (ops*1000/t) : 0;
685
686 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
687 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
688 server.ops_sec_last_sample_time = mstime();
689 server.ops_sec_last_sample_ops = server.stat_numcommands;
690 }
691
692 /* Return the mean of all the samples. */
693 long long getOperationsPerSecond(void) {
694 int j;
695 long long sum = 0;
696
697 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
698 sum += server.ops_sec_samples[j];
699 return sum / REDIS_OPS_SEC_SAMPLES;
700 }
701
702 /* Check for timeouts. Returns non-zero if the client was terminated */
703 int clientsCronHandleTimeout(redisClient *c) {
704 time_t now = server.unixtime;
705
706 if (server.maxidletime &&
707 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
708 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
709 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
710 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
711 listLength(c->pubsub_patterns) == 0 &&
712 (now - c->lastinteraction > server.maxidletime))
713 {
714 redisLog(REDIS_VERBOSE,"Closing idle client");
715 freeClient(c);
716 return 1;
717 } else if (c->flags & REDIS_BLOCKED) {
718 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
719 addReply(c,shared.nullmultibulk);
720 unblockClientWaitingData(c);
721 }
722 }
723 return 0;
724 }
725
726 /* The client query buffer is an sds.c string that can end with a lot of
727 * free space not used, this function reclaims space if needed.
728 *
729 * The funciton always returns 0 as it never terminates the client. */
730 int clientsCronResizeQueryBuffer(redisClient *c) {
731 size_t querybuf_size = sdsAllocSize(c->querybuf);
732 time_t idletime = server.unixtime - c->lastinteraction;
733
734 /* There are two conditions to resize the query buffer:
735 * 1) Query buffer is > BIG_ARG and too big for latest peak.
736 * 2) Client is inactive and the buffer is bigger than 1k. */
737 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
738 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
739 (querybuf_size > 1024 && idletime > 2))
740 {
741 /* Only resize the query buffer if it is actually wasting space. */
742 if (sdsavail(c->querybuf) > 1024) {
743 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
744 }
745 }
746 /* Reset the peak again to capture the peak memory usage in the next
747 * cycle. */
748 c->querybuf_peak = 0;
749 return 0;
750 }
751
752 void clientsCron(void) {
753 /* Make sure to process at least 1/(REDIS_HZ*10) of clients per call.
754 * Since this function is called REDIS_HZ times per second we are sure that
755 * in the worst case we process all the clients in 10 seconds.
756 * In normal conditions (a reasonable number of clients) we process
757 * all the clients in a shorter time. */
758 int numclients = listLength(server.clients);
759 int iterations = numclients/(REDIS_HZ*10);
760
761 if (iterations < 50)
762 iterations = (numclients < 50) ? numclients : 50;
763 while(listLength(server.clients) && iterations--) {
764 redisClient *c;
765 listNode *head;
766
767 /* Rotate the list, take the current head, process.
768 * This way if the client must be removed from the list it's the
769 * first element and we don't incur into O(N) computation. */
770 listRotate(server.clients);
771 head = listFirst(server.clients);
772 c = listNodeValue(head);
773 /* The following functions do different service checks on the client.
774 * The protocol is that they return non-zero if the client was
775 * terminated. */
776 if (clientsCronHandleTimeout(c)) continue;
777 if (clientsCronResizeQueryBuffer(c)) continue;
778 }
779 }
780
781 /* This is our timer interrupt, called REDIS_HZ times per second.
782 * Here is where we do a number of things that need to be done asynchronously.
783 * For instance:
784 *
785 * - Active expired keys collection (it is also performed in a lazy way on
786 * lookup).
787 * - Software watchdong.
788 * - Update some statistic.
789 * - Incremental rehashing of the DBs hash tables.
790 * - Triggering BGSAVE / AOF rewrite, and handling of terminated children.
791 * - Clients timeout of differnet kinds.
792 * - Replication reconnection.
793 * - Many more...
794 *
795 * Everything directly called here will be called REDIS_HZ times per second,
796 * so in order to throttle execution of things we want to do less frequently
797 * a macro is used: run_with_period(milliseconds) { .... }
798 */
799
800 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
801 int j;
802 REDIS_NOTUSED(eventLoop);
803 REDIS_NOTUSED(id);
804 REDIS_NOTUSED(clientData);
805
806 /* Software watchdog: deliver the SIGALRM that will reach the signal
807 * handler if we don't return here fast enough. */
808 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
809
810 /* We take a cached value of the unix time in the global state because
811 * with virtual memory and aging there is to store the current time
812 * in objects at every object access, and accuracy is not needed.
813 * To access a global var is faster than calling time(NULL) */
814 server.unixtime = time(NULL);
815
816 run_with_period(100) trackOperationsPerSecond();
817
818 /* We have just 22 bits per object for LRU information.
819 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
820 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
821 *
822 * Note that even if this will wrap after 1.5 years it's not a problem,
823 * everything will still work but just some object will appear younger
824 * to Redis. But for this to happen a given object should never be touched
825 * for 1.5 years.
826 *
827 * Note that you can change the resolution altering the
828 * REDIS_LRU_CLOCK_RESOLUTION define.
829 */
830 updateLRUClock();
831
832 /* Record the max memory used since the server was started. */
833 if (zmalloc_used_memory() > server.stat_peak_memory)
834 server.stat_peak_memory = zmalloc_used_memory();
835
836 /* We received a SIGTERM, shutting down here in a safe way, as it is
837 * not ok doing so inside the signal handler. */
838 if (server.shutdown_asap) {
839 if (prepareForShutdown(0) == REDIS_OK) exit(0);
840 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
841 }
842
843 /* Show some info about non-empty databases */
844 run_with_period(5000) {
845 for (j = 0; j < server.dbnum; j++) {
846 long long size, used, vkeys;
847
848 size = dictSlots(server.db[j].dict);
849 used = dictSize(server.db[j].dict);
850 vkeys = dictSize(server.db[j].expires);
851 if (used || vkeys) {
852 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
853 /* dictPrintStats(server.dict); */
854 }
855 }
856 }
857
858 /* We don't want to resize the hash tables while a bacground saving
859 * is in progress: the saving child is created using fork() that is
860 * implemented with a copy-on-write semantic in most modern systems, so
861 * if we resize the HT while there is the saving child at work actually
862 * a lot of memory movements in the parent will cause a lot of pages
863 * copied. */
864 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
865 tryResizeHashTables();
866 if (server.activerehashing) incrementallyRehash();
867 }
868
869 /* Show information about connected clients */
870 if (!server.sentinel_mode) {
871 run_with_period(5000) {
872 redisLog(REDIS_VERBOSE,
873 "%d clients connected (%d slaves), %zu bytes in use",
874 listLength(server.clients)-listLength(server.slaves),
875 listLength(server.slaves),
876 zmalloc_used_memory());
877 }
878 }
879
880 /* We need to do a few operations on clients asynchronously. */
881 clientsCron();
882
883 /* Start a scheduled AOF rewrite if this was requested by the user while
884 * a BGSAVE was in progress. */
885 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
886 server.aof_rewrite_scheduled)
887 {
888 rewriteAppendOnlyFileBackground();
889 }
890
891 /* Check if a background saving or AOF rewrite in progress terminated. */
892 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
893 int statloc;
894 pid_t pid;
895
896 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
897 int exitcode = WEXITSTATUS(statloc);
898 int bysignal = 0;
899
900 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
901
902 if (pid == server.rdb_child_pid) {
903 backgroundSaveDoneHandler(exitcode,bysignal);
904 } else {
905 backgroundRewriteDoneHandler(exitcode,bysignal);
906 }
907 updateDictResizePolicy();
908 }
909 } else {
910 /* If there is not a background saving/rewrite in progress check if
911 * we have to save/rewrite now */
912 for (j = 0; j < server.saveparamslen; j++) {
913 struct saveparam *sp = server.saveparams+j;
914
915 if (server.dirty >= sp->changes &&
916 server.unixtime-server.lastsave > sp->seconds) {
917 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
918 sp->changes, sp->seconds);
919 rdbSaveBackground(server.rdb_filename);
920 break;
921 }
922 }
923
924 /* Trigger an AOF rewrite if needed */
925 if (server.rdb_child_pid == -1 &&
926 server.aof_child_pid == -1 &&
927 server.aof_rewrite_perc &&
928 server.aof_current_size > server.aof_rewrite_min_size)
929 {
930 long long base = server.aof_rewrite_base_size ?
