]> git.saurik.com Git - redis.git/blob - src/redis.c
45ed36c9de13659943f89c2ffc112d1e49499622
[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 */
110 struct redisCommand redisCommandTable[] = {
111 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
112 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
113 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
114 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
115 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
116 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
117 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
118 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
119 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
120 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
121 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
122 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
123 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
124 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
125 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
126 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
127 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
128 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
129 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
130 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
131 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
132 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
133 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
134 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
135 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
136 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
137 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
138 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
139 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
140 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
141 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
142 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
143 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
144 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
145 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
146 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
147 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
148 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
149 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
150 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
151 {"srandmember",srandmemberCommand,2,"rR",0,NULL,1,1,1,0,0},
152 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
153 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
154 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
155 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
156 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
157 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
158 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
159 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
160 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
161 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
162 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
163 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
164 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
165 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
166 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
167 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
168 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
169 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
170 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
171 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
172 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
173 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
174 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
175 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
176 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
177 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
178 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
179 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
180 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
182 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
183 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
184 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
185 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
186 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
187 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
188 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
189 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
190 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
191 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
192 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
193 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
194 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
195 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
196 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
197 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
198 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
199 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
200 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
201 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
202 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
203 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
204 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
205 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
206 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
207 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
208 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
209 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
210 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
211 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
212 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
213 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
214 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
215 {"exec",execCommand,1,"s",0,NULL,0,0,0,0,0},
216 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
217 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
218 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
219 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
220 {"sort",sortCommand,-2,"wmS",0,NULL,1,1,1,0,0},
221 {"info",infoCommand,-1,"r",0,NULL,0,0,0,0,0},
222 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
223 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
224 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
225 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
226 {"slaveof",slaveofCommand,3,"as",0,NULL,0,0,0,0,0},
227 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
228 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
229 {"subscribe",subscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
230 {"unsubscribe",unsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
231 {"psubscribe",psubscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
232 {"punsubscribe",punsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
233 {"publish",publishCommand,3,"pf",0,NULL,0,0,0,0,0},
234 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
235 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
236 {"cluster",clusterCommand,-2,"ar",0,NULL,0,0,0,0,0},
237 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
238 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
239 {"asking",askingCommand,1,"r",0,NULL,0,0,0,0,0},
240 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
241 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
242 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
243 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
244 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
245 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
246 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
247 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0}
248 };
249
250 /*============================ Utility functions ============================ */
251
252 /* Low level logging. To use only for very big messages, otherwise
253 * redisLog() is to prefer. */
254 void redisLogRaw(int level, const char *msg) {
255 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
256 const char *c = ".-*#";
257 FILE *fp;
258 char buf[64];
259 int rawmode = (level & REDIS_LOG_RAW);
260
261 level &= 0xff; /* clear flags */
262 if (level < server.verbosity) return;
263
264 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
265 if (!fp) return;
266
267 if (rawmode) {
268 fprintf(fp,"%s",msg);
269 } else {
270 int off;
271 struct timeval tv;
272
273 gettimeofday(&tv,NULL);
274 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
275 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
276 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
277 }
278 fflush(fp);
279
280 if (server.logfile) fclose(fp);
281
282 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
283 }
284
285 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
286 * is used across the code. The raw version is only used in order to dump
287 * the INFO output on crash. */
288 void redisLog(int level, const char *fmt, ...) {
289 va_list ap;
290 char msg[REDIS_MAX_LOGMSG_LEN];
291
292 if ((level&0xff) < server.verbosity) return;
293
294 va_start(ap, fmt);
295 vsnprintf(msg, sizeof(msg), fmt, ap);
296 va_end(ap);
297
298 redisLogRaw(level,msg);
299 }
300
301 /* Log a fixed message without printf-alike capabilities, in a way that is
302 * safe to call from a signal handler.
303 *
304 * We actually use this only for signals that are not fatal from the point
305 * of view of Redis. Signals that are going to kill the server anyway and
306 * where we need printf-alike features are served by redisLog(). */
307 void redisLogFromHandler(int level, const char *msg) {
308 int fd;
309 char buf[64];
310
311 if ((level&0xff) < server.verbosity ||
312 (server.logfile == NULL && server.daemonize)) return;
313 fd = server.logfile ?
314 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
315 STDOUT_FILENO;
316 if (fd == -1) return;
317 ll2string(buf,sizeof(buf),getpid());
318 if (write(fd,"[",1) == -1) goto err;
319 if (write(fd,buf,strlen(buf)) == -1) goto err;
320 if (write(fd," | signal handler] (",20) == -1) goto err;
321 ll2string(buf,sizeof(buf),time(NULL));
322 if (write(fd,buf,strlen(buf)) == -1) goto err;
323 if (write(fd,") ",2) == -1) goto err;
324 if (write(fd,msg,strlen(msg)) == -1) goto err;
325 if (write(fd,"\n",1) == -1) goto err;
326 err:
327 if (server.logfile) close(fd);
328 }
329
330 /* Redis generally does not try to recover from out of memory conditions
331 * when allocating objects or strings, it is not clear if it will be possible
332 * to report this condition to the client since the networking layer itself
333 * is based on heap allocation for send buffers, so we simply abort.
334 * At least the code will be simpler to read... */
335 void oom(const char *msg) {
336 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
337 sleep(1);
338 abort();
339 }
340
341 /* Return the UNIX time in microseconds */
342 long long ustime(void) {
343 struct timeval tv;
344 long long ust;
345
346 gettimeofday(&tv, NULL);
347 ust = ((long long)tv.tv_sec)*1000000;
348 ust += tv.tv_usec;
349 return ust;
350 }
351
352 /* Return the UNIX time in milliseconds */
353 long long mstime(void) {
354 return ustime()/1000;
355 }
356
357 /* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
358 * exit(), because the latter may interact with the same file objects used by
359 * the parent process. However if we are testing the coverage normal exit() is
360 * used in order to obtain the right coverage information. */
361 void exitFromChild(int retcode) {
362 #ifdef COVERAGE_TEST
363 exit(retcode);
364 #else
365 _exit(retcode);
366 #endif
367 }
368
369 /*====================== Hash table type implementation ==================== */
370
371 /* This is an hash table type that uses the SDS dynamic strings libary as
372 * keys and radis objects as values (objects can hold SDS strings,
373 * lists, sets). */
374
375 void dictVanillaFree(void *privdata, void *val)
376 {
377 DICT_NOTUSED(privdata);
378 zfree(val);
379 }
380
381 void dictListDestructor(void *privdata, void *val)
382 {
383 DICT_NOTUSED(privdata);
384 listRelease((list*)val);
385 }
386
