]> git.saurik.com Git - redis.git/blob - src/redis.c
Support EXISTS command queries against MDB archive.
[redis.git] / src / redis.c
1 /*
2 * Copyright (c) 2009-2012, Salvatore Sanfilippo <antirez at gmail dot com>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are met:
7 *
8 * * Redistributions of source code must retain the above copyright notice,
9 * this list of conditions and the following disclaimer.
10 * * Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * * Neither the name of Redis nor the names of its contributors may be used
14 * to endorse or promote products derived from this software without
15 * specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27 * POSSIBILITY OF SUCH DAMAGE.
28 */
29
30 #include "redis.h"
31 #include "slowlog.h"
32 #include "bio.h"
33
34 #include <time.h>
35 #include <signal.h>
36 #include <sys/wait.h>
37 #include <errno.h>
38 #include <assert.h>
39 #include <ctype.h>
40 #include <stdarg.h>
41 #include <arpa/inet.h>
42 #include <sys/stat.h>
43 #include <fcntl.h>
44 #include <sys/time.h>
45 #include <sys/resource.h>
46 #include <sys/uio.h>
47 #include <limits.h>
48 #include <float.h>
49 #include <math.h>
50 #include <sys/resource.h>
51 #include <sys/utsname.h>
52 #include <lmdb.h>
53
54 /* Our shared "common" objects */
55
56 struct sharedObjectsStruct shared;
57
58 /* Global vars that are actually used as constants. The following double
59 * values are used for double on-disk serialization, and are initialized
60 * at runtime to avoid strange compiler optimizations. */
61
62 double R_Zero, R_PosInf, R_NegInf, R_Nan;
63
64 /*================================= Globals ================================= */
65
66 /* Global vars */
67 struct redisServer server; /* server global state */
68 struct redisCommand *commandTable;
69
70 /* Our command table.
71 *
72 * Every entry is composed of the following fields:
73 *
74 * name: a string representing the command name.
75 * function: pointer to the C function implementing the command.
76 * arity: number of arguments, it is possible to use -N to say >= N
77 * sflags: command flags as string. See below for a table of flags.
78 * flags: flags as bitmask. Computed by Redis using the 'sflags' field.
79 * get_keys_proc: an optional function to get key arguments from a command.
80 * This is only used when the following three fields are not
81 * enough to specify what arguments are keys.
82 * first_key_index: first argument that is a key
83 * last_key_index: last argument that is a key
84 * key_step: step to get all the keys from first to last argument. For instance
85 * in MSET the step is two since arguments are key,val,key,val,...
86 * microseconds: microseconds of total execution time for this command.
87 * calls: total number of calls of this command.
88 *
89 * The flags, microseconds and calls fields are computed by Redis and should
90 * always be set to zero.
91 *
92 * Command flags are expressed using strings where every character represents
93 * a flag. Later the populateCommandTable() function will take care of
94 * populating the real 'flags' field using this characters.
95 *
96 * This is the meaning of the flags:
97 *
98 * w: write command (may modify the key space).
99 * r: read command (will never modify the key space).
100 * m: may increase memory usage once called. Don't allow if out of memory.
101 * a: admin command, like SAVE or SHUTDOWN.
102 * p: Pub/Sub related command.
103 * f: force replication of this command, regarless of server.dirty.
104 * s: command not allowed in scripts.
105 * R: random command. Command is not deterministic, that is, the same command
106 * with the same arguments, with the same key space, may have different
107 * results. For instance SPOP and RANDOMKEY are two random commands.
108 * S: Sort command output array if called from script, so that the output
109 * is deterministic.
110 * l: Allow command while loading the database.
111 * t: Allow command while a slave has stale data but is not allowed to
112 * server this data. Normally no command is accepted in this condition
113 * but just a few.
114 * M: Do not automatically propagate the command on MONITOR.
115 */
116 struct redisCommand redisCommandTable[] = {
117 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
118 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
119 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
120 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
121 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
122 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
123 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
124 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
125 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
126 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
127 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
128 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
129 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
130 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
131 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
132 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
133 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
134 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
135 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
136 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
137 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
138 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
139 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
140 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
141 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
142 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
143 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
144 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
145 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
146 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
147 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
148 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
149 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
150 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
151 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
152 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
153 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
154 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
155 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
156 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
157 {"srandmember",srandmemberCommand,-2,"rR",0,NULL,1,1,1,0,0},
158 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
159 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
160 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
161 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
162 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
163 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
164 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
165 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
166 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
167 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
168 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
169 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
170 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
171 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
172 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
173 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
174 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
175 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
176 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
177 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
178 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
179 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
180 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
181 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
182 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
183 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
184 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
185 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
186 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
187 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
188 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
189 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
190 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
191 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
192 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
193 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
194 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
195 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
196 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
197 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
198 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
199 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
200 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
201 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
202 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
203 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
204 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
205 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
206 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
207 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
208 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
209 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
210 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
211 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
212 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
213 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
214 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
215 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
216 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
217 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
218 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
219 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
220 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
221 {"exec",execCommand,1,"sM",0,NULL,0,0,0,0,0},
222 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
223 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
224 {"replconf",replconfCommand,-1,"ars",0,NULL,0,0,0,0,0},
225 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
226 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
227 {"sort",sortCommand,-2,"wm",0,NULL,1,1,1,0,0},
228 {"info",infoCommand,-1,"rlt",0,NULL,0,0,0,0,0},
229 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
230 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
231 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
232 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
233 {"slaveof",slaveofCommand,3,"ast",0,NULL,0,0,0,0,0},
234 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
235 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
236 {"subscribe",subscribeCommand,-2,"rpslt",0,NULL,0,0,0,0,0},
237 {"unsubscribe",unsubscribeCommand,-1,"rpslt",0,NULL,0,0,0,0,0},
238 {"psubscribe",psubscribeCommand,-2,"rpslt",0,NULL,0,0,0,0,0},
239 {"punsubscribe",punsubscribeCommand,-1,"rpslt",0,NULL,0,0,0,0,0},
240 {"publish",publishCommand,3,"pflt",0,NULL,0,0,0,0,0},
241 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
242 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
243 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
244 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
245 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
246 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
247 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
248 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
249 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
250 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
251 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
252 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0},
253 {"bitop",bitopCommand,-4,"wm",0,NULL,2,-1,1,0,0},
254 {"bitcount",bitcountCommand,-2,"r",0,NULL,1,1,1,0,0}
255 };
256
257 /*============================ Utility functions ============================ */
258
259 /* Low level logging. To use only for very big messages, otherwise
260 * redisLog() is to prefer. */
261 void redisLogRaw(int level, const char *msg) {
262 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
263 const char *c = ".-*#";
264 FILE *fp;
265 char buf[64];
266 int rawmode = (level & REDIS_LOG_RAW);
267
268 level &= 0xff; /* clear flags */
269 if (level < server.verbosity) return;
270
271 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
272 if (!fp) return;
273
274 if (rawmode) {
275 fprintf(fp,"%s",msg);
276 } else {
277 int off;
278 struct timeval tv;
279
280 gettimeofday(&tv,NULL);
281 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
282 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
283 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
284 }
285 fflush(fp);
286
287 if (server.logfile) fclose(fp);
288
289 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
290 }
291
292 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
293 * is used across the code. The raw version is only used in order to dump
294 * the INFO output on crash. */
295 void redisLog(int level, const char *fmt, ...) {
296 va_list ap;
297 char msg[REDIS_MAX_LOGMSG_LEN];
298
299 if ((level&0xff) < server.verbosity) return;
300
301 va_start(ap, fmt);
302 vsnprintf(msg, sizeof(msg), fmt, ap);
303 va_end(ap);
304
305 redisLogRaw(level,msg);
306 }
307
308 /* Log a fixed message without printf-alike capabilities, in a way that is
309 * safe to call from a signal handler.