931 server.aof_rewrite_base_size : 1;
932 long long growth = (server.aof_current_size*100/base) - 100;
933 if (growth >= server.aof_rewrite_perc) {
934 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
935 rewriteAppendOnlyFileBackground();
936 }
937 }
938 }
939
940
941 /* If we postponed an AOF buffer flush, let's try to do it every time the
942 * cron function is called. */
943 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
944
945 /* Expire a few keys per cycle, only if this is a master.
946 * On slaves we wait for DEL operations synthesized by the master
947 * in order to guarantee a strict consistency. */
948 if (server.masterhost == NULL) activeExpireCycle();
949
950 /* Close clients that need to be closed asynchronous */
951 freeClientsInAsyncFreeQueue();
952
953 /* Replication cron function -- used to reconnect to master and
954 * to detect transfer failures. */
955 run_with_period(1000) replicationCron();
956
957 /* Run the sentinel timer if we are in sentinel mode. */
958 run_with_period(100) {
959 if (server.sentinel_mode) sentinelTimer();
960 }
961
962 server.cronloops++;
963 return 1000/REDIS_HZ;
964 }
965
966 /* This function gets called every time Redis is entering the
967 * main loop of the event driven library, that is, before to sleep
968 * for ready file descriptors. */
969 void beforeSleep(struct aeEventLoop *eventLoop) {
970 REDIS_NOTUSED(eventLoop);
971 listNode *ln;
972 redisClient *c;
973
974 /* Try to process pending commands for clients that were just unblocked. */
975 while (listLength(server.unblocked_clients)) {
976 ln = listFirst(server.unblocked_clients);
977 redisAssert(ln != NULL);
978 c = ln->value;
979 listDelNode(server.unblocked_clients,ln);
980 c->flags &= ~REDIS_UNBLOCKED;
981
982 /* Process remaining data in the input buffer. */
983 if (c->querybuf && sdslen(c->querybuf) > 0) {
984 server.current_client = c;
985 processInputBuffer(c);
986 server.current_client = NULL;
987 }
988 }
989
990 /* Write the AOF buffer on disk */
991 flushAppendOnlyFile(0);
992 }
993
994 /* =========================== Server initialization ======================== */
995
996 void createSharedObjects(void) {
997 int j;
998
999 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
1000 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
1001 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
1002 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
1003 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
1004 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
1005 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
1006 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
1007 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
1008 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
1009 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
1010 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
1011 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
1012 "-ERR Operation against a key holding the wrong kind of value\r\n"));
1013 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
1014 "-ERR no such key\r\n"));
1015 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
1016 "-ERR syntax error\r\n"));
1017 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
1018 "-ERR source and destination objects are the same\r\n"));
1019 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
1020 "-ERR index out of range\r\n"));
1021 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
1022 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
1023 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
1024 "-LOADING Redis is loading the dataset in memory\r\n"));
1025 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
1026 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
1027 shared.masterdownerr = createObject(REDIS_STRING,sdsnew(
1028 "-MASTERDOWN Link with MASTER is down and slave-serve-stale-data is set to 'no'.\r\n"));
1029 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
1030 "-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"));
1031 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
1032 "-READONLY You can't write against a read only slave.\r\n"));
1033 shared.oomerr = createObject(REDIS_STRING,sdsnew(
1034 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
1035 shared.space = createObject(REDIS_STRING,sdsnew(" "));
1036 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1037 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1038
1039 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1040 shared.select[j] = createObject(REDIS_STRING,
1041 sdscatprintf(sdsempty(),"select %d\r\n", j));
1042 }
1043 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1044 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1045 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1046 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1047 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1048 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1049 shared.del = createStringObject("DEL",3);
1050 shared.rpop = createStringObject("RPOP",4);
1051 shared.lpop = createStringObject("LPOP",4);
1052 shared.lpush = createStringObject("LPUSH",5);
1053 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1054 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1055 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1056 }
1057 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1058 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1059 sdscatprintf(sdsempty(),"*%d\r\n",j));
1060 shared.bulkhdr[j] = createObject(REDIS_STRING,
1061 sdscatprintf(sdsempty(),"$%d\r\n",j));
1062 }
1063 }
1064
1065 void initServerConfig() {
1066 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1067 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1068 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1069 server.port = REDIS_SERVERPORT;
1070 server.bindaddr = NULL;
1071 server.unixsocket = NULL;
1072 server.unixsocketperm = 0;
1073 server.ipfd = -1;
1074 server.sofd = -1;
1075 server.dbnum = REDIS_DEFAULT_DBNUM;
1076 server.verbosity = REDIS_NOTICE;
1077 server.maxidletime = REDIS_MAXIDLETIME;
1078 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1079 server.saveparams = NULL;
1080 server.loading = 0;
1081 server.logfile = NULL; /* NULL = log on standard output */
1082 server.syslog_enabled = 0;
1083 server.syslog_ident = zstrdup("redis");
1084 server.syslog_facility = LOG_LOCAL0;
1085 server.daemonize = 0;
1086 server.aof_state = REDIS_AOF_OFF;
1087 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1088 server.aof_no_fsync_on_rewrite = 0;
1089 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1090 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1091 server.aof_rewrite_base_size = 0;
1092 server.aof_rewrite_scheduled = 0;
1093 server.aof_last_fsync = time(NULL);
1094 server.aof_rewrite_time_last = -1;
1095 server.aof_rewrite_time_start = -1;
1096 server.aof_lastbgrewrite_status = REDIS_OK;
1097 server.aof_delayed_fsync = 0;
1098 server.aof_fd = -1;
1099 server.aof_selected_db = -1; /* Make sure the first time will not match */
1100 server.aof_flush_postponed_start = 0;
1101 server.pidfile = zstrdup("/var/run/redis.pid");
1102 server.rdb_filename = zstrdup("dump.rdb");
1103 server.aof_filename = zstrdup("appendonly.aof");
1104 server.requirepass = NULL;
1105 server.rdb_compression = 1;
1106 server.rdb_checksum = 1;
1107 server.activerehashing = 1;
1108 server.maxclients = REDIS_MAX_CLIENTS;
1109 server.bpop_blocked_clients = 0;
1110 server.maxmemory = 0;
1111 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1112 server.maxmemory_samples = 3;
1113 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1114 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1115 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1116 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1117 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1118 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1119 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1120 server.shutdown_asap = 0;
1121 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1122 server.repl_timeout = REDIS_REPL_TIMEOUT;
1123 server.lua_caller = NULL;
1124 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1125 server.lua_client = NULL;
1126 server.lua_timedout = 0;
1127
1128 updateLRUClock();
1129 resetServerSaveParams();
1130
1131 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1132 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1133 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1134 /* Replication related */
1135 server.masterauth = NULL;
1136 server.masterhost = NULL;
1137 server.masterport = 6379;
1138 server.master = NULL;
1139 server.repl_state = REDIS_REPL_NONE;
1140 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1141 server.repl_serve_stale_data = 1;
1142 server.repl_slave_ro = 1;
1143 server.repl_down_since = time(NULL);
1144 server.slave_priority = REDIS_DEFAULT_SLAVE_PRIORITY;
1145
1146 /* Client output buffer limits */
1147 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1148 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1149 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1150 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1151 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1152 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1153 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1154 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1155 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1156
1157 /* Double constants initialization */
1158 R_Zero = 0.0;
1159 R_PosInf = 1.0/R_Zero;
1160 R_NegInf = -1.0/R_Zero;
1161 R_Nan = R_Zero/R_Zero;
1162
1163 /* Command table -- we intiialize it here as it is part of the
1164 * initial configuration, since command names may be changed via
1165 * redis.conf using the rename-command directive. */
1166 server.commands = dictCreate(&commandTableDictType,NULL);
1167 populateCommandTable();
1168 server.delCommand = lookupCommandByCString("del");
1169 server.multiCommand = lookupCommandByCString("multi");
1170 server.lpushCommand = lookupCommandByCString("lpush");
1171 server.lpopCommand = lookupCommandByCString("lpop");
1172 server.rpopCommand = lookupCommandByCString("rpop");
1173
1174 /* Slow log */
1175 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1176 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1177
1178 /* Debugging */
1179 server.assert_failed = "<no assertion failed>";
1180 server.assert_file = "<no file>";
1181 server.assert_line = 0;
1182 server.bug_report_start = 0;
1183 server.watchdog_period = 0;
1184 }
1185
1186 /* This function will try to raise the max number of open files accordingly to
1187 * the configured max number of clients. It will also account for 32 additional
1188 * file descriptors as we need a few more for persistence, listening
1189 * sockets, log files and so forth.