387 int dictSdsKeyCompare(void *privdata, const void *key1,
388 const void *key2)
389 {
390 int l1,l2;
391 DICT_NOTUSED(privdata);
392
393 l1 = sdslen((sds)key1);
394 l2 = sdslen((sds)key2);
395 if (l1 != l2) return 0;
396 return memcmp(key1, key2, l1) == 0;
397 }
398
399 /* A case insensitive version used for the command lookup table. */
400 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
401 const void *key2)
402 {
403 DICT_NOTUSED(privdata);
404
405 return strcasecmp(key1, key2) == 0;
406 }
407
408 void dictRedisObjectDestructor(void *privdata, void *val)
409 {
410 DICT_NOTUSED(privdata);
411
412 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
413 decrRefCount(val);
414 }
415
416 void dictSdsDestructor(void *privdata, void *val)
417 {
418 DICT_NOTUSED(privdata);
419
420 sdsfree(val);
421 }
422
423 int dictObjKeyCompare(void *privdata, const void *key1,
424 const void *key2)
425 {
426 const robj *o1 = key1, *o2 = key2;
427 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
428 }
429
430 unsigned int dictObjHash(const void *key) {
431 const robj *o = key;
432 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
433 }
434
435 unsigned int dictSdsHash(const void *key) {
436 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
437 }
438
439 unsigned int dictSdsCaseHash(const void *key) {
440 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
441 }
442
443 int dictEncObjKeyCompare(void *privdata, const void *key1,
444 const void *key2)
445 {
446 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
447 int cmp;
448
449 if (o1->encoding == REDIS_ENCODING_INT &&
450 o2->encoding == REDIS_ENCODING_INT)
451 return o1->ptr == o2->ptr;
452
453 o1 = getDecodedObject(o1);
454 o2 = getDecodedObject(o2);
455 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
456 decrRefCount(o1);
457 decrRefCount(o2);
458 return cmp;
459 }
460
461 unsigned int dictEncObjHash(const void *key) {
462 robj *o = (robj*) key;
463
464 if (o->encoding == REDIS_ENCODING_RAW) {
465 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
466 } else {
467 if (o->encoding == REDIS_ENCODING_INT) {
468 char buf[32];
469 int len;
470
471 len = ll2string(buf,32,(long)o->ptr);
472 return dictGenHashFunction((unsigned char*)buf, len);
473 } else {
474 unsigned int hash;
475
476 o = getDecodedObject(o);
477 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
478 decrRefCount(o);
479 return hash;
480 }
481 }
482 }
483
484 /* Sets type hash table */
485 dictType setDictType = {
486 dictEncObjHash, /* hash function */
487 NULL, /* key dup */
488 NULL, /* val dup */
489 dictEncObjKeyCompare, /* key compare */
490 dictRedisObjectDestructor, /* key destructor */
491 NULL /* val destructor */
492 };
493
494 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
495 dictType zsetDictType = {
496 dictEncObjHash, /* hash function */
497 NULL, /* key dup */
498 NULL, /* val dup */
499 dictEncObjKeyCompare, /* key compare */
500 dictRedisObjectDestructor, /* key destructor */
501 NULL /* val destructor */
502 };
503
504 /* Db->dict, keys are sds strings, vals are Redis objects. */
505 dictType dbDictType = {
506 dictSdsHash, /* hash function */
507 NULL, /* key dup */
508 NULL, /* val dup */
509 dictSdsKeyCompare, /* key compare */
510 dictSdsDestructor, /* key destructor */
511 dictRedisObjectDestructor /* val destructor */
512 };
513
514 /* Db->expires */
515 dictType keyptrDictType = {
516 dictSdsHash, /* hash function */
517 NULL, /* key dup */
518 NULL, /* val dup */
519 dictSdsKeyCompare, /* key compare */
520 NULL, /* key destructor */
521 NULL /* val destructor */
522 };
523
524 /* Command table. sds string -> command struct pointer. */
525 dictType commandTableDictType = {
526 dictSdsCaseHash, /* hash function */
527 NULL, /* key dup */
528 NULL, /* val dup */
529 dictSdsKeyCaseCompare, /* key compare */
530 dictSdsDestructor, /* key destructor */
531 NULL /* val destructor */
532 };
533
534 /* Hash type hash table (note that small hashes are represented with zimpaps) */
535 dictType hashDictType = {
536 dictEncObjHash, /* hash function */
537 NULL, /* key dup */
538 NULL, /* val dup */
539 dictEncObjKeyCompare, /* key compare */
540 dictRedisObjectDestructor, /* key destructor */
541 dictRedisObjectDestructor /* val destructor */
542 };
543
544 /* Keylist hash table type has unencoded redis objects as keys and
545 * lists as values. It's used for blocking operations (BLPOP) and to
546 * map swapped keys to a list of clients waiting for this keys to be loaded. */
547 dictType keylistDictType = {
548 dictObjHash, /* hash function */
549 NULL, /* key dup */
550 NULL, /* val dup */
551 dictObjKeyCompare, /* key compare */
552 dictRedisObjectDestructor, /* key destructor */
553 dictListDestructor /* val destructor */
554 };
555
556 /* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
557 * clusterNode structures. */
558 dictType clusterNodesDictType = {
559 dictSdsHash, /* hash function */
560 NULL, /* key dup */
561 NULL, /* val dup */
562 dictSdsKeyCompare, /* key compare */
563 dictSdsDestructor, /* key destructor */
564 NULL /* val destructor */
565 };
566
567 int htNeedsResize(dict *dict) {
568 long long size, used;
569
570 size = dictSlots(dict);
571 used = dictSize(dict);
572 return (size && used && size > DICT_HT_INITIAL_SIZE &&
573 (used*100/size < REDIS_HT_MINFILL));
574 }
575
576 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
577 * we resize the hash table to save memory */
578 void tryResizeHashTables(void) {
579 int j;
580
581 for (j = 0; j < server.dbnum; j++) {
582 if (htNeedsResize(server.db[j].dict))
583 dictResize(server.db[j].dict);
584 if (htNeedsResize(server.db[j].expires))
585 dictResize(server.db[j].expires);
586 }
587 }
588
589 /* Our hash table implementation performs rehashing incrementally while
590 * we write/read from the hash table. Still if the server is idle, the hash
591 * table will use two tables for a long time. So we try to use 1 millisecond
592 * of CPU time at every serverCron() loop in order to rehash some key. */
593 void incrementallyRehash(void) {
594 int j;
595
596 for (j = 0; j < server.dbnum; j++) {
597 /* Keys dictionary */
598 if (dictIsRehashing(server.db[j].dict)) {
599 dictRehashMilliseconds(server.db[j].dict,1);
600 break; /* already used our millisecond for this loop... */
601 }
602 /* Expires */
603 if (dictIsRehashing(server.db[j].expires)) {
604 dictRehashMilliseconds(server.db[j].expires,1);
605 break; /* already used our millisecond for this loop... */
606 }
607 }
608 }
609
610 /* This function is called once a background process of some kind terminates,
611 * as we want to avoid resizing the hash tables when there is a child in order
612 * to play well with copy-on-write (otherwise when a resize happens lots of
613 * memory pages are copied). The goal of this function is to update the ability
614 * for dict.c to resize the hash tables accordingly to the fact we have o not
615 * running childs. */
616 void updateDictResizePolicy(void) {
617 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
618 dictEnableResize();
619 else
620 dictDisableResize();
621 }
622
623 /* ======================= Cron: called every 100 ms ======================== */
624
625 /* Try to expire a few timed out keys. The algorithm used is adaptive and
626 * will use few CPU cycles if there are few expiring keys, otherwise
627 * it will get more aggressive to avoid that too much memory is used by
628 * keys that can be removed from the keyspace. */
629 void activeExpireCycle(void) {
630 int j;
631 long long start = ustime(), timelimit;
632
633 /* We can use at max REDIS_EXPIRELOOKUPS_TIME_PERC percentage of CPU time
634 * per iteration. Since this function gets called with a frequency of
635 * REDIS_HZ times per second, the following is the max amount of
636 * microseconds we can spend in this function. */
637 timelimit = 1000000*REDIS_EXPIRELOOKUPS_TIME_PERC/REDIS_HZ/100;
638 if (timelimit <= 0) timelimit = 1;
639
640 for (j = 0; j < server.dbnum; j++) {
641 int expired, iteration = 0;
642 redisDb *db = server.db+j;
643
644 /* Continue to expire if at the end of the cycle more than 25%
645 * of the keys were expired. */
646 do {
647 unsigned long num = dictSize(db->expires);
648 unsigned long slots = dictSlots(db->expires);
649 long long now = mstime();
650
651 /* When there are less than 1% filled slots getting random
652 * keys is expensive, so stop here waiting for better times...
653 * The dictionary will be resized asap. */
654 if (num && slots > DICT_HT_INITIAL_SIZE &&
655 (num*100/slots < 1)) break;
656
657 expired = 0;
658 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
659 num = REDIS_EXPIRELOOKUPS_PER_CRON;
660 while (num--) {
661 dictEntry *de;
662 long long t;
663
664 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
665 t = dictGetSignedIntegerVal(de);
666 if (now > t) {
667 sds key = dictGetKey(de);
668 robj *keyobj = createStringObject(key,sdslen(key));
669
670 propagateExpire(db,keyobj);
671 dbDelete(db,keyobj);
672 decrRefCount(keyobj);
673 expired++;
674 server.stat_expiredkeys++;
675 }
676 }
677 /* We can't block forever here even if there are many keys to
678 * expire. So after a given amount of milliseconds return to the
679 * caller waiting for the other active expire cycle. */
680 iteration++;
681 if ((iteration & 0xff) == 0 && /* Check once every 255 iterations */
682 (ustime()-start) > timelimit) return;
683 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
684 }
685 }
686
687 void updateLRUClock(void) {
688 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
689 REDIS_LRU_CLOCK_MAX;
690 }
691
692
693 /* Add a sample to the operations per second array of samples. */
694 void trackOperationsPerSecond(void) {
695 long long t = mstime() - server.ops_sec_last_sample_time;
696 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
697 long long ops_sec;
698
699 ops_sec = t > 0 ? (ops*1000/t) : 0;
700
701 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
702 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
703 server.ops_sec_last_sample_time = mstime();
704 server.ops_sec_last_sample_ops = server.stat_numcommands;
705 }
706
707 /* Return the mean of all the samples. */
708 long long getOperationsPerSecond(void) {
709 int j;
710 long long sum = 0;
711
712 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
713 sum += server.ops_sec_samples[j];
714 return sum / REDIS_OPS_SEC_SAMPLES;
715 }
716
717 /* Check for timeouts. Returns non-zero if the client was terminated */
718 int clientsCronHandleTimeout(redisClient *c) {
719 time_t now = server.unixtime;
720
721 if (server.maxidletime &&
722 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
723 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
724 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
725 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
726 listLength(c->pubsub_patterns) == 0 &&
727 (now - c->lastinteraction > server.maxidletime))
728 {
729 redisLog(REDIS_VERBOSE,"Closing idle client");
730 freeClient(c);
731 return 1;
732 } else if (c->flags & REDIS_BLOCKED) {
733 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
734 addReply(c,shared.nullmultibulk);
735 unblockClientWaitingData(c);
736 }
737 }
738 return 0;
739 }
740
741 /* The client query buffer is an sds.c string that can end with a lot of
742 * free space not used, this function reclaims space if needed.
743 *
744 * The funciton always returns 0 as it never terminates the client. */
745 int clientsCronResizeQueryBuffer(redisClient *c) {
746 size_t querybuf_size = sdsAllocSize(c->querybuf);
747 time_t idletime = server.unixtime - c->lastinteraction;
748
749 /* There are two conditions to resize the query buffer:
750 * 1) Query buffer is > BIG_ARG and too big for latest peak.
751 * 2) Client is inactive and the buffer is bigger than 1k. */
752 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
753 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
754 (querybuf_size > 1024 && idletime > 2))
755 {
756 /* Only resize the query buffer if it is actually wasting space. */
757 if (sdsavail(c->querybuf) > 1024) {
758 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
759 }
760 }
761 /* Reset the peak again to capture the peak memory usage in the next
762 * cycle. */
763 c->querybuf_peak = 0;
764 return 0;
765 }
766
767 void clientsCron(void) {
768 /* Make sure to process at least 1/(REDIS_HZ*10) of clients per call.
769 * Since this function is called REDIS_HZ times per second we are sure that
770 * in the worst case we process all the clients in 10 seconds.