310 *
311 * We actually use this only for signals that are not fatal from the point
312 * of view of Redis. Signals that are going to kill the server anyway and
313 * where we need printf-alike features are served by redisLog(). */
314 void redisLogFromHandler(int level, const char *msg) {
315 int fd;
316 char buf[64];
317
318 if ((level&0xff) < server.verbosity ||
319 (server.logfile == NULL && server.daemonize)) return;
320 fd = server.logfile ?
321 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
322 STDOUT_FILENO;
323 if (fd == -1) return;
324 ll2string(buf,sizeof(buf),getpid());
325 if (write(fd,"[",1) == -1) goto err;
326 if (write(fd,buf,strlen(buf)) == -1) goto err;
327 if (write(fd," | signal handler] (",20) == -1) goto err;
328 ll2string(buf,sizeof(buf),time(NULL));
329 if (write(fd,buf,strlen(buf)) == -1) goto err;
330 if (write(fd,") ",2) == -1) goto err;
331 if (write(fd,msg,strlen(msg)) == -1) goto err;
332 if (write(fd,"\n",1) == -1) goto err;
333 err:
334 if (server.logfile) close(fd);
335 }
336
337 /* Return the UNIX time in microseconds */
338 long long ustime(void) {
339 struct timeval tv;
340 long long ust;
341
342 gettimeofday(&tv, NULL);
343 ust = ((long long)tv.tv_sec)*1000000;
344 ust += tv.tv_usec;
345 return ust;
346 }
347
348 /* Return the UNIX time in milliseconds */
349 long long mstime(void) {
350 return ustime()/1000;
351 }
352
353 /* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
354 * exit(), because the latter may interact with the same file objects used by
355 * the parent process. However if we are testing the coverage normal exit() is
356 * used in order to obtain the right coverage information. */
357 void exitFromChild(int retcode) {
358 #ifdef COVERAGE_TEST
359 exit(retcode);
360 #else
361 _exit(retcode);
362 #endif
363 }
364
365 /*====================== Hash table type implementation ==================== */
366
367 /* This is an hash table type that uses the SDS dynamic strings libary as
368 * keys and radis objects as values (objects can hold SDS strings,
369 * lists, sets). */
370
371 void dictVanillaFree(void *privdata, void *val)
372 {
373 DICT_NOTUSED(privdata);
374 zfree(val);
375 }
376
377 void dictListDestructor(void *privdata, void *val)
378 {
379 DICT_NOTUSED(privdata);
380 listRelease((list*)val);
381 }
382
383 int dictSdsKeyCompare(void *privdata, const void *key1,
384 const void *key2)
385 {
386 int l1,l2;
387 DICT_NOTUSED(privdata);
388
389 l1 = sdslen((sds)key1);
390 l2 = sdslen((sds)key2);
391 if (l1 != l2) return 0;
392 return memcmp(key1, key2, l1) == 0;
393 }
394
395 /* A case insensitive version used for the command lookup table and other
396 * places where case insensitive non binary-safe comparison is needed. */
397 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
398 const void *key2)
399 {
400 DICT_NOTUSED(privdata);
401
402 return strcasecmp(key1, key2) == 0;
403 }
404
405 void dictRedisObjectDestructor(void *privdata, void *val)
406 {
407 DICT_NOTUSED(privdata);
408
409 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
410 decrRefCount(val);
411 }
412
413 void dictSdsDestructor(void *privdata, void *val)
414 {
415 DICT_NOTUSED(privdata);
416
417 sdsfree(val);
418 }
419
420 int dictObjKeyCompare(void *privdata, const void *key1,
421 const void *key2)
422 {
423 const robj *o1 = key1, *o2 = key2;
424 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
425 }
426
427 unsigned int dictObjHash(const void *key) {
428 const robj *o = key;
429 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
430 }
431
432 unsigned int dictSdsHash(const void *key) {
433 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
434 }
435
436 unsigned int dictSdsCaseHash(const void *key) {
437 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
438 }
439
440 int dictEncObjKeyCompare(void *privdata, const void *key1,
441 const void *key2)
442 {
443 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
444 int cmp;
445
446 if (o1->encoding == REDIS_ENCODING_INT &&
447 o2->encoding == REDIS_ENCODING_INT)
448 return o1->ptr == o2->ptr;
449
450 o1 = getDecodedObject(o1);
451 o2 = getDecodedObject(o2);
452 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
453 decrRefCount(o1);
454 decrRefCount(o2);
455 return cmp;
456 }
457
458 unsigned int dictEncObjHash(const void *key) {
459 robj *o = (robj*) key;
460
461 if (o->encoding == REDIS_ENCODING_RAW) {
462 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
463 } else {
464 if (o->encoding == REDIS_ENCODING_INT) {
465 char buf[32];
466 int len;
467
468 len = ll2string(buf,32,(long)o->ptr);
469 return dictGenHashFunction((unsigned char*)buf, len);
470 } else {
471 unsigned int hash;
472
473 o = getDecodedObject(o);
474 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
475 decrRefCount(o);
476 return hash;
477 }
478 }
479 }
480
481 /* Sets type hash table */
482 dictType setDictType = {
483 dictEncObjHash, /* hash function */
484 NULL, /* key dup */
485 NULL, /* val dup */
486 dictEncObjKeyCompare, /* key compare */
487 dictRedisObjectDestructor, /* key destructor */
488 NULL /* val destructor */
489 };
490
491 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
492 dictType zsetDictType = {
493 dictEncObjHash, /* hash function */
494 NULL, /* key dup */
495 NULL, /* val dup */
496 dictEncObjKeyCompare, /* key compare */
497 dictRedisObjectDestructor, /* key destructor */
498 NULL /* val destructor */
499 };
500