1190 *
1191 * If it will not be possible to set the limit accordingly to the configured
1192 * max number of clients, the function will do the reverse setting
1193 * server.maxclients to the value that we can actually handle. */
1194 void adjustOpenFilesLimit(void) {
1195 rlim_t maxfiles = server.maxclients+32;
1196 struct rlimit limit;
1197
1198 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1199 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1200 strerror(errno));
1201 server.maxclients = 1024-32;
1202 } else {
1203 rlim_t oldlimit = limit.rlim_cur;
1204
1205 /* Set the max number of files if the current limit is not enough
1206 * for our needs. */
1207 if (oldlimit < maxfiles) {
1208 rlim_t f;
1209
1210 f = maxfiles;
1211 while(f > oldlimit) {
1212 limit.rlim_cur = f;
1213 limit.rlim_max = f;
1214 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1215 f -= 128;
1216 }
1217 if (f < oldlimit) f = oldlimit;
1218 if (f != maxfiles) {
1219 server.maxclients = f-32;
1220 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1221 (int) maxfiles, strerror(errno), (int) server.maxclients);
1222 } else {
1223 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1224 (int) maxfiles);
1225 }
1226 }
1227 }
1228 }
1229
1230 void initServer() {
1231 int j;
1232
1233 signal(SIGHUP, SIG_IGN);
1234 signal(SIGPIPE, SIG_IGN);
1235 setupSignalHandlers();
1236
1237 if (server.syslog_enabled) {
1238 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1239 server.syslog_facility);
1240 }
1241
1242 server.current_client = NULL;
1243 server.clients = listCreate();
1244 server.clients_to_close = listCreate();
1245 server.slaves = listCreate();
1246 server.monitors = listCreate();
1247 server.unblocked_clients = listCreate();
1248 server.ready_keys = listCreate();
1249
1250 createSharedObjects();
1251 adjustOpenFilesLimit();
1252 server.el = aeCreateEventLoop(server.maxclients+1024);
1253 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1254
1255 if (server.port != 0) {
1256 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1257 if (server.ipfd == ANET_ERR) {
1258 redisLog(REDIS_WARNING, "Opening port %d: %s",
1259 server.port, server.neterr);
1260 exit(1);
1261 }
1262 }
1263 if (server.unixsocket != NULL) {
1264 unlink(server.unixsocket); /* don't care if this fails */
1265 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1266 if (server.sofd == ANET_ERR) {
1267 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1268 exit(1);
1269 }
1270 }
1271 if (server.ipfd < 0 && server.sofd < 0) {
1272 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1273 exit(1);
1274 }
1275 for (j = 0; j < server.dbnum; j++) {
1276 server.db[j].dict = dictCreate(&dbDictType,NULL);
1277 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1278 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1279 server.db[j].ready_keys = dictCreate(&setDictType,NULL);
1280 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1281 server.db[j].id = j;
1282 }
1283 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1284 server.pubsub_patterns = listCreate();
1285 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1286 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1287 server.cronloops = 0;
1288 server.rdb_child_pid = -1;
1289 server.aof_child_pid = -1;
1290 aofRewriteBufferReset();
1291 server.aof_buf = sdsempty();
1292 server.lastsave = time(NULL);
1293 server.rdb_save_time_last = -1;
1294 server.rdb_save_time_start = -1;
1295 server.dirty = 0;
1296 server.stat_numcommands = 0;
1297 server.stat_numconnections = 0;
1298 server.stat_expiredkeys = 0;
1299 server.stat_evictedkeys = 0;
1300 server.stat_starttime = time(NULL);
1301 server.stat_keyspace_misses = 0;
1302 server.stat_keyspace_hits = 0;
1303 server.stat_peak_memory = 0;
1304 server.stat_fork_time = 0;
1305 server.stat_rejected_conn = 0;
1306 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1307 server.ops_sec_idx = 0;
1308 server.ops_sec_last_sample_time = mstime();
1309 server.ops_sec_last_sample_ops = 0;
1310 server.unixtime = time(NULL);
1311 server.lastbgsave_status = REDIS_OK;
1312 server.stop_writes_on_bgsave_err = 1;
1313 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1314 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1315 acceptTcpHandler,NULL) == AE_ERR) redisPanic("Unrecoverable error creating server.ipfd file event.");
1316 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1317 acceptUnixHandler,NULL) == AE_ERR) redisPanic("Unrecoverable error creating server.sofd file event.");
1318
1319 if (server.aof_state == REDIS_AOF_ON) {
1320 server.aof_fd = open(server.aof_filename,
1321 O_WRONLY|O_APPEND|O_CREAT,0644);
1322 if (server.aof_fd == -1) {
1323 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1324 strerror(errno));
1325 exit(1);
1326 }
1327 }
1328
1329 /* 32 bit instances are limited to 4GB of address space, so if there is
1330 * no explicit limit in the user provided configuration we set a limit
1331 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1332 * useless crashes of the Redis instance. */
1333 if (server.arch_bits == 32 && server.maxmemory == 0) {
1334 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1335 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1336 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1337 }
1338
1339 scriptingInit();
1340 slowlogInit();
1341 bioInit();
1342 }
1343
1344 /* Populates the Redis Command Table starting from the hard coded list
1345 * we have on top of redis.c file. */
1346 void populateCommandTable(void) {
1347 int j;
1348 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1349
1350 for (j = 0; j < numcommands; j++) {
1351 struct redisCommand *c = redisCommandTable+j;
1352 char *f = c->sflags;
1353 int retval;
1354
1355 while(*f != '\0') {
1356 switch(*f) {
1357 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1358 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1359 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1360 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1361 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1362 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1363 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1364 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1365 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1366 case 'l': c->flags |= REDIS_CMD_LOADING; break;
1367 case 't': c->flags |= REDIS_CMD_STALE; break;
1368 default: redisPanic("Unsupported command flag"); break;
1369 }
1370 f++;
1371 }
1372
1373 retval = dictAdd(server.commands, sdsnew(c->name), c);
1374 assert(retval == DICT_OK);
1375 }
1376 }
1377
1378 void resetCommandTableStats(void) {
1379 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1380 int j;
1381
1382 for (j = 0; j < numcommands; j++) {
1383 struct redisCommand *c = redisCommandTable+j;
1384
1385 c->microseconds = 0;
1386 c->calls = 0;
1387 }
1388 }
1389
1390 /* ========================== Redis OP Array API ============================ */
1391
1392 void redisOpArrayInit(redisOpArray *oa) {
1393 oa->ops = NULL;
1394 oa->numops = 0;
1395 }
1396
1397 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1398 robj **argv, int argc, int target)
1399 {
1400 redisOp *op;
1401
1402 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1403 op = oa->ops+oa->numops;
1404 op->cmd = cmd;
1405 op->dbid = dbid;
1406 op->argv = argv;
1407 op->argc = argc;
1408 op->target = target;
1409 oa->numops++;
1410 return oa->numops;
1411 }
1412
1413 void redisOpArrayFree(redisOpArray *oa) {
1414 while(oa->numops) {
1415 int j;
1416 redisOp *op;
1417
1418 oa->numops--;
1419 op = oa->ops+oa->numops;
1420 for (j = 0; j < op->argc; j++)
1421 decrRefCount(op->argv[j]);
1422 zfree(op->argv);
1423 }
1424 zfree(oa->ops);
1425 }
1426
1427 /* ====================== Commands lookup and execution ===================== */
1428
1429 struct redisCommand *lookupCommand(sds name) {
1430 return dictFetchValue(server.commands, name);
1431 }
1432
1433 struct redisCommand *lookupCommandByCString(char *s) {
1434 struct redisCommand *cmd;
1435 sds name = sdsnew(s);
1436
1437 cmd = dictFetchValue(server.commands, name);
1438 sdsfree(name);
1439 return cmd;
1440 }
1441
1442 /* Propagate the specified command (in the context of the specified database id)
1443 * to AOF, Slaves and Monitors.