771 * In normal conditions (a reasonable number of clients) we process
772 * all the clients in a shorter time. */
773 int numclients = listLength(server.clients);
774 int iterations = numclients/(REDIS_HZ*10);
775
776 if (iterations < 50)
777 iterations = (numclients < 50) ? numclients : 50;
778 while(listLength(server.clients) && iterations--) {
779 redisClient *c;
780 listNode *head;
781
782 /* Rotate the list, take the current head, process.
783 * This way if the client must be removed from the list it's the
784 * first element and we don't incur into O(N) computation. */
785 listRotate(server.clients);
786 head = listFirst(server.clients);
787 c = listNodeValue(head);
788 /* The following functions do different service checks on the client.
789 * The protocol is that they return non-zero if the client was
790 * terminated. */
791 if (clientsCronHandleTimeout(c)) continue;
792 if (clientsCronResizeQueryBuffer(c)) continue;
793 }
794 }
795
796 /* This is our timer interrupt, called REDIS_HZ times per second.
797 * Here is where we do a number of things that need to be done asynchronously.
798 * For instance:
799 *
800 * - Active expired keys collection (it is also performed in a lazy way on
801 * lookup).
802 * - Software watchdong.
803 * - Update some statistic.
804 * - Incremental rehashing of the DBs hash tables.
805 * - Triggering BGSAVE / AOF rewrite, and handling of terminated children.
806 * - Clients timeout of differnet kinds.
807 * - Replication reconnection.
808 * - Many more...
809 *
810 * Everything directly called here will be called REDIS_HZ times per second,
811 * so in order to throttle execution of things we want to do less frequently
812 * a macro is used: run_with_period(milliseconds) { .... }
813 */
814
815 /* Using the following macro you can run code inside serverCron() with the
816 * specified period, specified in milliseconds.
817 * The actual resolution depends on REDIS_HZ. */
818 #define run_with_period(_ms_) if (!(loops % ((_ms_)/(1000/REDIS_HZ))))
819
820 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
821 int j, loops = server.cronloops;
822 REDIS_NOTUSED(eventLoop);
823 REDIS_NOTUSED(id);
824 REDIS_NOTUSED(clientData);
825
826 /* Software watchdog: deliver the SIGALRM that will reach the signal
827 * handler if we don't return here fast enough. */
828 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
829
830 /* We take a cached value of the unix time in the global state because
831 * with virtual memory and aging there is to store the current time
832 * in objects at every object access, and accuracy is not needed.
833 * To access a global var is faster than calling time(NULL) */
834 server.unixtime = time(NULL);
835
836 run_with_period(100) trackOperationsPerSecond();
837
838 /* We have just 22 bits per object for LRU information.
839 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
840 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
841 *
842 * Note that even if this will wrap after 1.5 years it's not a problem,
843 * everything will still work but just some object will appear younger
844 * to Redis. But for this to happen a given object should never be touched
845 * for 1.5 years.
846 *
847 * Note that you can change the resolution altering the
848 * REDIS_LRU_CLOCK_RESOLUTION define.
849 */
850 updateLRUClock();
851
852 /* Record the max memory used since the server was started. */
853 if (zmalloc_used_memory() > server.stat_peak_memory)
854 server.stat_peak_memory = zmalloc_used_memory();
855
856 /* We received a SIGTERM, shutting down here in a safe way, as it is
857 * not ok doing so inside the signal handler. */
858 if (server.shutdown_asap) {
859 if (prepareForShutdown(0) == REDIS_OK) exit(0);
860 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
861 }
862
863 /* Show some info about non-empty databases */
864 run_with_period(5000) {
865 for (j = 0; j < server.dbnum; j++) {
866 long long size, used, vkeys;
867
868 size = dictSlots(server.db[j].dict);
869 used = dictSize(server.db[j].dict);
870 vkeys = dictSize(server.db[j].expires);
871 if (used || vkeys) {
872 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
873 /* dictPrintStats(server.dict); */
874 }
875 }
876 }
877
878 /* We don't want to resize the hash tables while a bacground saving
879 * is in progress: the saving child is created using fork() that is
880 * implemented with a copy-on-write semantic in most modern systems, so
881 * if we resize the HT while there is the saving child at work actually
882 * a lot of memory movements in the parent will cause a lot of pages
883 * copied. */
884 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
885 tryResizeHashTables();
886 if (server.activerehashing) incrementallyRehash();
887 }
888
889 /* Show information about connected clients */
890 run_with_period(5000) {
891 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
892 listLength(server.clients)-listLength(server.slaves),
893 listLength(server.slaves),
894 zmalloc_used_memory());
895 }
896
897 /* We need to do a few operations on clients asynchronously. */
898 clientsCron();
899
900 /* Start a scheduled AOF rewrite if this was requested by the user while
901 * a BGSAVE was in progress. */
902 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
903 server.aof_rewrite_scheduled)
904 {
905 rewriteAppendOnlyFileBackground();
906 }
907
908 /* Check if a background saving or AOF rewrite in progress terminated. */
909 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
910 int statloc;
911 pid_t pid;
912
913 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
914 int exitcode = WEXITSTATUS(statloc);
915 int bysignal = 0;
916
917 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
918
919 if (pid == server.rdb_child_pid) {
920 backgroundSaveDoneHandler(exitcode,bysignal);
921 } else {
922 backgroundRewriteDoneHandler(exitcode,bysignal);
923 }
924 updateDictResizePolicy();
925 }
926 } else {
927 /* If there is not a background saving/rewrite in progress check if
928 * we have to save/rewrite now */
929 for (j = 0; j < server.saveparamslen; j++) {
930 struct saveparam *sp = server.saveparams+j;
931
932 if (server.dirty >= sp->changes &&
933 server.unixtime-server.lastsave > sp->seconds) {
934 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
935 sp->changes, sp->seconds);
936 rdbSaveBackground(server.rdb_filename);
937 break;
938 }
939 }
940
941 /* Trigger an AOF rewrite if needed */
942 if (server.rdb_child_pid == -1 &&
943 server.aof_child_pid == -1 &&
944 server.aof_rewrite_perc &&
945 server.aof_current_size > server.aof_rewrite_min_size)
946 {
947 long long base = server.aof_rewrite_base_size ?
948 server.aof_rewrite_base_size : 1;
949 long long growth = (server.aof_current_size*100/base) - 100;
950 if (growth >= server.aof_rewrite_perc) {
951 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
952 rewriteAppendOnlyFileBackground();
953 }
954 }
955 }
956
957
958 /* If we postponed an AOF buffer flush, let's try to do it every time the
959 * cron function is called. */
960 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
961
962 /* Expire a few keys per cycle, only if this is a master.
963 * On slaves we wait for DEL operations synthesized by the master
964 * in order to guarantee a strict consistency. */
965 if (server.masterhost == NULL) activeExpireCycle();
966
967 /* Close clients that need to be closed asynchronous */
968 freeClientsInAsyncFreeQueue();
969
970 /* Replication cron function -- used to reconnect to master and
971 * to detect transfer failures. */
972 run_with_period(1000) replicationCron();
973
974 /* Run other sub-systems specific cron jobs */
975 run_with_period(1000) {
976 if (server.cluster_enabled) clusterCron();
977 }
978
979 server.cronloops++;
980 return 1000/REDIS_HZ;
981 }
982
983 /* This function gets called every time Redis is entering the
984 * main loop of the event driven library, that is, before to sleep
985 * for ready file descriptors. */
986 void beforeSleep(struct aeEventLoop *eventLoop) {
987 REDIS_NOTUSED(eventLoop);
988 listNode *ln;
989 redisClient *c;
990
991 /* Try to process pending commands for clients that were just unblocked. */
992 while (listLength(server.unblocked_clients)) {
993 ln = listFirst(server.unblocked_clients);
994 redisAssert(ln != NULL);
995 c = ln->value;
996 listDelNode(server.unblocked_clients,ln);
997 c->flags &= ~REDIS_UNBLOCKED;
998
999 /* Process remaining data in the input buffer. */
1000 if (c->querybuf && sdslen(c->querybuf) > 0) {
1001 server.current_client = c;
1002 processInputBuffer(c);
1003 server.current_client = NULL;
1004 }
1005 }
1006
1007 /* Write the AOF buffer on disk */
1008 flushAppendOnlyFile(0);
1009 }
1010
1011 /* =========================== Server initialization ======================== */
1012
1013 void createSharedObjects(void) {
1014 int j;
1015
1016 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
1017 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
1018 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
1019 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
1020 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
1021 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
1022 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
1023 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
1024 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
1025 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
1026 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
1027 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
1028 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
1029 "-ERR Operation against a key holding the wrong kind of value\r\n"));
1030 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
1031 "-ERR no such key\r\n"));
1032 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
1033 "-ERR syntax error\r\n"));
1034 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
1035 "-ERR source and destination objects are the same\r\n"));
1036 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
1037 "-ERR index out of range\r\n"));
1038 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
1039 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
1040 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
1041 "-LOADING Redis is loading the dataset in memory\r\n"));
1042 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
1043 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
1044 shared.masterdownerr = createObject(REDIS_STRING,sdsnew(
1045 "-MASTERDOWN Link with MASTER is down and slave-serve-stale-data is set to 'no'.\r\n"));
1046 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
1047 "-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"));
1048 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
1049 "-READONLY You can't write against a read only slave.\r\n"));
1050 shared.oomerr = createObject(REDIS_STRING,sdsnew(
1051 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
1052 shared.space = createObject(REDIS_STRING,sdsnew(" "));
1053 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1054 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1055
1056 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1057 shared.select[j] = createObject(REDIS_STRING,
1058 sdscatprintf(sdsempty(),"select %d\r\n", j));
1059 }
1060 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1061 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1062 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1063 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1064 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1065 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1066 shared.del = createStringObject("DEL",3);
1067 shared.rpop = createStringObject("RPOP",4);
1068 shared.lpop = createStringObject("LPOP",4);
1069 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1070 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1071 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1072 }
1073 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1074 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1075 sdscatprintf(sdsempty(),"*%d\r\n",j));
1076 shared.bulkhdr[j] = createObject(REDIS_STRING,
1077 sdscatprintf(sdsempty(),"$%d\r\n",j));
1078 }
1079 }
1080
1081 void initServerConfig() {
1082 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1083 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1084 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1085 server.port = REDIS_SERVERPORT;
1086 server.bindaddr = NULL;
1087 server.unixsocket = NULL;
1088 server.unixsocketperm = 0;
1089 server.ipfd = -1;
1090 server.sofd = -1;
1091 server.dbnum = REDIS_DEFAULT_DBNUM;