501 /* Db->dict, keys are sds strings, vals are Redis objects. */
502 dictType dbDictType = {
503 dictSdsHash, /* hash function */
504 NULL, /* key dup */
505 NULL, /* val dup */
506 dictSdsKeyCompare, /* key compare */
507 dictSdsDestructor, /* key destructor */
508 dictRedisObjectDestructor /* val destructor */
509 };
510
511 /* server.lua_scripts sha (as sds string) -> scripts (as robj) cache. */
512 dictType shaScriptObjectDictType = {
513 dictSdsCaseHash, /* hash function */
514 NULL, /* key dup */
515 NULL, /* val dup */
516 dictSdsKeyCaseCompare, /* key compare */
517 dictSdsDestructor, /* key destructor */
518 dictRedisObjectDestructor /* val destructor */
519 };
520
521 /* Db->expires */
522 dictType keyptrDictType = {
523 dictSdsHash, /* hash function */
524 NULL, /* key dup */
525 NULL, /* val dup */
526 dictSdsKeyCompare, /* key compare */
527 NULL, /* key destructor */
528 NULL /* val destructor */
529 };
530
531 /* Command table. sds string -> command struct pointer. */
532 dictType commandTableDictType = {
533 dictSdsCaseHash, /* hash function */
534 NULL, /* key dup */
535 NULL, /* val dup */
536 dictSdsKeyCaseCompare, /* key compare */
537 dictSdsDestructor, /* key destructor */
538 NULL /* val destructor */
539 };
540
541 /* Hash type hash table (note that small hashes are represented with zimpaps) */
542 dictType hashDictType = {
543 dictEncObjHash, /* hash function */
544 NULL, /* key dup */
545 NULL, /* val dup */
546 dictEncObjKeyCompare, /* key compare */
547 dictRedisObjectDestructor, /* key destructor */
548 dictRedisObjectDestructor /* val destructor */
549 };
550
551 /* Keylist hash table type has unencoded redis objects as keys and
552 * lists as values. It's used for blocking operations (BLPOP) and to
553 * map swapped keys to a list of clients waiting for this keys to be loaded. */
554 dictType keylistDictType = {
555 dictObjHash, /* hash function */
556 NULL, /* key dup */
557 NULL, /* val dup */
558 dictObjKeyCompare, /* key compare */
559 dictRedisObjectDestructor, /* key destructor */
560 dictListDestructor /* val destructor */
561 };
562
563 int htNeedsResize(dict *dict) {
564 long long size, used;
565
566 size = dictSlots(dict);
567 used = dictSize(dict);
568 return (size && used && size > DICT_HT_INITIAL_SIZE &&
569 (used*100/size < REDIS_HT_MINFILL));
570 }
571
572 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
573 * we resize the hash table to save memory */
574 void tryResizeHashTables(void) {
575 int j;
576
577 for (j = 0; j < server.dbnum; j++) {
578 if (htNeedsResize(server.db[j].dict))
579 dictResize(server.db[j].dict);
580 if (htNeedsResize(server.db[j].expires))
581 dictResize(server.db[j].expires);
582 }
583 }
584
585 /* Our hash table implementation performs rehashing incrementally while
586 * we write/read from the hash table. Still if the server is idle, the hash
587 * table will use two tables for a long time. So we try to use 1 millisecond
588 * of CPU time at every serverCron() loop in order to rehash some key. */
589 void incrementallyRehash(void) {
590 int j;
591
592 for (j = 0; j < server.dbnum; j++) {
593 /* Keys dictionary */
594 if (dictIsRehashing(server.db[j].dict)) {
595 dictRehashMilliseconds(server.db[j].dict,1);
596 break; /* already used our millisecond for this loop... */
597 }
598 /* Expires */
599 if (dictIsRehashing(server.db[j].expires)) {
600 dictRehashMilliseconds(server.db[j].expires,1);
601 break; /* already used our millisecond for this loop... */
602 }
603 }
604 }
605
606 /* This function is called once a background process of some kind terminates,
607 * as we want to avoid resizing the hash tables when there is a child in order
608 * to play well with copy-on-write (otherwise when a resize happens lots of
609 * memory pages are copied). The goal of this function is to update the ability
610 * for dict.c to resize the hash tables accordingly to the fact we have o not
611 * running childs. */
612 void updateDictResizePolicy(void) {
613 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
614 dictEnableResize();
615 else
616 dictDisableResize();
617 }
618
619 /* ======================= Cron: called every 100 ms ======================== */
620
621 /* Try to expire a few timed out keys. The algorithm used is adaptive and
622 * will use few CPU cycles if there are few expiring keys, otherwise
623 * it will get more aggressive to avoid that too much memory is used by
624 * keys that can be removed from the keyspace. */
625 void activeExpireCycle(void) {
626 int j, iteration = 0;
627 long long start = ustime(), timelimit;
628
629 /* We can use at max REDIS_EXPIRELOOKUPS_TIME_PERC percentage of CPU time
630 * per iteration. Since this function gets called with a frequency of
631 * REDIS_HZ times per second, the following is the max amount of
632 * microseconds we can spend in this function. */
633 timelimit = 1000000*REDIS_EXPIRELOOKUPS_TIME_PERC/REDIS_HZ/100;
634 if (timelimit <= 0) timelimit = 1;
635
636 for (j = 0; j < server.dbnum; j++) {
637 int expired;
638 redisDb *db = server.db+j;
639
640 /* Continue to expire if at the end of the cycle more than 25%
641 * of the keys were expired. */
642 do {
643 unsigned long num = dictSize(db->expires);
644 unsigned long slots = dictSlots(db->expires);
645 long long now = mstime();
646
647 /* When there are less than 1% filled slots getting random
648 * keys is expensive, so stop here waiting for better times...