1444 *
1445 * flags are an xor between:
1446 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1447 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1448 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1449 */
1450 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1451 int flags)
1452 {
1453 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1454 feedAppendOnlyFile(cmd,dbid,argv,argc);
1455 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1456 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1457 }
1458
1459 /* Used inside commands to schedule the propagation of additional commands
1460 * after the current command is propagated to AOF / Replication. */
1461 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1462 int target)
1463 {
1464 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1465 }
1466
1467 /* Call() is the core of Redis execution of a command */
1468 void call(redisClient *c, int flags) {
1469 long long dirty, start = ustime(), duration;
1470
1471 /* Sent the command to clients in MONITOR mode, only if the commands are
1472 * not geneated from reading an AOF. */
1473 if (listLength(server.monitors) && !server.loading)
1474 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1475
1476 /* Call the command. */
1477 redisOpArrayInit(&server.also_propagate);
1478 dirty = server.dirty;
1479 c->cmd->proc(c);
1480 dirty = server.dirty-dirty;
1481 duration = ustime()-start;
1482
1483 /* When EVAL is called loading the AOF we don't want commands called
1484 * from Lua to go into the slowlog or to populate statistics. */
1485 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1486 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1487
1488 /* Log the command into the Slow log if needed, and populate the
1489 * per-command statistics that we show in INFO commandstats. */
1490 if (flags & REDIS_CALL_SLOWLOG)
1491 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1492 if (flags & REDIS_CALL_STATS) {
1493 c->cmd->microseconds += duration;
1494 c->cmd->calls++;
1495 }
1496
1497 /* Propagate the command into the AOF and replication link */
1498 if (flags & REDIS_CALL_PROPAGATE) {
1499 int flags = REDIS_PROPAGATE_NONE;
1500
1501 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1502 flags |= REDIS_PROPAGATE_REPL;
1503 if (dirty)
1504 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1505 if (flags != REDIS_PROPAGATE_NONE)
1506 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1507 }
1508 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1509 * PUSH command. */
1510 if (server.also_propagate.numops) {
1511 int j;
1512 redisOp *rop;
1513
1514 for (j = 0; j < server.also_propagate.numops; j++) {
1515 rop = &server.also_propagate.ops[j];
1516 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1517 }
1518 redisOpArrayFree(&server.also_propagate);
1519 }
1520 server.stat_numcommands++;
1521 }
1522
1523 /* If this function gets called we already read a whole
1524 * command, argments are in the client argv/argc fields.
1525 * processCommand() execute the command or prepare the
1526 * server for a bulk read from the client.
1527 *
1528 * If 1 is returned the client is still alive and valid and
1529 * and other operations can be performed by the caller. Otherwise
1530 * if 0 is returned the client was destroied (i.e. after QUIT). */
1531 int processCommand(redisClient *c) {
1532 /* The QUIT command is handled separately. Normal command procs will
1533 * go through checking for replication and QUIT will cause trouble
1534 * when FORCE_REPLICATION is enabled and would be implemented in
1535 * a regular command proc. */
1536 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1537 addReply(c,shared.ok);
1538 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1539 return REDIS_ERR;
1540 }
1541
1542 /* Now lookup the command and check ASAP about trivial error conditions
1543 * such as wrong arity, bad command name and so forth. */
1544 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1545 if (!c->cmd) {
1546 addReplyErrorFormat(c,"unknown command '%s'",
1547 (char*)c->argv[0]->ptr);
1548 return REDIS_OK;
1549 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1550 (c->argc < -c->cmd->arity)) {
1551 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1552 c->cmd->name);
1553 return REDIS_OK;
1554 }
1555
1556 /* Check if the user is authenticated */
1557 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1558 {
1559 addReplyError(c,"operation not permitted");
1560 return REDIS_OK;
1561 }
1562
1563 /* Handle the maxmemory directive.
1564 *
1565 * First we try to free some memory if possible (if there are volatile
1566 * keys in the dataset). If there are not the only thing we can do
1567 * is returning an error. */
1568 if (server.maxmemory) {
1569 int retval = freeMemoryIfNeeded();
1570 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1571 addReply(c, shared.oomerr);
1572 return REDIS_OK;
1573 }
1574 }
1575
1576 /* Don't accept write commands if there are problems persisting on disk. */
1577 if (server.stop_writes_on_bgsave_err &&
1578 server.saveparamslen > 0
1579 && server.lastbgsave_status == REDIS_ERR &&
1580 c->cmd->flags & REDIS_CMD_WRITE)
1581 {
1582 addReply(c, shared.bgsaveerr);
1583 return REDIS_OK;
1584 }
1585
1586 /* Don't accept write commands if this is a read only slave. But
1587 * accept write commands if this is our master. */
1588 if (server.masterhost && server.repl_slave_ro &&
1589 !(c->flags & REDIS_MASTER) &&
1590 c->cmd->flags & REDIS_CMD_WRITE)
1591 {
1592 addReply(c, shared.roslaveerr);
1593 return REDIS_OK;
1594 }
1595
1596 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1597 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1598 &&
1599 c->cmd->proc != subscribeCommand &&
1600 c->cmd->proc != unsubscribeCommand &&
1601 c->cmd->proc != psubscribeCommand &&
1602 c->cmd->proc != punsubscribeCommand) {
1603 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1604 return REDIS_OK;
1605 }
1606
1607 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1608 * we are a slave with a broken link with master. */
1609 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1610 server.repl_serve_stale_data == 0 &&
1611 !(c->cmd->flags & REDIS_CMD_STALE))
1612 {
1613 addReply(c, shared.masterdownerr);
1614 return REDIS_OK;
1615 }
1616
1617 /* Loading DB? Return an error if the command has not the
1618 * REDIS_CMD_LOADING flag. */
1619 if (server.loading && !(c->cmd->flags & REDIS_CMD_LOADING)) {
1620 addReply(c, shared.loadingerr);
1621 return REDIS_OK;
1622 }
1623
1624 /* Lua script too slow? Only allow commands with REDIS_CMD_STALE flag. */
1625 if (server.lua_timedout &&
1626 !(c->cmd->proc == shutdownCommand &&
1627 c->argc == 2 &&
1628 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1629 !(c->cmd->proc == scriptCommand &&
1630 c->argc == 2 &&
1631 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1632 {
1633 addReply(c, shared.slowscripterr);
1634 return REDIS_OK;
1635 }
1636
1637 /* Exec the command */
1638 if (c->flags & REDIS_MULTI &&
1639 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1640 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1641 {
1642 queueMultiCommand(c);
1643 addReply(c,shared.queued);
1644 } else {
1645 call(c,REDIS_CALL_FULL);
1646 if (listLength(server.ready_keys))
1647 handleClientsBlockedOnLists();
1648 }
1649 return REDIS_OK;
1650 }
1651
1652 /*================================== Shutdown =============================== */
1653
1654 int prepareForShutdown(int flags) {
1655 int save = flags & REDIS_SHUTDOWN_SAVE;
1656 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1657
1658 redisLog(REDIS_WARNING,"User requested shutdown...");
1659 /* Kill the saving child if there is a background saving in progress.