1092 server.verbosity = REDIS_NOTICE;
1093 server.maxidletime = REDIS_MAXIDLETIME;
1094 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1095 server.saveparams = NULL;
1096 server.loading = 0;
1097 server.logfile = NULL; /* NULL = log on standard output */
1098 server.syslog_enabled = 0;
1099 server.syslog_ident = zstrdup("redis");
1100 server.syslog_facility = LOG_LOCAL0;
1101 server.daemonize = 0;
1102 server.aof_state = REDIS_AOF_OFF;
1103 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1104 server.aof_no_fsync_on_rewrite = 0;
1105 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1106 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1107 server.aof_rewrite_base_size = 0;
1108 server.aof_rewrite_scheduled = 0;
1109 server.aof_last_fsync = time(NULL);
1110 server.aof_delayed_fsync = 0;
1111 server.aof_fd = -1;
1112 server.aof_selected_db = -1; /* Make sure the first time will not match */
1113 server.aof_flush_postponed_start = 0;
1114 server.pidfile = zstrdup("/var/run/redis.pid");
1115 server.rdb_filename = zstrdup("dump.rdb");
1116 server.aof_filename = zstrdup("appendonly.aof");
1117 server.requirepass = NULL;
1118 server.rdb_compression = 1;
1119 server.rdb_checksum = 1;
1120 server.activerehashing = 1;
1121 server.maxclients = REDIS_MAX_CLIENTS;
1122 server.bpop_blocked_clients = 0;
1123 server.maxmemory = 0;
1124 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1125 server.maxmemory_samples = 3;
1126 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1127 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1128 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1129 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1130 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1131 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1132 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1133 server.shutdown_asap = 0;
1134 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1135 server.repl_timeout = REDIS_REPL_TIMEOUT;
1136 server.cluster_enabled = 0;
1137 server.cluster.configfile = zstrdup("nodes.conf");
1138 server.lua_caller = NULL;
1139 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1140 server.lua_client = NULL;
1141 server.lua_timedout = 0;
1142
1143 updateLRUClock();
1144 resetServerSaveParams();
1145
1146 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1147 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1148 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1149 /* Replication related */
1150 server.masterauth = NULL;
1151 server.masterhost = NULL;
1152 server.masterport = 6379;
1153 server.master = NULL;
1154 server.repl_state = REDIS_REPL_NONE;
1155 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1156 server.repl_serve_stale_data = 1;
1157 server.repl_slave_ro = 1;
1158 server.repl_down_since = time(NULL);
1159
1160 /* Client output buffer limits */
1161 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1162 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1163 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1164 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1165 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1166 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1167 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1168 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1169 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1170
1171 /* Double constants initialization */
1172 R_Zero = 0.0;
1173 R_PosInf = 1.0/R_Zero;
1174 R_NegInf = -1.0/R_Zero;
1175 R_Nan = R_Zero/R_Zero;
1176
1177 /* Command table -- we intiialize it here as it is part of the
1178 * initial configuration, since command names may be changed via
1179 * redis.conf using the rename-command directive. */
1180 server.commands = dictCreate(&commandTableDictType,NULL);
1181 populateCommandTable();
1182 server.delCommand = lookupCommandByCString("del");
1183 server.multiCommand = lookupCommandByCString("multi");
1184 server.lpushCommand = lookupCommandByCString("lpush");
1185
1186 /* Slow log */
1187 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1188 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1189
1190 /* Debugging */
1191 server.assert_failed = "<no assertion failed>";
1192 server.assert_file = "<no file>";
1193 server.assert_line = 0;
1194 server.bug_report_start = 0;
1195 server.watchdog_period = 0;
1196 }
1197
1198 /* This function will try to raise the max number of open files accordingly to
1199 * the configured max number of clients. It will also account for 32 additional
1200 * file descriptors as we need a few more for persistence, listening
1201 * sockets, log files and so forth.
1202 *
1203 * If it will not be possible to set the limit accordingly to the configured
1204 * max number of clients, the function will do the reverse setting
1205 * server.maxclients to the value that we can actually handle. */
1206 void adjustOpenFilesLimit(void) {
1207 rlim_t maxfiles = server.maxclients+32;
1208 struct rlimit limit;
1209
1210 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1211 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1212 strerror(errno));
1213 server.maxclients = 1024-32;
1214 } else {
1215 rlim_t oldlimit = limit.rlim_cur;
1216
1217 /* Set the max number of files if the current limit is not enough
1218 * for our needs. */
1219 if (oldlimit < maxfiles) {
1220 rlim_t f;
1221
1222 f = maxfiles;
1223 while(f > oldlimit) {
1224 limit.rlim_cur = f;
1225 limit.rlim_max = f;
1226 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1227 f -= 128;
1228 }
1229 if (f < oldlimit) f = oldlimit;
1230 if (f != maxfiles) {
1231 server.maxclients = f-32;
1232 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1233 (int) maxfiles, strerror(errno), (int) server.maxclients);
1234 } else {
1235 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1236 (int) maxfiles);
1237 }
1238 }
1239 }
1240 }
1241
1242 void initServer() {
1243 int j;
1244
1245 signal(SIGHUP, SIG_IGN);
1246 signal(SIGPIPE, SIG_IGN);
1247 setupSignalHandlers();
1248
1249 if (server.syslog_enabled) {
1250 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1251 server.syslog_facility);
1252 }
1253
1254 server.current_client = NULL;
1255 server.clients = listCreate();
1256 server.clients_to_close = listCreate();
1257 server.slaves = listCreate();
1258 server.monitors = listCreate();
1259 server.unblocked_clients = listCreate();
1260
1261 createSharedObjects();
1262 adjustOpenFilesLimit();
1263 server.el = aeCreateEventLoop(server.maxclients+1024);
1264 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1265
1266 if (server.port != 0) {
1267 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1268 if (server.ipfd == ANET_ERR) {
1269 redisLog(REDIS_WARNING, "Opening port %d: %s",
1270 server.port, server.neterr);
1271 exit(1);
1272 }
1273 }
1274 if (server.unixsocket != NULL) {
1275 unlink(server.unixsocket); /* don't care if this fails */
1276 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1277 if (server.sofd == ANET_ERR) {
1278 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1279 exit(1);
1280 }
1281 }
1282 if (server.ipfd < 0 && server.sofd < 0) {
1283 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1284 exit(1);
1285 }
1286 for (j = 0; j < server.dbnum; j++) {
1287 server.db[j].dict = dictCreate(&dbDictType,NULL);
1288 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1289 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1290 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1291 server.db[j].id = j;
1292 }
1293 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1294 server.pubsub_patterns = listCreate();
1295 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1296 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1297 server.cronloops = 0;
1298 server.rdb_child_pid = -1;
1299 server.aof_child_pid = -1;
1300 server.aof_rewrite_buf = sdsempty();
1301 server.aof_buf = sdsempty();
1302 server.lastsave = time(NULL);
1303 server.dirty = 0;
1304 server.stat_numcommands = 0;
1305 server.stat_numconnections = 0;
1306 server.stat_expiredkeys = 0;
1307 server.stat_evictedkeys = 0;
1308 server.stat_starttime = time(NULL);
1309 server.stat_keyspace_misses = 0;
1310 server.stat_keyspace_hits = 0;
1311 server.stat_peak_memory = 0;
1312 server.stat_fork_time = 0;
1313 server.stat_rejected_conn = 0;
1314 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1315 server.ops_sec_idx = 0;
1316 server.ops_sec_last_sample_time = mstime();
1317 server.ops_sec_last_sample_ops = 0;
1318 server.unixtime = time(NULL);
1319 server.lastbgsave_status = REDIS_OK;
1320 server.stop_writes_on_bgsave_err = 1;
1321 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1322 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1323 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1324 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1325 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1326
1327 if (server.aof_state == REDIS_AOF_ON) {
1328 server.aof_fd = open(server.aof_filename,
1329 O_WRONLY|O_APPEND|O_CREAT,0644);
1330 if (server.aof_fd == -1) {
1331 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1332 strerror(errno));
1333 exit(1);
1334 }
1335 }
1336
1337 /* 32 bit instances are limited to 4GB of address space, so if there is
1338 * no explicit limit in the user provided configuration we set a limit
1339 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1340 * useless crashes of the Redis instance. */
1341 if (server.arch_bits == 32 && server.maxmemory == 0) {
1342 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1343 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1344 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1345 }
1346
1347 if (server.cluster_enabled) clusterInit();
1348 scriptingInit();
1349 slowlogInit();
1350 bioInit();
1351 }
1352
1353 /* Populates the Redis Command Table starting from the hard coded list
1354 * we have on top of redis.c file. */
1355 void populateCommandTable(void) {
1356 int j;
1357 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1358
1359 for (j = 0; j < numcommands; j++) {
1360 struct redisCommand *c = redisCommandTable+j;
1361 char *f = c->sflags;
1362 int retval;
1363
1364 while(*f != '\0') {
1365 switch(*f) {
1366 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1367 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1368 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1369 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1370 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1371 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1372 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1373 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1374 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1375 default: redisPanic("Unsupported command flag"); break;
1376 }
1377 f++;
1378 }
1379
1380 retval = dictAdd(server.commands, sdsnew(c->name), c);
1381 assert(retval == DICT_OK);
1382 }
1383 }
1384
1385 void resetCommandTableStats(void) {
1386 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1387 int j;
1388
1389 for (j = 0; j < numcommands; j++) {
1390 struct redisCommand *c = redisCommandTable+j;
1391
1392 c->microseconds = 0;
1393 c->calls = 0;
1394 }
1395 }
1396
1397 /* ========================== Redis OP Array API ============================ */
1398
1399 void redisOpArrayInit(redisOpArray *oa) {
1400 oa->ops = NULL;
1401 oa->numops = 0;
1402 }
1403
1404 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1405 robj **argv, int argc, int target)
1406 {
1407 redisOp *op;
1408
1409 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1410 op = oa->ops+oa->numops;
1411 op->cmd = cmd;
1412 op->dbid = dbid;
1413 op->argv = argv;
1414 op->argc = argc;
1415 op->target = target;
1416 oa->numops++;
1417 return oa->numops;
1418 }
1419
1420 void redisOpArrayFree(redisOpArray *oa) {
1421 while(oa->numops) {
1422 int j;
1423 redisOp *op;
1424
1425 oa->numops--;
1426 op = oa->ops+oa->numops;
1427 for (j = 0; j < op->argc; j++)
1428 decrRefCount(op->argv[j]);
1429 zfree(op->argv);
1430 }
1431 zfree(oa->ops);
1432 }
1433
1434 /* ====================== Commands lookup and execution ===================== */
1435
1436 struct redisCommand *lookupCommand(sds name) {
1437 return dictFetchValue(server.commands, name);
1438 }
1439
1440 struct redisCommand *lookupCommandByCString(char *s) {
1441 struct redisCommand *cmd;
1442 sds name = sdsnew(s);
1443
1444 cmd = dictFetchValue(server.commands, name);
1445 sdsfree(name);
1446 return cmd;
1447 }
1448
1449 /* Propagate the specified command (in the context of the specified database id)
1450 * to AOF, Slaves and Monitors.