649 * The dictionary will be resized asap. */
650 if (num && slots > DICT_HT_INITIAL_SIZE &&
651 (num*100/slots < 1)) break;
652
653 /* The main collection cycle. Sample random keys among keys
654 * with an expire set, checking for expired ones. */
655 expired = 0;
656 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
657 num = REDIS_EXPIRELOOKUPS_PER_CRON;
658 while (num--) {
659 dictEntry *de;
660 long long t;
661
662 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
663 t = dictGetSignedIntegerVal(de);
664 if (now > t) {
665 sds key = dictGetKey(de);
666 robj *keyobj = createStringObject(key,sdslen(key));
667
668 propagateExpire(db,keyobj);
669 dbDelete(db,keyobj);
670 decrRefCount(keyobj);
671 expired++;
672 server.stat_expiredkeys++;
673 }
674 }
675 /* We can't block forever here even if there are many keys to
676 * expire. So after a given amount of milliseconds return to the
677 * caller waiting for the other active expire cycle. */
678 iteration++;
679 if ((iteration & 0xf) == 0 && /* check once every 16 cycles. */
680 (ustime()-start) > timelimit) return;
681 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
682 }
683 }
684
685 void updateLRUClock(void) {
686 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
687 REDIS_LRU_CLOCK_MAX;
688 }
689
690
691 /* Add a sample to the operations per second array of samples. */
692 void trackOperationsPerSecond(void) {
693 long long t = mstime() - server.ops_sec_last_sample_time;
694 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
695 long long ops_sec;
696
697 ops_sec = t > 0 ? (ops*1000/t) : 0;
698
699 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
700 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
701 server.ops_sec_last_sample_time = mstime();
702 server.ops_sec_last_sample_ops = server.stat_numcommands;
703 }
704
705 /* Return the mean of all the samples. */
706 long long getOperationsPerSecond(void) {
707 int j;
708 long long sum = 0;
709
710 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
711 sum += server.ops_sec_samples[j];
712 return sum / REDIS_OPS_SEC_SAMPLES;
713 }
714
715 /* Check for timeouts. Returns non-zero if the client was terminated */
716 int clientsCronHandleTimeout(redisClient *c) {
717 time_t now = server.unixtime;
718
719 if (server.maxidletime &&
720 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
721 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
722 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
723 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
724 listLength(c->pubsub_patterns) == 0 &&
725 (now - c->lastinteraction > server.maxidletime))
726 {
727 redisLog(REDIS_VERBOSE,"Closing idle client");
728 freeClient(c);
729 return 1;
730 } else if (c->flags & REDIS_BLOCKED) {
731 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
732 addReply(c,shared.nullmultibulk);
733 unblockClientWaitingData(c);
734 }
735 }
736 return 0;
737 }
738
739 /* The client query buffer is an sds.c string that can end with a lot of
740 * free space not used, this function reclaims space if needed.
741 *
742 * The funciton always returns 0 as it never terminates the client. */
743 int clientsCronResizeQueryBuffer(redisClient *c) {
744 size_t querybuf_size = sdsAllocSize(c->querybuf);
745 time_t idletime = server.unixtime - c->lastinteraction;
746
747 /* There are two conditions to resize the query buffer:
748 * 1) Query buffer is > BIG_ARG and too big for latest peak.
749 * 2) Client is inactive and the buffer is bigger than 1k. */
750 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
751 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
752 (querybuf_size > 1024 && idletime > 2))
753 {
754 /* Only resize the query buffer if it is actually wasting space. */
755 if (sdsavail(c->querybuf) > 1024) {
756 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
757 }
758 }
759 /* Reset the peak again to capture the peak memory usage in the next
760 * cycle. */
761 c->querybuf_peak = 0;
762 return 0;
763 }
764
765 void clientsCron(void) {
766 /* Make sure to process at least 1/(REDIS_HZ*10) of clients per call.
767 * Since this function is called REDIS_HZ times per second we are sure that
768 * in the worst case we process all the clients in 10 seconds.
769 * In normal conditions (a reasonable number of clients) we process
770 * all the clients in a shorter time. */
771 int numclients = listLength(server.clients);
772 int iterations = numclients/(REDIS_HZ*10);
773
774 if (iterations < 50)
775 iterations = (numclients < 50) ? numclients : 50;
776 while(listLength(server.clients) && iterations--) {
777 redisClient *c;
778 listNode *head;
779
780 /* Rotate the list, take the current head, process.
781 * This way if the client must be removed from the list it's the
782 * first element and we don't incur into O(N) computation. */
783 listRotate(server.clients);
784 head = listFirst(server.clients);
785 c = listNodeValue(head);
786 /* The following functions do different service checks on the client.
787 * The protocol is that they return non-zero if the client was
788 * terminated. */
789 if (clientsCronHandleTimeout(c)) continue;
790 if (clientsCronResizeQueryBuffer(c)) continue;
791 }
792 }
793
794 /* This is our timer interrupt, called REDIS_HZ times per second.
795 * Here is where we do a number of things that need to be done asynchronously.
796 * For instance:
797 *
798 * - Active expired keys collection (it is also performed in a lazy way on
799 * lookup).
800 * - Software watchdong.
801 * - Update some statistic.
802 * - Incremental rehashing of the DBs hash tables.
803 * - Triggering BGSAVE / AOF rewrite, and handling of terminated children.
804 * - Clients timeout of differnet kinds.
805 * - Replication reconnection.
806 * - Many more...
807 *
808 * Everything directly called here will be called REDIS_HZ times per second,
809 * so in order to throttle execution of things we want to do less frequently
810 * a macro is used: run_with_period(milliseconds) { .... }
811 */
812
813 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
814 int j;
815 REDIS_NOTUSED(eventLoop);
816 REDIS_NOTUSED(id);
817 REDIS_NOTUSED(clientData);
818
819 /* Software watchdog: deliver the SIGALRM that will reach the signal
820 * handler if we don't return here fast enough. */
821 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
822
823 /* We take a cached value of the unix time in the global state because
824 * with virtual memory and aging there is to store the current time
825 * in objects at every object access, and accuracy is not needed.
826 * To access a global var is faster than calling time(NULL) */
827 server.unixtime = time(NULL);
828
829 run_with_period(100) trackOperationsPerSecond();
830
831 /* We have just 22 bits per object for LRU information.
832 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
833 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
834 *
835 * Note that even if this will wrap after 1.5 years it's not a problem,
836 * everything will still work but just some object will appear younger
837 * to Redis. But for this to happen a given object should never be touched
838 * for 1.5 years.
839 *
840 * Note that you can change the resolution altering the
841 * REDIS_LRU_CLOCK_RESOLUTION define.