1660 We want to avoid race conditions, for instance our saving child may
1661 overwrite the synchronous saving did by SHUTDOWN. */
1662 if (server.rdb_child_pid != -1) {
1663 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1664 kill(server.rdb_child_pid,SIGKILL);
1665 rdbRemoveTempFile(server.rdb_child_pid);
1666 }
1667 if (server.aof_state != REDIS_AOF_OFF) {
1668 /* Kill the AOF saving child as the AOF we already have may be longer
1669 * but contains the full dataset anyway. */
1670 if (server.aof_child_pid != -1) {
1671 redisLog(REDIS_WARNING,
1672 "There is a child rewriting the AOF. Killing it!");
1673 kill(server.aof_child_pid,SIGKILL);
1674 }
1675 /* Append only file: fsync() the AOF and exit */
1676 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1677 aof_fsync(server.aof_fd);
1678 }
1679 if ((server.saveparamslen > 0 && !nosave) || save) {
1680 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1681 /* Snapshotting. Perform a SYNC SAVE and exit */
1682 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1683 /* Ooops.. error saving! The best we can do is to continue
1684 * operating. Note that if there was a background saving process,
1685 * in the next cron() Redis will be notified that the background
1686 * saving aborted, handling special stuff like slaves pending for
1687 * synchronization... */
1688 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1689 return REDIS_ERR;
1690 }
1691 }
1692 if (server.daemonize) {
1693 redisLog(REDIS_NOTICE,"Removing the pid file.");
1694 unlink(server.pidfile);
1695 }
1696 /* Close the listening sockets. Apparently this allows faster restarts. */
1697 if (server.ipfd != -1) close(server.ipfd);
1698 if (server.sofd != -1) close(server.sofd);
1699 if (server.unixsocket) {
1700 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1701 unlink(server.unixsocket); /* don't care if this fails */
1702 }
1703
1704 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1705 return REDIS_OK;
1706 }
1707
1708 /*================================== Commands =============================== */
1709
1710 /* Return zero if strings are the same, non-zero if they are not.
1711 * The comparison is performed in a way that prevents an attacker to obtain
1712 * information about the nature of the strings just monitoring the execution
1713 * time of the function.
1714 *
1715 * Note that limiting the comparison length to strings up to 512 bytes we
1716 * can avoid leaking any information about the password length and any
1717 * possible branch misprediction related leak.
1718 */
1719 int time_independent_strcmp(char *a, char *b) {
1720 char bufa[REDIS_AUTHPASS_MAX_LEN], bufb[REDIS_AUTHPASS_MAX_LEN];
1721 /* The above two strlen perform len(a) + len(b) operations where either
1722 * a or b are fixed (our password) length, and the difference is only
1723 * relative to the length of the user provided string, so no information
1724 * leak is possible in the following two lines of code. */
1725 int alen = strlen(a);
1726 int blen = strlen(b);
1727 int j;
1728 int diff = 0;
1729
1730 /* We can't compare strings longer than our static buffers.
1731 * Note that this will never pass the first test in practical circumstances
1732 * so there is no info leak. */
1733 if (alen > sizeof(bufa) || blen > sizeof(bufb)) return 1;
1734
1735 memset(bufa,0,sizeof(bufa)); /* Constant time. */
1736 memset(bufb,0,sizeof(bufb)); /* Constant time. */
1737 /* Again the time of the following two copies is proportional to
1738 * len(a) + len(b) so no info is leaked. */
1739 memcpy(bufa,a,alen);
1740 memcpy(bufb,b,blen);
1741
1742 /* Always compare all the chars in the two buffers without
1743 * conditional expressions. */
1744 for (j = 0; j < sizeof(bufa); j++) {
1745 diff |= (bufa[j] ^ bufb[j]);
1746 }
1747 /* Length must be equal as well. */
1748 diff |= alen ^ blen;
1749 return diff; /* If zero strings are the same. */
1750 }
1751
1752 void authCommand(redisClient *c) {
1753 if (!server.requirepass) {
1754 addReplyError(c,"Client sent AUTH, but no password is set");
1755 } else if (!time_independent_strcmp(c->argv[1]->ptr, server.requirepass)) {
1756 c->authenticated = 1;
1757 addReply(c,shared.ok);
1758 } else {
1759 c->authenticated = 0;
1760 addReplyError(c,"invalid password");
1761 }
1762 }
1763
1764 void pingCommand(redisClient *c) {
1765 addReply(c,shared.pong);
1766 }
1767
1768 void echoCommand(redisClient *c) {
1769 addReplyBulk(c,c->argv[1]);
1770 }
1771
1772 void timeCommand(redisClient *c) {
1773 struct timeval tv;
1774
1775 /* gettimeofday() can only fail if &tv is a bad addresss so we
1776 * don't check for errors. */
1777 gettimeofday(&tv,NULL);
1778 addReplyMultiBulkLen(c,2);
1779 addReplyBulkLongLong(c,tv.tv_sec);
1780 addReplyBulkLongLong(c,tv.tv_usec);
1781 }
1782
1783 /* Convert an amount of bytes into a human readable string in the form
1784 * of 100B, 2G, 100M, 4K, and so forth. */
1785 void bytesToHuman(char *s, unsigned long long n) {
1786 double d;
1787
1788 if (n < 1024) {
1789 /* Bytes */
1790 sprintf(s,"%lluB",n);
1791 return;
1792 } else if (n < (1024*1024)) {
1793 d = (double)n/(1024);
1794 sprintf(s,"%.2fK",d);
1795 } else if (n < (1024LL*1024*1024)) {
1796 d = (double)n/(1024*1024);
1797 sprintf(s,"%.2fM",d);
1798 } else if (n < (1024LL*1024*1024*1024)) {
1799 d = (double)n/(1024LL*1024*1024);
1800 sprintf(s,"%.2fG",d);
1801 }
1802 }
1803
1804 /* Create the string returned by the INFO command. This is decoupled
1805 * by the INFO command itself as we need to report the same information
1806 * on memory corruption problems. */
1807 sds genRedisInfoString(char *section) {
1808 sds info = sdsempty();
1809 time_t uptime = server.unixtime-server.stat_starttime;
1810 int j, numcommands;
1811 struct rusage self_ru, c_ru;
1812 unsigned long lol, bib;
1813 int allsections = 0, defsections = 0;
1814 int sections = 0;
1815
1816 if (section) {
1817 allsections = strcasecmp(section,"all") == 0;
1818 defsections = strcasecmp(section,"default") == 0;
1819 }
1820
1821 getrusage(RUSAGE_SELF, &self_ru);
1822 getrusage(RUSAGE_CHILDREN, &c_ru);
1823 getClientsMaxBuffers(&lol,&bib);
1824
1825 /* Server */
1826 if (allsections || defsections || !strcasecmp(section,"server")) {
1827 struct utsname name;
1828
1829 if (sections++) info = sdscat(info,"\r\n");
1830 uname(&name);
1831 info = sdscatprintf(info,
1832 "# Server\r\n"
1833 "redis_version:%s\r\n"
1834 "redis_git_sha1:%s\r\n"
1835 "redis_git_dirty:%d\r\n"
1836 "os:%s %s %s\r\n"
1837 "arch_bits:%d\r\n"
1838 "multiplexing_api:%s\r\n"
1839 "gcc_version:%d.%d.%d\r\n"
1840 "process_id:%ld\r\n"
1841 "run_id:%s\r\n"
1842 "tcp_port:%d\r\n"
1843 "uptime_in_seconds:%ld\r\n"
1844 "uptime_in_days:%ld\r\n"
1845 "lru_clock:%ld\r\n",
1846 REDIS_VERSION,
1847 redisGitSHA1(),
1848 strtol(redisGitDirty(),NULL,10) > 0,
1849 name.sysname, name.release, name.machine,
1850 server.arch_bits,
1851 aeGetApiName(),
1852 #ifdef __GNUC__
1853 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1854 #else
1855 0,0,0,
1856 #endif
1857 (long) getpid(),
1858 server.runid,
1859 server.port,
1860 uptime,
1861 uptime/(3600*24),
1862 (unsigned long) server.lruclock);
1863 }
1864
1865 /* Clients */
1866 if (allsections || defsections || !strcasecmp(section,"clients")) {
1867 if (sections++) info = sdscat(info,"\r\n");
1868 info = sdscatprintf(info,
1869 "# Clients\r\n"
1870 "connected_clients:%lu\r\n"
1871 "client_longest_output_list:%lu\r\n"
1872 "client_biggest_input_buf:%lu\r\n"
1873 "blocked_clients:%d\r\n",
1874 listLength(server.clients)-listLength(server.slaves),
1875 lol, bib,
1876 server.bpop_blocked_clients);
1877 }
1878
1879 /* Memory */
1880 if (allsections || defsections || !strcasecmp(section,"memory")) {
1881 char hmem[64];
1882 char peak_hmem[64];
1883
1884 bytesToHuman(hmem,zmalloc_used_memory());
1885 bytesToHuman(peak_hmem,server.stat_peak_memory);
1886 if (sections++) info = sdscat(info,"\r\n");
1887 info = sdscatprintf(info,
1888 "# Memory\r\n"
1889 "used_memory:%zu\r\n"
1890 "used_memory_human:%s\r\n"
1891 "used_memory_rss:%zu\r\n"
1892 "used_memory_peak:%zu\r\n"
1893 "used_memory_peak_human:%s\r\n"
1894 "used_memory_lua:%lld\r\n"
1895 "mem_fragmentation_ratio:%.2f\r\n"
1896 "mem_allocator:%s\r\n",
1897 zmalloc_used_memory(),
1898 hmem,
1899 zmalloc_get_rss(),
1900 server.stat_peak_memory,
1901 peak_hmem,
1902 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1903 zmalloc_get_fragmentation_ratio(),
1904 ZMALLOC_LIB
1905 );
1906 }
1907
1908 /* Persistence */
1909 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1910 if (sections++) info = sdscat(info,"\r\n");
1911 info = sdscatprintf(info,
1912 "# Persistence\r\n"
1913 "loading:%d\r\n"
1914 "rdb_changes_since_last_save:%lld\r\n"
1915 "rdb_bgsave_in_progress:%d\r\n"
1916 "rdb_last_save_time:%ld\r\n"
1917 "rdb_last_bgsave_status:%s\r\n"
1918 "rdb_last_bgsave_time_sec:%ld\r\n"
1919 "rdb_current_bgsave_time_sec:%ld\r\n"
1920 "aof_enabled:%d\r\n"
1921 "aof_rewrite_in_progress:%d\r\n"
1922 "aof_rewrite_scheduled:%d\r\n"
1923 "aof_last_rewrite_time_sec:%ld\r\n"
1924 "aof_current_rewrite_time_sec:%ld\r\n"
1925 "aof_last_bgrewrite_status:%s\r\n",
1926 server.loading,
1927 server.dirty,
1928 server.rdb_child_pid != -1,
1929 server.lastsave,
1930 (server.lastbgsave_status == REDIS_OK) ? "ok" : "err",
1931 server.rdb_save_time_last,
1932 (server.rdb_child_pid == -1) ?