1451 *
1452 * flags are an xor between:
1453 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1454 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1455 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1456 */
1457 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1458 int flags)
1459 {
1460 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1461 feedAppendOnlyFile(cmd,dbid,argv,argc);
1462 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1463 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1464 }
1465
1466 /* Used inside commands to schedule the propagation of additional commands
1467 * after the current command is propagated to AOF / Replication. */
1468 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1469 int target)
1470 {
1471 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1472 }
1473
1474 /* Call() is the core of Redis execution of a command */
1475 void call(redisClient *c, int flags) {
1476 long long dirty, start = ustime(), duration;
1477
1478 /* Sent the command to clients in MONITOR mode, only if the commands are
1479 * not geneated from reading an AOF. */
1480 if (listLength(server.monitors) && !server.loading)
1481 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1482
1483 /* Call the command. */
1484 redisOpArrayInit(&server.also_propagate);
1485 dirty = server.dirty;
1486 c->cmd->proc(c);
1487 dirty = server.dirty-dirty;
1488 duration = ustime()-start;
1489
1490 /* When EVAL is called loading the AOF we don't want commands called
1491 * from Lua to go into the slowlog or to populate statistics. */
1492 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1493 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1494
1495 /* Log the command into the Slow log if needed, and populate the
1496 * per-command statistics that we show in INFO commandstats. */
1497 if (flags & REDIS_CALL_SLOWLOG)
1498 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1499 if (flags & REDIS_CALL_STATS) {
1500 c->cmd->microseconds += duration;
1501 c->cmd->calls++;
1502 }
1503
1504 /* Propagate the command into the AOF and replication link */
1505 if (flags & REDIS_CALL_PROPAGATE) {
1506 int flags = REDIS_PROPAGATE_NONE;
1507
1508 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1509 flags |= REDIS_PROPAGATE_REPL;
1510 if (dirty)
1511 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1512 if (flags != REDIS_PROPAGATE_NONE)
1513 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1514 }
1515 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1516 * PUSH command. */
1517 if (server.also_propagate.numops) {
1518 int j;
1519 redisOp *rop;
1520
1521 for (j = 0; j < server.also_propagate.numops; j++) {
1522 rop = &server.also_propagate.ops[j];
1523 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1524 }
1525 redisOpArrayFree(&server.also_propagate);
1526 }
1527 server.stat_numcommands++;
1528 }
1529
1530 /* If this function gets called we already read a whole
1531 * command, argments are in the client argv/argc fields.
1532 * processCommand() execute the command or prepare the
1533 * server for a bulk read from the client.
1534 *
1535 * If 1 is returned the client is still alive and valid and
1536 * and other operations can be performed by the caller. Otherwise
1537 * if 0 is returned the client was destroied (i.e. after QUIT). */
1538 int processCommand(redisClient *c) {
1539 /* The QUIT command is handled separately. Normal command procs will
1540 * go through checking for replication and QUIT will cause trouble
1541 * when FORCE_REPLICATION is enabled and would be implemented in
1542 * a regular command proc. */
1543 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1544 addReply(c,shared.ok);
1545 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1546 return REDIS_ERR;
1547 }
1548
1549 /* Now lookup the command and check ASAP about trivial error conditions
1550 * such as wrong arity, bad command name and so forth. */
1551 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1552 if (!c->cmd) {
1553 addReplyErrorFormat(c,"unknown command '%s'",
1554 (char*)c->argv[0]->ptr);
1555 return REDIS_OK;
1556 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1557 (c->argc < -c->cmd->arity)) {
1558 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1559 c->cmd->name);
1560 return REDIS_OK;
1561 }
1562
1563 /* Check if the user is authenticated */
1564 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1565 {
1566 addReplyError(c,"operation not permitted");
1567 return REDIS_OK;
1568 }
1569
1570 /* If cluster is enabled, redirect here */
1571 if (server.cluster_enabled &&
1572 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1573 int hashslot;
1574
1575 if (server.cluster.state != REDIS_CLUSTER_OK) {
1576 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1577 return REDIS_OK;
1578 } else {
1579 int ask;
1580 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1581 if (n == NULL) {
1582 addReplyError(c,"Multi keys request invalid in cluster");
1583 return REDIS_OK;
1584 } else if (n != server.cluster.myself) {
1585 addReplySds(c,sdscatprintf(sdsempty(),
1586 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1587 hashslot,n->ip,n->port));
1588 return REDIS_OK;
1589 }
1590 }
1591 }
1592
1593 /* Handle the maxmemory directive.
1594 *
1595 * First we try to free some memory if possible (if there are volatile
1596 * keys in the dataset). If there are not the only thing we can do
1597 * is returning an error. */
1598 if (server.maxmemory) {
1599 int retval = freeMemoryIfNeeded();
1600 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1601 addReply(c, shared.oomerr);
1602 return REDIS_OK;
1603 }
1604 }
1605
1606 /* Don't accept write commands if there are problems persisting on disk. */
1607 if (server.stop_writes_on_bgsave_err &&
1608 server.saveparamslen > 0
1609 && server.lastbgsave_status == REDIS_ERR &&
1610 c->cmd->flags & REDIS_CMD_WRITE)
1611 {
1612 addReply(c, shared.bgsaveerr);
1613 return REDIS_OK;
1614 }
1615
1616 /* Don't accept wirte commands if this is a read only slave. But
1617 * accept write commands if this is our master. */
1618 if (server.masterhost && server.repl_slave_ro &&
1619 !(c->flags & REDIS_MASTER) &&
1620 c->cmd->flags & REDIS_CMD_WRITE)
1621 {
1622 addReply(c, shared.roslaveerr);
1623 return REDIS_OK;
1624 }
1625
1626 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1627 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1628 &&
1629 c->cmd->proc != subscribeCommand &&
1630 c->cmd->proc != unsubscribeCommand &&
1631 c->cmd->proc != psubscribeCommand &&
1632 c->cmd->proc != punsubscribeCommand) {
1633 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1634 return REDIS_OK;
1635 }
1636
1637 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1638 * we are a slave with a broken link with master. */
1639 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1640 server.repl_serve_stale_data == 0 &&
1641 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1642 {
1643 addReply(c, shared.masterdownerr);
1644 return REDIS_OK;
1645 }
1646
1647 /* Loading DB? Return an error if the command is not INFO */
1648 if (server.loading && c->cmd->proc != infoCommand) {
1649 addReply(c, shared.loadingerr);
1650 return REDIS_OK;
1651 }
1652
1653 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1654 if (server.lua_timedout &&
1655 !(c->cmd->proc == shutdownCommand &&
1656 c->argc == 2 &&
1657 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1658 !(c->cmd->proc == scriptCommand &&
1659 c->argc == 2 &&
1660 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1661 {
1662 addReply(c, shared.slowscripterr);
1663 return REDIS_OK;
1664 }
1665
1666 /* Exec the command */
1667 if (c->flags & REDIS_MULTI &&
1668 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1669 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1670 {
1671 queueMultiCommand(c);
1672 addReply(c,shared.queued);
1673 } else {
1674 call(c,REDIS_CALL_FULL);
1675 }
1676 return REDIS_OK;
1677 }
1678
1679 /*================================== Shutdown =============================== */
1680
1681 int prepareForShutdown(int flags) {
1682 int save = flags & REDIS_SHUTDOWN_SAVE;
1683 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1684
1685 redisLog(REDIS_WARNING,"User requested shutdown...");
1686 /* Kill the saving child if there is a background saving in progress.