842 */
843 updateLRUClock();
844
845 /* Record the max memory used since the server was started. */
846 if (zmalloc_used_memory() > server.stat_peak_memory)
847 server.stat_peak_memory = zmalloc_used_memory();
848
849 /* We received a SIGTERM, shutting down here in a safe way, as it is
850 * not ok doing so inside the signal handler. */
851 if (server.shutdown_asap) {
852 if (prepareForShutdown(0) == REDIS_OK) exit(0);
853 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
854 }
855
856 /* Show some info about non-empty databases */
857 run_with_period(5000) {
858 for (j = 0; j < server.dbnum; j++) {
859 long long size, used, vkeys;
860
861 size = dictSlots(server.db[j].dict);
862 used = dictSize(server.db[j].dict);
863 vkeys = dictSize(server.db[j].expires);
864 if (used || vkeys) {
865 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
866 /* dictPrintStats(server.dict); */
867 }
868 }
869 }
870
871 /* We don't want to resize the hash tables while a bacground saving
872 * is in progress: the saving child is created using fork() that is
873 * implemented with a copy-on-write semantic in most modern systems, so
874 * if we resize the HT while there is the saving child at work actually
875 * a lot of memory movements in the parent will cause a lot of pages
876 * copied. */
877 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
878 tryResizeHashTables();
879 if (server.activerehashing) incrementallyRehash();
880 }
881
882 /* Show information about connected clients */
883 if (!server.sentinel_mode) {
884 run_with_period(5000) {
885 redisLog(REDIS_VERBOSE,
886 "%d clients connected (%d slaves), %zu bytes in use",
887 listLength(server.clients)-listLength(server.slaves),
888 listLength(server.slaves),
889 zmalloc_used_memory());
890 }
891 }
892
893 /* We need to do a few operations on clients asynchronously. */
894 clientsCron();
895
896 /* Start a scheduled AOF rewrite if this was requested by the user while
897 * a BGSAVE was in progress. */
898 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
899 server.aof_rewrite_scheduled)
900 {
901 rewriteAppendOnlyFileBackground();
902 }
903
904 /* Check if a background saving or AOF rewrite in progress terminated. */
905 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
906 int statloc;
907 pid_t pid;
908
909 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
910 int exitcode = WEXITSTATUS(statloc);
911 int bysignal = 0;
912
913 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
914
915 if (pid == server.rdb_child_pid) {
916 backgroundSaveDoneHandler(exitcode,bysignal);
917 } else if (pid == server.aof_child_pid) {
918 backgroundRewriteDoneHandler(exitcode,bysignal);
919 } else {
920 redisLog(REDIS_WARNING,
921 "Warning, detected child with unmatched pid: %ld",
922 (long)pid);
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 the sentinel timer if we are in sentinel mode. */
975 run_with_period(100) {
976 if (server.sentinel_mode) sentinelTimer();
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.execaborterr = createObject(REDIS_STRING,sdsnew(
1053 "-EXECABORT Transaction discarded because of previous errors.\r\n"));
1054 shared.space = createObject(REDIS_STRING,sdsnew(" "));
1055 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1056 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1057
1058 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1059 shared.select[j] = createObject(REDIS_STRING,
1060 sdscatprintf(sdsempty(),"select %d\r\n", j));
1061 }
1062 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1063 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1064 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1065 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1066 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1067 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1068 shared.del = createStringObject("DEL",3);
1069 shared.rpop = createStringObject("RPOP",4);
1070 shared.lpop = createStringObject("LPOP",4);
1071 shared.lpush = createStringObject("LPUSH",5);
1072 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1073 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1074 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1075 }
1076 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1077 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1078 sdscatprintf(sdsempty(),"*%d\r\n",j));
1079 shared.bulkhdr[j] = createObject(REDIS_STRING,
1080 sdscatprintf(sdsempty(),"$%d\r\n",j));
1081 }
1082 }
1083
1084 void initServerConfig() {
1085 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1086 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1087 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1088 server.port = REDIS_SERVERPORT;
1089 server.bindaddr = NULL;
1090 server.unixsocket = NULL;
1091 server.unixsocketperm = 0;
1092 server.ipfd = -1;
1093 server.sofd = -1;
1094 server.dbnum = REDIS_DEFAULT_DBNUM;
1095 server.verbosity = REDIS_NOTICE;
1096 server.maxidletime = REDIS_MAXIDLETIME;
1097 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1098 server.saveparams = NULL;
1099 server.loading = 0;
1100 server.logfile = NULL; /* NULL = log on standard output */
1101 server.syslog_enabled = 0;
1102 server.syslog_ident = zstrdup("redis");
1103 server.syslog_facility = LOG_LOCAL0;
1104 server.daemonize = 0;
1105 server.aof_state = REDIS_AOF_OFF;
1106 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1107 server.aof_no_fsync_on_rewrite = 0;
1108 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1109 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1110 server.aof_rewrite_base_size = 0;
1111 server.aof_rewrite_scheduled = 0;
1112 server.aof_last_fsync = time(NULL);
1113 server.aof_rewrite_time_last = -1;
1114 server.aof_rewrite_time_start = -1;
1115 server.aof_lastbgrewrite_status = REDIS_OK;
1116 server.aof_delayed_fsync = 0;
1117 server.aof_fd = -1;
1118 server.aof_selected_db = -1; /* Make sure the first time will not match */
1119 server.aof_flush_postponed_start = 0;
1120 server.pidfile = zstrdup("/var/run/redis.pid");
1121 server.rdb_filename = zstrdup("dump.rdb");
1122 server.aof_filename = zstrdup("appendonly.aof");
1123 server.mdb_state = REDIS_MDB_OFF;
1124 server.mdb_environment = zstrdup("archive");
1125 server.mdb_mapsize = 10485760;
1126 server.requirepass = NULL;
1127 server.rdb_compression = 1;
1128 server.rdb_checksum = 1;
1129 server.activerehashing = 1;
1130 server.maxclients = REDIS_MAX_CLIENTS;
1131 server.bpop_blocked_clients = 0;
1132 server.maxmemory = 0;
1133 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1134 server.maxmemory_samples = 3;
1135 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1136 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1137 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1138 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1139 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1140 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1141 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1142 server.shutdown_asap = 0;
1143 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1144 server.repl_timeout = REDIS_REPL_TIMEOUT;
1145 server.lua_caller = NULL;