1933 -1 : time(NULL)-server.rdb_save_time_start,
1934 server.aof_state != REDIS_AOF_OFF,
1935 server.aof_child_pid != -1,
1936 server.aof_rewrite_scheduled,
1937 server.aof_rewrite_time_last,
1938 (server.aof_child_pid == -1) ?
1939 -1 : time(NULL)-server.aof_rewrite_time_start,
1940 (server.aof_lastbgrewrite_status == REDIS_OK) ? "ok" : "err");
1941
1942 if (server.aof_state != REDIS_AOF_OFF) {
1943 info = sdscatprintf(info,
1944 "aof_current_size:%lld\r\n"
1945 "aof_base_size:%lld\r\n"
1946 "aof_pending_rewrite:%d\r\n"
1947 "aof_buffer_length:%zu\r\n"
1948 "aof_rewrite_buffer_length:%lu\r\n"
1949 "aof_pending_bio_fsync:%llu\r\n"
1950 "aof_delayed_fsync:%lu\r\n",
1951 (long long) server.aof_current_size,
1952 (long long) server.aof_rewrite_base_size,
1953 server.aof_rewrite_scheduled,
1954 sdslen(server.aof_buf),
1955 aofRewriteBufferSize(),
1956 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1957 server.aof_delayed_fsync);
1958 }
1959
1960 if (server.loading) {
1961 double perc;
1962 time_t eta, elapsed;
1963 off_t remaining_bytes = server.loading_total_bytes-
1964 server.loading_loaded_bytes;
1965
1966 perc = ((double)server.loading_loaded_bytes /
1967 server.loading_total_bytes) * 100;
1968
1969 elapsed = server.unixtime-server.loading_start_time;
1970 if (elapsed == 0) {
1971 eta = 1; /* A fake 1 second figure if we don't have
1972 enough info */
1973 } else {
1974 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1975 }
1976
1977 info = sdscatprintf(info,
1978 "loading_start_time:%ld\r\n"
1979 "loading_total_bytes:%llu\r\n"
1980 "loading_loaded_bytes:%llu\r\n"
1981 "loading_loaded_perc:%.2f\r\n"
1982 "loading_eta_seconds:%ld\r\n"
1983 ,(unsigned long) server.loading_start_time,
1984 (unsigned long long) server.loading_total_bytes,
1985 (unsigned long long) server.loading_loaded_bytes,
1986 perc,
1987 eta
1988 );
1989 }
1990 }
1991
1992 /* Stats */
1993 if (allsections || defsections || !strcasecmp(section,"stats")) {
1994 if (sections++) info = sdscat(info,"\r\n");
1995 info = sdscatprintf(info,
1996 "# Stats\r\n"
1997 "total_connections_received:%lld\r\n"
1998 "total_commands_processed:%lld\r\n"
1999 "instantaneous_ops_per_sec:%lld\r\n"
2000 "rejected_connections:%lld\r\n"
2001 "expired_keys:%lld\r\n"
2002 "evicted_keys:%lld\r\n"
2003 "keyspace_hits:%lld\r\n"
2004 "keyspace_misses:%lld\r\n"
2005 "pubsub_channels:%ld\r\n"
2006 "pubsub_patterns:%lu\r\n"
2007 "latest_fork_usec:%lld\r\n",
2008 server.stat_numconnections,
2009 server.stat_numcommands,
2010 getOperationsPerSecond(),
2011 server.stat_rejected_conn,
2012 server.stat_expiredkeys,
2013 server.stat_evictedkeys,
2014 server.stat_keyspace_hits,
2015 server.stat_keyspace_misses,
2016 dictSize(server.pubsub_channels),
2017 listLength(server.pubsub_patterns),
2018 server.stat_fork_time);
2019 }
2020
2021 /* Replication */
2022 if (allsections || defsections || !strcasecmp(section,"replication")) {
2023 if (sections++) info = sdscat(info,"\r\n");
2024 info = sdscatprintf(info,
2025 "# Replication\r\n"
2026 "role:%s\r\n",
2027 server.masterhost == NULL ? "master" : "slave");
2028 if (server.masterhost) {
2029 info = sdscatprintf(info,
2030 "master_host:%s\r\n"
2031 "master_port:%d\r\n"
2032 "master_link_status:%s\r\n"
2033 "master_last_io_seconds_ago:%d\r\n"
2034 "master_sync_in_progress:%d\r\n"
2035 ,server.masterhost,
2036 server.masterport,
2037 (server.repl_state == REDIS_REPL_CONNECTED) ?
2038 "up" : "down",
2039 server.master ?