1687 We want to avoid race conditions, for instance our saving child may
1688 overwrite the synchronous saving did by SHUTDOWN. */
1689 if (server.rdb_child_pid != -1) {
1690 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1691 kill(server.rdb_child_pid,SIGKILL);
1692 rdbRemoveTempFile(server.rdb_child_pid);
1693 }
1694 if (server.aof_state != REDIS_AOF_OFF) {
1695 /* Kill the AOF saving child as the AOF we already have may be longer
1696 * but contains the full dataset anyway. */
1697 if (server.aof_child_pid != -1) {
1698 redisLog(REDIS_WARNING,
1699 "There is a child rewriting the AOF. Killing it!");
1700 kill(server.aof_child_pid,SIGKILL);
1701 }
1702 /* Append only file: fsync() the AOF and exit */
1703 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1704 aof_fsync(server.aof_fd);
1705 }
1706 if ((server.saveparamslen > 0 && !nosave) || save) {
1707 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1708 /* Snapshotting. Perform a SYNC SAVE and exit */
1709 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1710 /* Ooops.. error saving! The best we can do is to continue
1711 * operating. Note that if there was a background saving process,
1712 * in the next cron() Redis will be notified that the background
1713 * saving aborted, handling special stuff like slaves pending for
1714 * synchronization... */
1715 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1716 return REDIS_ERR;
1717 }
1718 }
1719 if (server.daemonize) {
1720 redisLog(REDIS_NOTICE,"Removing the pid file.");
1721 unlink(server.pidfile);
1722 }
1723 /* Close the listening sockets. Apparently this allows faster restarts. */
1724 if (server.ipfd != -1) close(server.ipfd);
1725 if (server.sofd != -1) close(server.sofd);
1726 if (server.unixsocket) {
1727 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1728 unlink(server.unixsocket); /* don't care if this fails */
1729 }
1730
1731 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1732 return REDIS_OK;
1733 }
1734
1735 /*================================== Commands =============================== */
1736
1737 void authCommand(redisClient *c) {
1738 if (!server.requirepass) {
1739 addReplyError(c,"Client sent AUTH, but no password is set");
1740 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1741 c->authenticated = 1;
1742 addReply(c,shared.ok);
1743 } else {
1744 c->authenticated = 0;
1745 addReplyError(c,"invalid password");
1746 }
1747 }
1748
1749 void pingCommand(redisClient *c) {
1750 addReply(c,shared.pong);
1751 }
1752
1753 void echoCommand(redisClient *c) {
1754 addReplyBulk(c,c->argv[1]);
1755 }
1756
1757 void timeCommand(redisClient *c) {
1758 struct timeval tv;
1759
1760 /* gettimeofday() can only fail if &tv is a bad addresss so we
1761 * don't check for errors. */
1762 gettimeofday(&tv,NULL);
1763 addReplyMultiBulkLen(c,2);
1764 addReplyBulkLongLong(c,tv.tv_sec);
1765 addReplyBulkLongLong(c,tv.tv_usec);
1766 }
1767
1768 /* Convert an amount of bytes into a human readable string in the form
1769 * of 100B, 2G, 100M, 4K, and so forth. */
1770 void bytesToHuman(char *s, unsigned long long n) {
1771 double d;
1772
1773 if (n < 1024) {
1774 /* Bytes */
1775 sprintf(s,"%lluB",n);
1776 return;
1777 } else if (n < (1024*1024)) {
1778 d = (double)n/(1024);
1779 sprintf(s,"%.2fK",d);
1780 } else if (n < (1024LL*1024*1024)) {
1781 d = (double)n/(1024*1024);
1782 sprintf(s,"%.2fM",d);
1783 } else if (n < (1024LL*1024*1024*1024)) {
1784 d = (double)n/(1024LL*1024*1024);
1785 sprintf(s,"%.2fG",d);
1786 }
1787 }
1788
1789 /* Create the string returned by the INFO command. This is decoupled
1790 * by the INFO command itself as we need to report the same information
1791 * on memory corruption problems. */
1792 sds genRedisInfoString(char *section) {
1793 sds info = sdsempty();
1794 time_t uptime = server.unixtime-server.stat_starttime;
1795 int j, numcommands;
1796 struct rusage self_ru, c_ru;
1797 unsigned long lol, bib;
1798 int allsections = 0, defsections = 0;
1799 int sections = 0;
1800
1801 if (section) {
1802 allsections = strcasecmp(section,"all") == 0;
1803 defsections = strcasecmp(section,"default") == 0;
1804 }
1805
1806 getrusage(RUSAGE_SELF, &self_ru);
1807 getrusage(RUSAGE_CHILDREN, &c_ru);
1808 getClientsMaxBuffers(&lol,&bib);
1809
1810 /* Server */
1811 if (allsections || defsections || !strcasecmp(section,"server")) {
1812 struct utsname name;
1813
1814 if (sections++) info = sdscat(info,"\r\n");
1815 uname(&name);
1816 info = sdscatprintf(info,
1817 "# Server\r\n"
1818 "redis_version:%s\r\n"
1819 "redis_git_sha1:%s\r\n"
1820 "redis_git_dirty:%d\r\n"
1821 "os:%s %s %s\r\n"
1822 "arch_bits:%d\r\n"
1823 "multiplexing_api:%s\r\n"
1824 "gcc_version:%d.%d.%d\r\n"
1825 "process_id:%ld\r\n"
1826 "run_id:%s\r\n"
1827 "tcp_port:%d\r\n"
1828 "uptime_in_seconds:%ld\r\n"
1829 "uptime_in_days:%ld\r\n"
1830 "lru_clock:%ld\r\n",
1831 REDIS_VERSION,
1832 redisGitSHA1(),
1833 strtol(redisGitDirty(),NULL,10) > 0,
1834 name.sysname, name.release, name.machine,
1835 server.arch_bits,
1836 aeGetApiName(),
1837 #ifdef __GNUC__
1838 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1839 #else
1840 0,0,0,
1841 #endif
1842 (long) getpid(),
1843 server.runid,
1844 server.port,
1845 uptime,
1846 uptime/(3600*24),
1847 (unsigned long) server.lruclock);
1848 }
1849
1850 /* Clients */
1851 if (allsections || defsections || !strcasecmp(section,"clients")) {
1852 if (sections++) info = sdscat(info,"\r\n");
1853 info = sdscatprintf(info,
1854 "# Clients\r\n"
1855 "connected_clients:%lu\r\n"
1856 "client_longest_output_list:%lu\r\n"
1857 "client_biggest_input_buf:%lu\r\n"
1858 "blocked_clients:%d\r\n",
1859 listLength(server.clients)-listLength(server.slaves),
1860 lol, bib,
1861 server.bpop_blocked_clients);
1862 }
1863
1864 /* Memory */
1865 if (allsections || defsections || !strcasecmp(section,"memory")) {
1866 char hmem[64];
1867 char peak_hmem[64];
1868
1869 bytesToHuman(hmem,zmalloc_used_memory());
1870 bytesToHuman(peak_hmem,server.stat_peak_memory);
1871 if (sections++) info = sdscat(info,"\r\n");
1872 info = sdscatprintf(info,
1873 "# Memory\r\n"
1874 "used_memory:%zu\r\n"
1875 "used_memory_human:%s\r\n"
1876 "used_memory_rss:%zu\r\n"
1877 "used_memory_peak:%zu\r\n"
1878 "used_memory_peak_human:%s\r\n"
1879 "used_memory_lua:%lld\r\n"
1880 "mem_fragmentation_ratio:%.2f\r\n"
1881 "mem_allocator:%s\r\n",
1882 zmalloc_used_memory(),
1883 hmem,
1884 zmalloc_get_rss(),
1885 server.stat_peak_memory,
1886 peak_hmem,
1887 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1888 zmalloc_get_fragmentation_ratio(),
1889 ZMALLOC_LIB
1890 );
1891 }
1892
1893 /* Persistence */
1894 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1895 if (sections++) info = sdscat(info,"\r\n");
1896 info = sdscatprintf(info,
1897 "# Persistence\r\n"
1898 "loading:%d\r\n"
1899 "aof_enabled:%d\r\n"
1900 "changes_since_last_save:%lld\r\n"
1901 "bgsave_in_progress:%d\r\n"
1902 "last_save_time:%ld\r\n"
1903 "last_bgsave_status:%s\r\n"
1904 "bgrewriteaof_in_progress:%d\r\n"
1905 "bgrewriteaof_scheduled:%d\r\n",
1906 server.loading,
1907 server.aof_state != REDIS_AOF_OFF,
1908 server.dirty,
1909 server.rdb_child_pid != -1,
1910 server.lastsave,
1911 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1912 server.aof_child_pid != -1,
1913 server.aof_rewrite_scheduled);
1914
1915 if (server.aof_state != REDIS_AOF_OFF) {
1916 info = sdscatprintf(info,
1917 "aof_current_size:%lld\r\n"
1918 "aof_base_size:%lld\r\n"
1919 "aof_pending_rewrite:%d\r\n"
1920 "aof_buffer_length:%zu\r\n"
1921 "aof_pending_bio_fsync:%llu\r\n"
1922 "aof_delayed_fsync:%lu\r\n",
1923 (long long) server.aof_current_size,