1146 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1147 server.lua_client = NULL;
1148 server.lua_timedout = 0;
1149
1150 updateLRUClock();
1151 resetServerSaveParams();
1152
1153 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1154 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1155 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1156 /* Replication related */
1157 server.masterauth = NULL;
1158 server.masterhost = NULL;
1159 server.masterport = 6379;
1160 server.master = NULL;
1161 server.repl_state = REDIS_REPL_NONE;
1162 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1163 server.repl_serve_stale_data = 1;
1164 server.repl_slave_ro = 1;
1165 server.repl_down_since = time(NULL);
1166 server.slave_priority = REDIS_DEFAULT_SLAVE_PRIORITY;
1167
1168 /* Client output buffer limits */
1169 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1170 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1171 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1172 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1173 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1174 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1175 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1176 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1177 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1178
1179 /* Double constants initialization */
1180 R_Zero = 0.0;
1181 R_PosInf = 1.0/R_Zero;
1182 R_NegInf = -1.0/R_Zero;
1183 R_Nan = R_Zero/R_Zero;
1184
1185 /* Command table -- we intiialize it here as it is part of the
1186 * initial configuration, since command names may be changed via
1187 * redis.conf using the rename-command directive. */
1188 server.commands = dictCreate(&commandTableDictType,NULL);
1189 populateCommandTable();
1190 server.delCommand = lookupCommandByCString("del");
1191 server.multiCommand = lookupCommandByCString("multi");
1192 server.lpushCommand = lookupCommandByCString("lpush");
1193 server.lpopCommand = lookupCommandByCString("lpop");
1194 server.rpopCommand = lookupCommandByCString("rpop");
1195
1196 /* Slow log */
1197 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1198 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1199
1200 /* Debugging */
1201 server.assert_failed = "<no assertion failed>";
1202 server.assert_file = "<no file>";
1203 server.assert_line = 0;
1204 server.bug_report_start = 0;
1205 server.watchdog_period = 0;
1206 }
1207
1208 /* This function will try to raise the max number of open files accordingly to
1209 * the configured max number of clients. It will also account for 32 additional
1210 * file descriptors as we need a few more for persistence, listening
1211 * sockets, log files and so forth.
1212 *
1213 * If it will not be possible to set the limit accordingly to the configured
1214 * max number of clients, the function will do the reverse setting
1215 * server.maxclients to the value that we can actually handle. */
1216 void adjustOpenFilesLimit(void) {
1217 rlim_t maxfiles = server.maxclients+32;
1218 struct rlimit limit;
1219
1220 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1221 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1222 strerror(errno));
1223 server.maxclients = 1024-32;
1224 } else {
1225 rlim_t oldlimit = limit.rlim_cur;
1226
1227 /* Set the max number of files if the current limit is not enough
1228 * for our needs. */
1229 if (oldlimit < maxfiles) {
1230 rlim_t f;
1231
1232 f = maxfiles;
1233 while(f > oldlimit) {
1234 limit.rlim_cur = f;
1235 limit.rlim_max = f;
1236 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1237 f -= 128;
1238 }
1239 if (f < oldlimit) f = oldlimit;
1240 if (f != maxfiles) {
1241 server.maxclients = f-32;
1242 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1243 (int) maxfiles, strerror(errno), (int) server.maxclients);
1244 } else {
1245 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1246 (int) maxfiles);
1247 }
1248 }
1249 }
1250 }
1251
1252 void initServer() {
1253 int j;
1254
1255 signal(SIGHUP, SIG_IGN);
1256 signal(SIGPIPE, SIG_IGN);
1257 setupSignalHandlers();
1258
1259 if (server.syslog_enabled) {
1260 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1261 server.syslog_facility);
1262 }
1263
1264 server.current_client = NULL;
1265 server.clients = listCreate();
1266 server.clients_to_close = listCreate();
1267 server.slaves = listCreate();
1268 server.monitors = listCreate();
1269 server.unblocked_clients = listCreate();
1270 server.ready_keys = listCreate();
1271
1272 createSharedObjects();
1273 adjustOpenFilesLimit();
1274 server.el = aeCreateEventLoop(server.maxclients+1024);
1275 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1276
1277 if (server.port != 0) {
1278 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1279 if (server.ipfd == ANET_ERR) {
1280 redisLog(REDIS_WARNING, "Opening port %d: %s",
1281 server.port, server.neterr);
1282 exit(1);
1283 }
1284 }
1285 if (server.unixsocket != NULL) {
1286 unlink(server.unixsocket); /* don't care if this fails */
1287 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1288 if (server.sofd == ANET_ERR) {
1289 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1290 exit(1);
1291 }
1292 }
1293 if (server.ipfd < 0 && server.sofd < 0) {
1294 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1295 exit(1);
1296 }
1297 for (j = 0; j < server.dbnum; j++) {
1298 server.db[j].dict = dictCreate(&dbDictType,NULL);
1299 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1300 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1301 server.db[j].ready_keys = dictCreate(&setDictType,NULL);
1302 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1303 server.db[j].id = j;
1304 }
1305 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1306 server.pubsub_patterns = listCreate();
1307 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1308 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1309 server.cronloops = 0;
1310 server.rdb_child_pid = -1;
1311 server.aof_child_pid = -1;
1312 aofRewriteBufferReset();
1313 server.aof_buf = sdsempty();
1314 server.lastsave = time(NULL);
1315 server.rdb_save_time_last = -1;
1316 server.rdb_save_time_start = -1;
1317 server.dirty = 0;
1318 server.stat_numcommands = 0;
1319 server.stat_numconnections = 0;
1320 server.stat_expiredkeys = 0;
1321 server.stat_evictedkeys = 0;
1322 server.stat_starttime = time(NULL);
1323 server.stat_keyspace_misses = 0;
1324 server.stat_keyspace_hits = 0;
1325 server.stat_peak_memory = 0;
1326 server.stat_fork_time = 0;
1327 server.stat_rejected_conn = 0;
1328 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1329 server.ops_sec_idx = 0;
1330 server.ops_sec_last_sample_time = mstime();
1331 server.ops_sec_last_sample_ops = 0;
1332 server.unixtime = time(NULL);
1333 server.lastbgsave_status = REDIS_OK;
1334 server.stop_writes_on_bgsave_err = 1;
1335 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1336 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1337 acceptTcpHandler,NULL) == AE_ERR) redisPanic("Unrecoverable error creating server.ipfd file event.");
1338 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1339 acceptUnixHandler,NULL) == AE_ERR) redisPanic("Unrecoverable error creating server.sofd file event.");
1340