2040 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
2041 server.repl_state == REDIS_REPL_TRANSFER
2042 );
2043
2044 if (server.repl_state == REDIS_REPL_TRANSFER) {
2045 info = sdscatprintf(info,
2046 "master_sync_left_bytes:%lld\r\n"
2047 "master_sync_last_io_seconds_ago:%d\r\n"
2048 , (long long)
2049 (server.repl_transfer_size - server.repl_transfer_read),
2050 (int)(server.unixtime-server.repl_transfer_lastio)
2051 );
2052 }
2053
2054 if (server.repl_state != REDIS_REPL_CONNECTED) {
2055 info = sdscatprintf(info,
2056 "master_link_down_since_seconds:%ld\r\n",
2057 (long)server.unixtime-server.repl_down_since);
2058 }
2059 info = sdscatprintf(info,
2060 "slave_priority:%d\r\n", server.slave_priority);
2061 }
2062 info = sdscatprintf(info,
2063 "connected_slaves:%lu\r\n",
2064 listLength(server.slaves));
2065 if (listLength(server.slaves)) {
2066 int slaveid = 0;
2067 listNode *ln;
2068 listIter li;
2069
2070 listRewind(server.slaves,&li);
2071 while((ln = listNext(&li))) {
2072 redisClient *slave = listNodeValue(ln);
2073 char *state = NULL;
2074 char ip[32];
2075 int port;
2076
2077 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
2078 switch(slave->replstate) {
2079 case REDIS_REPL_WAIT_BGSAVE_START:
2080 case REDIS_REPL_WAIT_BGSAVE_END:
2081 state = "wait_bgsave";
2082 break;
2083 case REDIS_REPL_SEND_BULK:
2084 state = "send_bulk";
2085 break;
2086 case REDIS_REPL_ONLINE:
2087 state = "online";
2088 break;
2089 }
2090 if (state == NULL) continue;
2091 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2092 slaveid,ip,slave->slave_listening_port,state);
2093 slaveid++;
2094 }
2095 }
2096 }
2097
2098 /* CPU */
2099 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2100 if (sections++) info = sdscat(info,"\r\n");
2101 info = sdscatprintf(info,
2102 "# CPU\r\n"
2103 "used_cpu_sys:%.2f\r\n"
2104 "used_cpu_user:%.2f\r\n"
2105 "used_cpu_sys_children:%.2f\r\n"
2106 "used_cpu_user_children:%.2f\r\n",
2107 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2108 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2109 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2110 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2111 }
2112
2113 /* cmdtime */
2114 if (allsections || !strcasecmp(section,"commandstats")) {
2115 if (sections++) info = sdscat(info,"\r\n");
2116 info = sdscatprintf(info, "# Commandstats\r\n");
2117 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2118 for (j = 0; j < numcommands; j++) {
2119 struct redisCommand *c = redisCommandTable+j;
2120
2121 if (!c->calls) continue;
2122 info = sdscatprintf(info,
2123 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2124 c->name, c->calls, c->microseconds,
2125 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2126 }
2127 }
2128
2129 /* Key space */
2130 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2131 if (sections++) info = sdscat(info,"\r\n");
2132 info = sdscatprintf(info, "# Keyspace\r\n");
2133 for (j = 0; j < server.dbnum; j++) {
2134 long long keys, vkeys;
2135
2136 keys = dictSize(server.db[j].dict);
2137 vkeys = dictSize(server.db[j].expires);
2138 if (keys || vkeys) {
2139 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2140 j, keys, vkeys);
2141 }
2142 }
2143 }
2144 return info;
2145 }
2146
2147 void infoCommand(redisClient *c) {
2148 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2149
2150 if (c->argc > 2) {
2151 addReply(c,shared.syntaxerr);
2152 return;
2153 }
2154 sds info = genRedisInfoString(section);
2155 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2156 (unsigned long)sdslen(info)));
2157 addReplySds(c,info);
2158 addReply(c,shared.crlf);
2159 }
2160
2161 void monitorCommand(redisClient *c) {
2162 /* ignore MONITOR if aleady slave or in monitor mode */
2163 if (c->flags & REDIS_SLAVE) return;
2164
2165 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2166 c->slaveseldb = 0;
2167 listAddNodeTail(server.monitors,c);
2168 addReply(c,shared.ok);
2169 }
2170
2171 /* ============================ Maxmemory directive ======================== */
2172
2173 /* This function gets called when 'maxmemory' is set on the config file to limit
2174 * the max memory used by the server, before processing a command.
2175 *
2176 * The goal of the function is to free enough memory to keep Redis under the
2177 * configured memory limit.
2178 *
2179 * The function starts calculating how many bytes should be freed to keep
2180 * Redis under the limit, and enters a loop selecting the best keys to
2181 * evict accordingly to the configured policy.
2182 *
2183 * If all the bytes needed to return back under the limit were freed the
2184 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2185 * should block the execution of commands that will result in more memory
2186 * used by the server.
2187 */
2188 int freeMemoryIfNeeded(void) {
2189 size_t mem_used, mem_tofree, mem_freed;
2190 int slaves = listLength(server.slaves);
2191
2192 /* Remove the size of slaves output buffers and AOF buffer from the
2193 * count of used memory. */
2194 mem_used = zmalloc_used_memory();
2195 if (slaves) {
2196 listIter li;
2197 listNode *ln;
2198
2199 listRewind(server.slaves,&li);
2200 while((ln = listNext(&li))) {
2201 redisClient *slave = listNodeValue(ln);
2202 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2203 if (obuf_bytes > mem_used)
2204 mem_used = 0;
2205 else
2206 mem_used -= obuf_bytes;
2207 }
2208 }
2209 if (server.aof_state != REDIS_AOF_OFF) {
2210 mem_used -= sdslen(server.aof_buf);
2211 mem_used -= aofRewriteBufferSize();
2212 }
2213
2214 /* Check if we are over the memory limit. */
2215 if (mem_used <= server.maxmemory) return REDIS_OK;
2216
2217 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2218 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2219
2220 /* Compute how much memory we need to free. */
2221 mem_tofree = mem_used - server.maxmemory;
2222 mem_freed = 0;
2223 while (mem_freed < mem_tofree) {
2224 int j, k, keys_freed = 0;
2225
2226 for (j = 0; j < server.dbnum; j++) {
2227 long bestval = 0; /* just to prevent warning */
2228 sds bestkey = NULL;
2229 struct dictEntry *de;
2230 redisDb *db = server.db+j;
2231 dict *dict;
2232
2233 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2234 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2235 {
2236 dict = server.db[j].dict;
2237 } else {
2238 dict = server.db[j].expires;
2239 }
2240 if (dictSize(dict) == 0) continue;
2241
2242 /* volatile-random and allkeys-random policy */
2243 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2244 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2245 {
2246 de = dictGetRandomKey(dict);
2247 bestkey = dictGetKey(de);
2248 }
2249
2250 /* volatile-lru and allkeys-lru policy */
2251 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2252 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2253 {
2254 for (k = 0; k < server.maxmemory_samples; k++) {
2255 sds thiskey;
2256 long thisval;
2257 robj *o;
2258
2259 de = dictGetRandomKey(dict);
2260 thiskey = dictGetKey(de);
2261 /* When policy is volatile-lru we need an additonal lookup
2262 * to locate the real key, as dict is set to db->expires. */
2263 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2264 de = dictFind(db->dict, thiskey);
2265 o = dictGetVal(de);
2266 thisval = estimateObjectIdleTime(o);
2267
2268 /* Higher idle time is better candidate for deletion */
2269 if (bestkey == NULL || thisval > bestval) {
2270 bestkey = thiskey;
2271 bestval = thisval;
2272 }
2273 }
2274 }
2275
2276 /* volatile-ttl */
2277 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2278 for (k = 0; k < server.maxmemory_samples; k++) {
2279 sds thiskey;
2280 long thisval;
2281
2282 de = dictGetRandomKey(dict);
2283 thiskey = dictGetKey(de);
2284 thisval = (long) dictGetVal(de);
2285
2286 /* Expire sooner (minor expire unix timestamp) is better
2287 * candidate for deletion */
2288 if (bestkey == NULL || thisval < bestval) {
2289 bestkey = thiskey;
2290 bestval = thisval;
2291 }
2292 }
2293 }
2294
2295 /* Finally remove the selected key. */
2296 if (bestkey) {
2297 long long delta;
2298
2299 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2300 propagateExpire(db,keyobj);
2301 /* We compute the amount of memory freed by dbDelete() alone.
2302 * It is possible that actually the memory needed to propagate
2303 * the DEL in AOF and replication link is greater than the one
2304 * we are freeing removing the key, but we can't account for
2305 * that otherwise we would never exit the loop.