1924 (long long) server.aof_rewrite_base_size,
1925 server.aof_rewrite_scheduled,
1926 sdslen(server.aof_buf),
1927 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1928 server.aof_delayed_fsync);
1929 }
1930
1931 if (server.loading) {
1932 double perc;
1933 time_t eta, elapsed;
1934 off_t remaining_bytes = server.loading_total_bytes-
1935 server.loading_loaded_bytes;
1936
1937 perc = ((double)server.loading_loaded_bytes /
1938 server.loading_total_bytes) * 100;
1939
1940 elapsed = server.unixtime-server.loading_start_time;
1941 if (elapsed == 0) {
1942 eta = 1; /* A fake 1 second figure if we don't have
1943 enough info */
1944 } else {
1945 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1946 }
1947
1948 info = sdscatprintf(info,
1949 "loading_start_time:%ld\r\n"
1950 "loading_total_bytes:%llu\r\n"
1951 "loading_loaded_bytes:%llu\r\n"
1952 "loading_loaded_perc:%.2f\r\n"
1953 "loading_eta_seconds:%ld\r\n"
1954 ,(unsigned long) server.loading_start_time,
1955 (unsigned long long) server.loading_total_bytes,
1956 (unsigned long long) server.loading_loaded_bytes,
1957 perc,
1958 eta
1959 );
1960 }
1961 }
1962
1963 /* Stats */
1964 if (allsections || defsections || !strcasecmp(section,"stats")) {
1965 if (sections++) info = sdscat(info,"\r\n");
1966 info = sdscatprintf(info,
1967 "# Stats\r\n"
1968 "total_connections_received:%lld\r\n"
1969 "total_commands_processed:%lld\r\n"
1970 "instantaneous_ops_per_sec:%lld\r\n"
1971 "rejected_connections:%lld\r\n"
1972 "expired_keys:%lld\r\n"
1973 "evicted_keys:%lld\r\n"
1974 "keyspace_hits:%lld\r\n"
1975 "keyspace_misses:%lld\r\n"
1976 "pubsub_channels:%ld\r\n"
1977 "pubsub_patterns:%lu\r\n"
1978 "latest_fork_usec:%lld\r\n",
1979 server.stat_numconnections,
1980 server.stat_numcommands,
1981 getOperationsPerSecond(),
1982 server.stat_rejected_conn,
1983 server.stat_expiredkeys,
1984 server.stat_evictedkeys,
1985 server.stat_keyspace_hits,
1986 server.stat_keyspace_misses,
1987 dictSize(server.pubsub_channels),
1988 listLength(server.pubsub_patterns),
1989 server.stat_fork_time);
1990 }
1991
1992 /* Replication */
1993 if (allsections || defsections || !strcasecmp(section,"replication")) {
1994 if (sections++) info = sdscat(info,"\r\n");
1995 info = sdscatprintf(info,
1996 "# Replication\r\n"
1997 "role:%s\r\n",
1998 server.masterhost == NULL ? "master" : "slave");
1999 if (server.masterhost) {
2000 info = sdscatprintf(info,
2001 "master_host:%s\r\n"
2002 "master_port:%d\r\n"
2003 "master_link_status:%s\r\n"
2004 "master_last_io_seconds_ago:%d\r\n"
2005 "master_sync_in_progress:%d\r\n"
2006 ,server.masterhost,
2007 server.masterport,
2008 (server.repl_state == REDIS_REPL_CONNECTED) ?
2009 "up" : "down",
2010 server.master ?
2011 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
2012 server.repl_state == REDIS_REPL_TRANSFER
2013 );
2014
2015 if (server.repl_state == REDIS_REPL_TRANSFER) {
2016 info = sdscatprintf(info,
2017 "master_sync_left_bytes:%ld\r\n"
2018 "master_sync_last_io_seconds_ago:%d\r\n"
2019 ,(long)server.repl_transfer_left,
2020 (int)(server.unixtime-server.repl_transfer_lastio)
2021 );
2022 }
2023
2024 if (server.repl_state != REDIS_REPL_CONNECTED) {
2025 info = sdscatprintf(info,
2026 "master_link_down_since_seconds:%ld\r\n",
2027 (long)server.unixtime-server.repl_down_since);
2028 }
2029 }
2030 info = sdscatprintf(info,
2031 "connected_slaves:%lu\r\n",
2032 listLength(server.slaves));
2033 if (listLength(server.slaves)) {
2034 int slaveid = 0;
2035 listNode *ln;
2036 listIter li;
2037
2038 listRewind(server.slaves,&li);
2039 while((ln = listNext(&li))) {
2040 redisClient *slave = listNodeValue(ln);
2041 char *state = NULL;
2042 char ip[32];
2043 int port;
2044
2045 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
2046 switch(slave->replstate) {
2047 case REDIS_REPL_WAIT_BGSAVE_START:
2048 case REDIS_REPL_WAIT_BGSAVE_END:
2049 state = "wait_bgsave";
2050 break;
2051 case REDIS_REPL_SEND_BULK:
2052 state = "send_bulk";
2053 break;
2054 case REDIS_REPL_ONLINE:
2055 state = "online";
2056 break;
2057 }
2058 if (state == NULL) continue;
2059 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2060 slaveid,ip,port,state);
2061 slaveid++;
2062 }
2063 }
2064 }
2065
2066 /* CPU */
2067 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2068 if (sections++) info = sdscat(info,"\r\n");
2069 info = sdscatprintf(info,
2070 "# CPU\r\n"
2071 "used_cpu_sys:%.2f\r\n"
2072 "used_cpu_user:%.2f\r\n"
2073 "used_cpu_sys_children:%.2f\r\n"
2074 "used_cpu_user_children:%.2f\r\n",
2075 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2076 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2077 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2078 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2079 }
2080
2081 /* cmdtime */
2082 if (allsections || !strcasecmp(section,"commandstats")) {
2083 if (sections++) info = sdscat(info,"\r\n");
2084 info = sdscatprintf(info, "# Commandstats\r\n");
2085 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2086 for (j = 0; j < numcommands; j++) {
2087 struct redisCommand *c = redisCommandTable+j;
2088
2089 if (!c->calls) continue;
2090 info = sdscatprintf(info,
2091 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2092 c->name, c->calls, c->microseconds,
2093 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2094 }
2095 }
2096
2097 /* Cluster */
2098 if (allsections || defsections || !strcasecmp(section,"cluster")) {
2099 if (sections++) info = sdscat(info,"\r\n");
2100 info = sdscatprintf(info,
2101 "# Cluster\r\n"
2102 "cluster_enabled:%d\r\n",
2103 server.cluster_enabled);
2104 }
2105
2106 /* Key space */
2107 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2108 if (sections++) info = sdscat(info,"\r\n");
2109 info = sdscatprintf(info, "# Keyspace\r\n");
2110 for (j = 0; j < server.dbnum; j++) {
2111 long long keys, vkeys;
2112
2113 keys = dictSize(server.db[j].dict);
2114 vkeys = dictSize(server.db[j].expires);
2115 if (keys || vkeys) {
2116 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2117 j, keys, vkeys);
2118 }
2119 }
2120 }
2121 return info;
2122 }
2123
2124 void infoCommand(redisClient *c) {
2125 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2126
2127 if (c->argc > 2) {
2128 addReply(c,shared.syntaxerr);
2129 return;
2130 }
2131 sds info = genRedisInfoString(section);
2132 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2133 (unsigned long)sdslen(info)));
2134 addReplySds(c,info);
2135 addReply(c,shared.crlf);
2136 }
2137
2138 void monitorCommand(redisClient *c) {
2139 /* ignore MONITOR if aleady slave or in monitor mode */
2140 if (c->flags & REDIS_SLAVE) return;
2141
2142 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2143 c->slaveseldb = 0;
2144 listAddNodeTail(server.monitors,c);
2145 addReply(c,shared.ok);
2146 }
2147
2148 /* ============================ Maxmemory directive ======================== */
2149
2150 /* This function gets called when 'maxmemory' is set on the config file to limit
2151 * the max memory used by the server, before processing a command.
2152 *
2153 * The goal of the function is to free enough memory to keep Redis under the
2154 * configured memory limit.
2155 *
2156 * The function starts calculating how many bytes should be freed to keep
2157 * Redis under the limit, and enters a loop selecting the best keys to
2158 * evict accordingly to the configured policy.
2159 *
2160 * If all the bytes needed to return back under the limit were freed the
2161 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2162 * should block the execution of commands that will result in more memory
2163 * used by the server.