1341 if (server.aof_state == REDIS_AOF_ON) {
1342 server.aof_fd = open(server.aof_filename,
1343 O_WRONLY|O_APPEND|O_CREAT,0644);
1344 if (server.aof_fd == -1) {
1345 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1346 strerror(errno));
1347 exit(1);
1348 }
1349 }
1350
1351 if (server.aof_state == REDIS_MDB_ON) {
1352 int retval = startKeyArchive();
1353 if (retval != 0) {
1354 redisLog(REDIS_WARNING, "Can't open the key-archive environment: %s",
1355 mdb_strerror(retval));
1356 exit(1);
1357 }
1358 }
1359
1360 /* 32 bit instances are limited to 4GB of address space, so if there is
1361 * no explicit limit in the user provided configuration we set a limit
1362 * at 3 GB using maxmemory with 'noeviction' policy'. This avoids
1363 * useless crashes of the Redis instance for out of memory. */
1364 if (server.arch_bits == 32 && server.maxmemory == 0) {
1365 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3 GB maxmemory limit with 'noeviction' policy now.");
1366 server.maxmemory = 3072LL*(1024*1024); /* 3 GB */
1367 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1368 }
1369
1370 scriptingInit();
1371 slowlogInit();
1372 bioInit();
1373 }
1374
1375 /* Populates the Redis Command Table starting from the hard coded list
1376 * we have on top of redis.c file. */
1377 void populateCommandTable(void) {
1378 int j;
1379 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1380
1381 for (j = 0; j < numcommands; j++) {
1382 struct redisCommand *c = redisCommandTable+j;
1383 char *f = c->sflags;
1384 int retval;
1385
1386 while(*f != '\0') {
1387 switch(*f) {
1388 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1389 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1390 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1391 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1392 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1393 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1394 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1395 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1396 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1397 case 'l': c->flags |= REDIS_CMD_LOADING; break;
1398 case 't': c->flags |= REDIS_CMD_STALE; break;
1399 case 'M': c->flags |= REDIS_CMD_SKIP_MONITOR; break;
1400 default: redisPanic("Unsupported command flag"); break;
1401 }
1402 f++;
1403 }
1404
1405 retval = dictAdd(server.commands, sdsnew(c->name), c);
1406 assert(retval == DICT_OK);
1407 }
1408 }
1409
1410 void resetCommandTableStats(void) {
1411 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1412 int j;
1413
1414 for (j = 0; j < numcommands; j++) {
1415 struct redisCommand *c = redisCommandTable+j;
1416
1417 c->microseconds = 0;
1418 c->calls = 0;
1419 }
1420 }
1421
1422 /* ========================== Redis OP Array API ============================ */
1423
1424 void redisOpArrayInit(redisOpArray *oa) {
1425 oa->ops = NULL;
1426 oa->numops = 0;
1427 }
1428
1429 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1430 robj **argv, int argc, int target)
1431 {
1432 redisOp *op;
1433
1434 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1435 op = oa->ops+oa->numops;
1436 op->cmd = cmd;
1437 op->dbid = dbid;
1438 op->argv = argv;
1439 op->argc = argc;
1440 op->target = target;
1441 oa->numops++;
1442 return oa->numops;
1443 }
1444
1445 void redisOpArrayFree(redisOpArray *oa) {
1446 while(oa->numops) {
1447 int j;
1448 redisOp *op;
1449
1450 oa->numops--;
1451 op = oa->ops+oa->numops;
1452 for (j = 0; j < op->argc; j++)
1453 decrRefCount(op->argv[j]);
1454 zfree(op->argv);
1455 }
1456 zfree(oa->ops);
1457 }
1458
1459 /* ====================== Commands lookup and execution ===================== */
1460
1461 struct redisCommand *lookupCommand(sds name) {
1462 return dictFetchValue(server.commands, name);
1463 }
1464
1465 struct redisCommand *lookupCommandByCString(char *s) {
1466 struct redisCommand *cmd;
1467 sds name = sdsnew(s);
1468
1469 cmd = dictFetchValue(server.commands, name);
1470 sdsfree(name);
1471 return cmd;
1472 }
1473
1474 /* Propagate the specified command (in the context of the specified database id)
1475 * to AOF and Slaves.
1476 *
1477 * flags are an xor between:
1478 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1479 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1480 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1481 */
1482 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1483 int flags)
1484 {
1485 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1486 feedAppendOnlyFile(cmd,dbid,argv,argc);
1487 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1488 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1489 }
1490
1491 /* Used inside commands to schedule the propagation of additional commands
1492 * after the current command is propagated to AOF / Replication. */
1493 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1494 int target)
1495 {
1496 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1497 }
1498
1499 /* Call() is the core of Redis execution of a command */
1500 void call(redisClient *c, int flags) {
1501 long long dirty, start = ustime(), duration;
1502
1503 /* Sent the command to clients in MONITOR mode, only if the commands are
1504 * not geneated from reading an AOF. */
1505 if (listLength(server.monitors) &&
1506 !server.loading &&
1507 !(c->cmd->flags & REDIS_CMD_SKIP_MONITOR))
1508 {
1509 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1510 }
1511
1512 /* Call the command. */
1513 redisOpArrayInit(&server.also_propagate);
1514 dirty = server.dirty;
1515 c->cmd->proc(c);
1516 dirty = server.dirty-dirty;
1517 duration = ustime()-start;
1518
1519 /* When EVAL is called loading the AOF we don't want commands called
1520 * from Lua to go into the slowlog or to populate statistics. */
1521 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1522 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1523
1524 /* Log the command into the Slow log if needed, and populate the
1525 * per-command statistics that we show in INFO commandstats. */
1526 if (flags & REDIS_CALL_SLOWLOG)
1527 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1528 if (flags & REDIS_CALL_STATS) {
1529 c->cmd->microseconds += duration;
1530 c->cmd->calls++;
1531 }
1532
1533 /* Propagate the command into the AOF and replication link */
1534 if (flags & REDIS_CALL_PROPAGATE) {
1535 int flags = REDIS_PROPAGATE_NONE;
1536
1537 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1538 flags |= REDIS_PROPAGATE_REPL;
1539 if (dirty)
1540 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1541 if (flags != REDIS_PROPAGATE_NONE)
1542 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1543 }
1544 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1545 * PUSH command. */
1546 if (server.also_propagate.numops) {
1547 int j;
1548 redisOp *rop;
1549
1550 for (j = 0; j < server.also_propagate.numops; j++) {
1551 rop = &server.also_propagate.ops[j];
1552 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1553 }
1554 redisOpArrayFree(&server.also_propagate);
1555 }
1556 server.stat_numcommands++;
1557 }
1558
1559 /* If this function gets called we already read a whole
1560 * command, arguments are in the client argv/argc fields.