2306 *
2307 * AOF and Output buffer memory will be freed eventually so
2308 * we only care about memory used by the key space. */
2309 delta = (long long) zmalloc_used_memory();
2310 dbDelete(db,keyobj);
2311 delta -= (long long) zmalloc_used_memory();
2312 mem_freed += delta;
2313 server.stat_evictedkeys++;
2314 decrRefCount(keyobj);
2315 keys_freed++;
2316
2317 /* When the memory to free starts to be big enough, we may
2318 * start spending so much time here that is impossible to
2319 * deliver data to the slaves fast enough, so we force the
2320 * transmission here inside the loop. */
2321 if (slaves) flushSlavesOutputBuffers();
2322 }
2323 }
2324 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2325 }
2326 return REDIS_OK;
2327 }
2328
2329 /* =================================== Main! ================================ */
2330
2331 #ifdef __linux__
2332 int linuxOvercommitMemoryValue(void) {
2333 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2334 char buf[64];
2335
2336 if (!fp) return -1;
2337 if (fgets(buf,64,fp) == NULL) {
2338 fclose(fp);
2339 return -1;
2340 }
2341 fclose(fp);
2342
2343 return atoi(buf);
2344 }
2345
2346 void linuxOvercommitMemoryWarning(void) {
2347 if (linuxOvercommitMemoryValue() == 0) {
2348 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.");
2349 }
2350 }
2351 #endif /* __linux__ */
2352
2353 void createPidFile(void) {
2354 /* Try to write the pid file in a best-effort way. */
2355 FILE *fp = fopen(server.pidfile,"w");
2356 if (fp) {
2357 fprintf(fp,"%d\n",(int)getpid());
2358 fclose(fp);
2359 }
2360 }
2361
2362 void daemonize(void) {
2363 int fd;
2364
2365 if (fork() != 0) exit(0); /* parent exits */
2366 setsid(); /* create a new session */
2367
2368 /* Every output goes to /dev/null. If Redis is daemonized but
2369 * the 'logfile' is set to 'stdout' in the configuration file
2370 * it will not log at all. */
2371 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2372 dup2(fd, STDIN_FILENO);
2373 dup2(fd, STDOUT_FILENO);
2374 dup2(fd, STDERR_FILENO);
2375 if (fd > STDERR_FILENO) close(fd);
2376 }
2377 }
2378
2379 void version() {
2380 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2381 REDIS_VERSION,
2382 redisGitSHA1(),
2383 atoi(redisGitDirty()) > 0,
2384 ZMALLOC_LIB,
2385 sizeof(long) == 4 ? 32 : 64);
2386 exit(0);
2387 }
2388
2389 void usage() {
2390 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2391 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2392 fprintf(stderr," ./redis-server -v or --version\n");
2393 fprintf(stderr," ./redis-server -h or --help\n");
2394 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2395 fprintf(stderr,"Examples:\n");
2396 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2397 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2398 fprintf(stderr," ./redis-server --port 7777\n");
2399 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2400 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n\n");
2401 fprintf(stderr,"Sentinel mode:\n");
2402 fprintf(stderr," ./redis-server /etc/sentinel.conf --sentinel\n");
2403 exit(1);
2404 }
2405
2406 void redisAsciiArt(void) {
2407 #include "asciilogo.h"
2408 char *buf = zmalloc(1024*16);
2409 char *mode = "stand alone";
2410
2411 if (server.cluster_enabled) mode = "cluster";
2412 else if (server.sentinel_mode) mode = "sentinel";
2413
2414 snprintf(buf,1024*16,ascii_logo,
2415 REDIS_VERSION,
2416 redisGitSHA1(),
2417 strtol(redisGitDirty(),NULL,10) > 0,
2418 (sizeof(long) == 8) ? "64" : "32",
2419 <<<<<<< HEAD
2420 "stand alone",
2421 server.port,
2422 =======
2423 mode, server.port,
2424 >>>>>>> 6b5daa2... First implementation of Redis Sentinel.
2425 (long) getpid()
2426 );
2427 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2428 zfree(buf);
2429 }
2430
2431 static void sigtermHandler(int sig) {
2432 REDIS_NOTUSED(sig);
2433
2434 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2435 server.shutdown_asap = 1;
2436 }
2437
2438 void setupSignalHandlers(void) {
2439 struct sigaction act;
2440
2441 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2442 * Otherwise, sa_handler is used. */
2443 sigemptyset(&act.sa_mask);
2444 act.sa_flags = 0;
2445 act.sa_handler = sigtermHandler;
2446 sigaction(SIGTERM, &act, NULL);
2447
2448 #ifdef HAVE_BACKTRACE
2449 sigemptyset(&act.sa_mask);
2450 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2451 act.sa_sigaction = sigsegvHandler;
2452 sigaction(SIGSEGV, &act, NULL);
2453 sigaction(SIGBUS, &act, NULL);
2454 sigaction(SIGFPE, &act, NULL);
2455 sigaction(SIGILL, &act, NULL);
2456 #endif
2457 return;
2458 }
2459
2460 void memtest(size_t megabytes, int passes);
2461
2462 <<<<<<< HEAD
2463 void redisOutOfMemoryHandler(size_t allocation_size) {
2464 redisLog(REDIS_WARNING,"Out Of Memory allocating %zu bytes!",
2465 allocation_size);
2466 redisPanic("OOM");
2467 =======
2468 /* Returns 1 if there is --sentinel among the arguments or if
2469 * argv[0] is exactly "redis-sentinel". */
2470 int checkForSentinelMode(int argc, char **argv) {
2471 int j;
2472
2473 if (strstr(argv[0],"redis-sentinel") != NULL) return 1;
2474 for (j = 1; j < argc; j++)
2475 if (!strcmp(argv[j],"--sentinel")) return 1;
2476 return 0;
2477 }
2478
2479 /* Function called at startup to load RDB or AOF file in memory. */
2480 void loadDataFromDisk(void) {
2481 long long start = ustime();
2482 if (server.aof_state == REDIS_AOF_ON) {
2483 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2484 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2485 } else {
2486 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2487 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2488 (float)(ustime()-start)/1000000);
2489 } else if (errno != ENOENT) {
2490 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2491 exit(1);
2492 }
2493 }
2494 >>>>>>> 6b5daa2... First implementation of Redis Sentinel.
2495 }
2496
2497 int main(int argc, char **argv) {
2498 struct timeval tv;
2499
2500 /* We need to initialize our libraries, and the server configuration. */
2501 zmalloc_enable_thread_safeness();
2502 zmalloc_set_oom_handler(redisOutOfMemoryHandler);
2503 srand(time(NULL)^getpid());
2504 gettimeofday(&tv,NULL);
2505 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2506 server.sentinel_mode = checkForSentinelMode(argc,argv);
2507 initServerConfig();
2508
2509 /* We need to init sentinel right now as parsing the configuration file
2510 * in sentinel mode will have the effect of populating the sentinel
2511 * data structures with master nodes to monitor. */
2512 if (server.sentinel_mode) {
2513 initSentinelConfig();
2514 initSentinel();
2515 }
2516
2517 if (argc >= 2) {
2518 int j = 1; /* First option to parse in argv[] */
2519 sds options = sdsempty();
2520 char *configfile = NULL;
2521
2522 /* Handle special options --help and --version */
2523 if (strcmp(argv[1], "-v") == 0 ||
2524 strcmp(argv[1], "--version") == 0) version();
2525 if (strcmp(argv[1], "--help") == 0 ||
2526 strcmp(argv[1], "-h") == 0) usage();
2527 if (strcmp(argv[1], "--test-memory") == 0) {
2528 if (argc == 3) {
2529 memtest(atoi(argv[2]),50);
2530 exit(0);
2531 } else {
2532 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2533 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2534 exit(1);
2535 }
2536 }
2537
2538 /* First argument is the config file name? */
2539 if (argv[j][0] != '-' || argv[j][1] != '-')
2540 configfile = argv[j++];
2541 /* All the other options are parsed and conceptually appended to the
2542 * configuration file. For instance --port 6380 will generate the
2543 * string "port 6380\n" to be parsed after the actual file name
2544 * is parsed, if any. */
2545 while(j != argc) {
2546 if (argv[j][0] == '-' && argv[j][1] == '-') {
2547 /* Option name */
2548 if (sdslen(options)) options = sdscat(options,"\n");
2549 options = sdscat(options,argv[j]+2);
2550 options = sdscat(options," ");
2551 } else {
2552 /* Option argument */
2553 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2554 options = sdscat(options," ");
2555 }
2556 j++;
2557 }
2558 resetServerSaveParams();
2559 loadServerConfig(configfile,options);
2560 sdsfree(options);
2561 } else {
2562 redisLog(REDIS_WARNING,"Warning: no config file specified, using the default config. In order to specify a config file use 'redis-server /path/to/redis.conf'");
2563 }
2564 if (server.daemonize) daemonize();
2565 initServer();
2566 if (server.daemonize) createPidFile();
2567 redisAsciiArt();
2568
2569 if (!server.sentinel_mode) {
2570 /* Things only needed when not runnign in Sentinel mode. */
2571 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2572 #ifdef __linux__
2573 linuxOvercommitMemoryWarning();
2574 #endif
2575 loadDataFromDisk();
2576 if (server.ipfd > 0)
2577 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2578 if (server.sofd > 0)
2579 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2580 }
2581
2582 aeSetBeforeSleepProc(server.el,beforeSleep);
2583 aeMain(server.el);
2584 aeDeleteEventLoop(server.el);
2585 return 0;
2586 }
2587
2588 /* The End */