2164 */
2165 int freeMemoryIfNeeded(void) {
2166 size_t mem_used, mem_tofree, mem_freed;
2167 int slaves = listLength(server.slaves);
2168
2169 /* Remove the size of slaves output buffers and AOF buffer from the
2170 * count of used memory. */
2171 mem_used = zmalloc_used_memory();
2172 if (slaves) {
2173 listIter li;
2174 listNode *ln;
2175
2176 listRewind(server.slaves,&li);
2177 while((ln = listNext(&li))) {
2178 redisClient *slave = listNodeValue(ln);
2179 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2180 if (obuf_bytes > mem_used)
2181 mem_used = 0;
2182 else
2183 mem_used -= obuf_bytes;
2184 }
2185 }
2186 if (server.aof_state != REDIS_AOF_OFF) {
2187 mem_used -= sdslen(server.aof_buf);
2188 mem_used -= sdslen(server.aof_rewrite_buf);
2189 }
2190
2191 /* Check if we are over the memory limit. */
2192 if (mem_used <= server.maxmemory) return REDIS_OK;
2193
2194 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2195 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2196
2197 /* Compute how much memory we need to free. */
2198 mem_tofree = mem_used - server.maxmemory;
2199 mem_freed = 0;
2200 while (mem_freed < mem_tofree) {
2201 int j, k, keys_freed = 0;
2202
2203 for (j = 0; j < server.dbnum; j++) {
2204 long bestval = 0; /* just to prevent warning */
2205 sds bestkey = NULL;
2206 struct dictEntry *de;
2207 redisDb *db = server.db+j;
2208 dict *dict;
2209
2210 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2211 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2212 {
2213 dict = server.db[j].dict;
2214 } else {
2215 dict = server.db[j].expires;
2216 }
2217 if (dictSize(dict) == 0) continue;
2218
2219 /* volatile-random and allkeys-random policy */
2220 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2221 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2222 {
2223 de = dictGetRandomKey(dict);
2224 bestkey = dictGetKey(de);
2225 }
2226
2227 /* volatile-lru and allkeys-lru policy */
2228 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2229 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2230 {
2231 for (k = 0; k < server.maxmemory_samples; k++) {
2232 sds thiskey;
2233 long thisval;
2234 robj *o;
2235
2236 de = dictGetRandomKey(dict);
2237 thiskey = dictGetKey(de);
2238 /* When policy is volatile-lru we need an additonal lookup
2239 * to locate the real key, as dict is set to db->expires. */
2240 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2241 de = dictFind(db->dict, thiskey);
2242 o = dictGetVal(de);
2243 thisval = estimateObjectIdleTime(o);
2244
2245 /* Higher idle time is better candidate for deletion */
2246 if (bestkey == NULL || thisval > bestval) {
2247 bestkey = thiskey;
2248 bestval = thisval;
2249 }
2250 }
2251 }
2252
2253 /* volatile-ttl */
2254 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2255 for (k = 0; k < server.maxmemory_samples; k++) {
2256 sds thiskey;
2257 long thisval;
2258
2259 de = dictGetRandomKey(dict);
2260 thiskey = dictGetKey(de);
2261 thisval = (long) dictGetVal(de);
2262
2263 /* Expire sooner (minor expire unix timestamp) is better
2264 * candidate for deletion */
2265 if (bestkey == NULL || thisval < bestval) {
2266 bestkey = thiskey;
2267 bestval = thisval;
2268 }
2269 }
2270 }
2271
2272 /* Finally remove the selected key. */
2273 if (bestkey) {
2274 long long delta;
2275
2276 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2277 propagateExpire(db,keyobj);
2278 /* We compute the amount of memory freed by dbDelete() alone.
2279 * It is possible that actually the memory needed to propagate
2280 * the DEL in AOF and replication link is greater than the one
2281 * we are freeing removing the key, but we can't account for
2282 * that otherwise we would never exit the loop.
2283 *
2284 * AOF and Output buffer memory will be freed eventually so
2285 * we only care about memory used by the key space. */
2286 delta = (long long) zmalloc_used_memory();
2287 dbDelete(db,keyobj);
2288 delta -= (long long) zmalloc_used_memory();
2289 mem_freed += delta;
2290 server.stat_evictedkeys++;
2291 decrRefCount(keyobj);
2292 keys_freed++;
2293
2294 /* When the memory to free starts to be big enough, we may
2295 * start spending so much time here that is impossible to
2296 * deliver data to the slaves fast enough, so we force the
2297 * transmission here inside the loop. */
2298 if (slaves) flushSlavesOutputBuffers();
2299 }
2300 }
2301 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2302 }
2303 return REDIS_OK;
2304 }
2305
2306 /* =================================== Main! ================================ */
2307
2308 #ifdef __linux__
2309 int linuxOvercommitMemoryValue(void) {
2310 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2311 char buf[64];
2312
2313 if (!fp) return -1;
2314 if (fgets(buf,64,fp) == NULL) {
2315 fclose(fp);
2316 return -1;
2317 }
2318 fclose(fp);
2319
2320 return atoi(buf);
2321 }
2322
2323 void linuxOvercommitMemoryWarning(void) {
2324 if (linuxOvercommitMemoryValue() == 0) {
2325 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.");
2326 }
2327 }
2328 #endif /* __linux__ */
2329
2330 void createPidFile(void) {
2331 /* Try to write the pid file in a best-effort way. */
2332 FILE *fp = fopen(server.pidfile,"w");
2333 if (fp) {
2334 fprintf(fp,"%d\n",(int)getpid());
2335 fclose(fp);
2336 }
2337 }
2338
2339 void daemonize(void) {
2340 int fd;
2341
2342 if (fork() != 0) exit(0); /* parent exits */
2343 setsid(); /* create a new session */
2344
2345 /* Every output goes to /dev/null. If Redis is daemonized but
2346 * the 'logfile' is set to 'stdout' in the configuration file
2347 * it will not log at all. */
2348 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2349 dup2(fd, STDIN_FILENO);
2350 dup2(fd, STDOUT_FILENO);
2351 dup2(fd, STDERR_FILENO);
2352 if (fd > STDERR_FILENO) close(fd);
2353 }
2354 }
2355
2356 void version() {
2357 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2358 REDIS_VERSION,
2359 redisGitSHA1(),
2360 atoi(redisGitDirty()) > 0,
2361 ZMALLOC_LIB,
2362 sizeof(long) == 4 ? 32 : 64);
2363 exit(0);
2364 }
2365
2366 void usage() {
2367 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2368 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2369 fprintf(stderr," ./redis-server -v or --version\n");
2370 fprintf(stderr," ./redis-server -h or --help\n");
2371 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2372 fprintf(stderr,"Examples:\n");
2373 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2374 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2375 fprintf(stderr," ./redis-server --port 7777\n");
2376 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2377 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2378 exit(1);
2379 }
2380
2381 void redisAsciiArt(void) {
2382 #include "asciilogo.h"
2383 char *buf = zmalloc(1024*16);
2384
2385 snprintf(buf,1024*16,ascii_logo,
2386 REDIS_VERSION,
2387 redisGitSHA1(),
2388 strtol(redisGitDirty(),NULL,10) > 0,
2389 (sizeof(long) == 8) ? "64" : "32",
2390 server.cluster_enabled ? "cluster" : "stand alone",
2391 server.port,
2392 (long) getpid()
2393 );
2394 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2395 zfree(buf);
2396 }
2397
2398 static void sigtermHandler(int sig) {
2399 REDIS_NOTUSED(sig);
2400
2401 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2402 server.shutdown_asap = 1;
2403 }
2404
2405 void setupSignalHandlers(void) {
2406 struct sigaction act;
2407
2408 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2409 * Otherwise, sa_handler is used. */
2410 sigemptyset(&act.sa_mask);
2411 act.sa_flags = 0;
2412 act.sa_handler = sigtermHandler;
2413 sigaction(SIGTERM, &act, NULL);
2414
2415 #ifdef HAVE_BACKTRACE
2416 sigemptyset(&act.sa_mask);
2417 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2418 act.sa_sigaction = sigsegvHandler;
2419 sigaction(SIGSEGV, &act, NULL);
2420 sigaction(SIGBUS, &act, NULL);
2421 sigaction(SIGFPE, &act, NULL);
2422 sigaction(SIGILL, &act, NULL);
2423 #endif
2424 return;
2425 }
2426
2427 void memtest(size_t megabytes, int passes);
2428
2429 int main(int argc, char **argv) {
2430 long long start;
2431 struct timeval tv;
2432
2433 /* We need to initialize our libraries, and the server configuration. */
2434 zmalloc_enable_thread_safeness();
2435 srand(time(NULL)^getpid());
2436 gettimeofday(&tv,NULL);
2437 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2438 initServerConfig();
2439
2440 if (argc >= 2) {
2441 int j = 1; /* First option to parse in argv[] */
2442 sds options = sdsempty();
2443 char *configfile = NULL;
2444
2445 /* Handle special options --help and --version */
2446 if (strcmp(argv[1], "-v") == 0 ||
2447 strcmp(argv[1], "--version") == 0) version();
2448 if (strcmp(argv[1], "--help") == 0 ||
2449 strcmp(argv[1], "-h") == 0) usage();
2450 if (strcmp(argv[1], "--test-memory") == 0) {
2451 if (argc == 3) {
2452 memtest(atoi(argv[2]),50);
2453 exit(0);
2454 } else {
2455 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2456 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2457 exit(1);
2458 }
2459 }
2460
2461 /* First argument is the config file name? */
2462 if (argv[j][0] != '-' || argv[j][1] != '-')
2463 configfile = argv[j++];
2464 /* All the other options are parsed and conceptually appended to the
2465 * configuration file. For instance --port 6380 will generate the
2466 * string "port 6380\n" to be parsed after the actual file name
2467 * is parsed, if any. */
2468 while(j != argc) {
2469 if (argv[j][0] == '-' && argv[j][1] == '-') {
2470 /* Option name */
2471 if (sdslen(options)) options = sdscat(options,"\n");
2472 options = sdscat(options,argv[j]+2);
2473 options = sdscat(options," ");
2474 } else {
2475 /* Option argument */
2476 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2477 options = sdscat(options," ");
2478 }
2479 j++;
2480 }
2481 resetServerSaveParams();
2482 loadServerConfig(configfile,options);
2483 sdsfree(options);
2484 } else {
2485 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'");
2486 }
2487 if (server.daemonize) daemonize();
2488 initServer();
2489 if (server.daemonize) createPidFile();
2490 redisAsciiArt();
2491 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2492 #ifdef __linux__
2493 linuxOvercommitMemoryWarning();
2494 #endif
2495 start = ustime();
2496 if (server.aof_state == REDIS_AOF_ON) {
2497 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2498 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2499 } else {
2500 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2501 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2502 (float)(ustime()-start)/1000000);
2503 } else if (errno != ENOENT) {
2504 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2505 exit(1);
2506 }
2507 }
2508 if (server.ipfd > 0)
2509 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2510 if (server.sofd > 0)
2511 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2512 aeSetBeforeSleepProc(server.el,beforeSleep);
2513 aeMain(server.el);
2514 aeDeleteEventLoop(server.el);
2515 return 0;
2516 }
2517
2518 /* The End */