1561 * processCommand() execute the command or prepare the
1562 * server for a bulk read from the client.
1563 *
1564 * If 1 is returned the client is still alive and valid and
1565 * and other operations can be performed by the caller. Otherwise
1566 * if 0 is returned the client was destroied (i.e. after QUIT). */
1567 int processCommand(redisClient *c) {
1568 /* The QUIT command is handled separately. Normal command procs will
1569 * go through checking for replication and QUIT will cause trouble
1570 * when FORCE_REPLICATION is enabled and would be implemented in
1571 * a regular command proc. */
1572 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1573 addReply(c,shared.ok);
1574 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1575 return REDIS_ERR;
1576 }
1577
1578 /* Now lookup the command and check ASAP about trivial error conditions
1579 * such as wrong arity, bad command name and so forth. */
1580 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1581 if (!c->cmd) {
1582 flagTransaction(c);
1583 addReplyErrorFormat(c,"unknown command '%s'",
1584 (char*)c->argv[0]->ptr);
1585 return REDIS_OK;
1586 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1587 (c->argc < -c->cmd->arity)) {
1588 flagTransaction(c);
1589 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1590 c->cmd->name);
1591 return REDIS_OK;
1592 }
1593
1594 /* Check if the user is authenticated */
1595 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1596 {
1597 flagTransaction(c);
1598 addReplyError(c,"operation not permitted");
1599 return REDIS_OK;
1600 }
1601
1602 /* Handle the maxmemory directive.
1603 *
1604 * First we try to free some memory if possible (if there are volatile
1605 * keys in the dataset). If there are not the only thing we can do
1606 * is returning an error. */
1607 if (server.maxmemory) {
1608 int retval = freeMemoryIfNeeded();
1609 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1610 flagTransaction(c);
1611 addReply(c, shared.oomerr);
1612 return REDIS_OK;
1613 }
1614 }
1615
1616 /* Don't accept write commands if there are problems persisting on disk. */
1617 if (server.stop_writes_on_bgsave_err &&
1618 server.saveparamslen > 0
1619 && server.lastbgsave_status == REDIS_ERR &&
1620 c->cmd->flags & REDIS_CMD_WRITE)
1621 {
1622 flagTransaction(c);
1623 addReply(c, shared.bgsaveerr);
1624 return REDIS_OK;
1625 }
1626
1627 /* Don't accept write commands if this is a read only slave. But
1628 * accept write commands if this is our master. */
1629 if (server.masterhost && server.repl_slave_ro &&
1630 !(c->flags & REDIS_MASTER) &&
1631 c->cmd->flags & REDIS_CMD_WRITE)
1632 {
1633 addReply(c, shared.roslaveerr);
1634 return REDIS_OK;
1635 }
1636
1637 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1638 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1639 &&
1640 c->cmd->proc != subscribeCommand &&
1641 c->cmd->proc != unsubscribeCommand &&
1642 c->cmd->proc != psubscribeCommand &&
1643 c->cmd->proc != punsubscribeCommand) {
1644 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1645 return REDIS_OK;
1646 }
1647
1648 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1649 * we are a slave with a broken link with master. */
1650 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1651 server.repl_serve_stale_data == 0 &&
1652 !(c->cmd->flags & REDIS_CMD_STALE))
1653 {
1654 flagTransaction(c);
1655 addReply(c, shared.masterdownerr);
1656 return REDIS_OK;
1657 }
1658
1659 /* Loading DB? Return an error if the command has not the
1660 * REDIS_CMD_LOADING flag. */
1661 if (server.loading && !(c->cmd->flags & REDIS_CMD_LOADING)) {
1662 addReply(c, shared.loadingerr);
1663 return REDIS_OK;
1664 }
1665
1666 /* Lua script too slow? Only allow commands with REDIS_CMD_STALE flag. */
1667 if (server.lua_timedout &&
1668 c->cmd->proc != authCommand &&
1669 !(c->cmd->proc == shutdownCommand &&
1670 c->argc == 2 &&
1671 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1672 !(c->cmd->proc == scriptCommand &&
1673 c->argc == 2 &&
1674 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1675 {
1676 flagTransaction(c);
1677 addReply(c, shared.slowscripterr);
1678 return REDIS_OK;
1679 }
1680
1681 /* Exec the command */
1682 if (c->flags & REDIS_MULTI &&
1683 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1684 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1685 {
1686 queueMultiCommand(c);
1687 addReply(c,shared.queued);
1688 } else {
1689 call(c,REDIS_CALL_FULL);
1690 if (listLength(server.ready_keys))
1691 handleClientsBlockedOnLists();
1692 }
1693 return REDIS_OK;
1694 }
1695
1696 /*================================== Shutdown =============================== */
1697
1698 int prepareForShutdown(int flags) {
1699 int save = flags & REDIS_SHUTDOWN_SAVE;
1700 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1701
1702 redisLog(REDIS_WARNING,"User requested shutdown...");
1703 /* Kill the saving child if there is a background saving in progress.
1704 We want to avoid race conditions, for instance our saving child may
1705 overwrite the synchronous saving did by SHUTDOWN. */
1706 if (server.rdb_child_pid != -1) {
1707 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1708 kill(server.rdb_child_pid,SIGKILL);
1709 rdbRemoveTempFile(server.rdb_child_pid);
1710 }
1711 if (server.aof_state != REDIS_AOF_OFF) {
1712 /* Kill the AOF saving child as the AOF we already have may be longer
1713 * but contains the full dataset anyway. */
1714 if (server.aof_child_pid != -1) {
1715 redisLog(REDIS_WARNING,
1716 "There is a child rewriting the AOF. Killing it!");
1717 kill(server.aof_child_pid,SIGKILL);
1718 }
1719 /* Append only file: fsync() the AOF and exit */
1720 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1721 aof_fsync(server.aof_fd);
1722 }
1723 if ((server.saveparamslen > 0 && !nosave) || save) {
1724 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1725 /* Snapshotting. Perform a SYNC SAVE and exit */
1726 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1727 /* Ooops.. error saving! The best we can do is to continue
1728 * operating. Note that if there was a background saving process,
1729 * in the next cron() Redis will be notified that the background
1730 * saving aborted, handling special stuff like slaves pending for
1731 * synchronization... */
1732 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1733 return REDIS_ERR;
1734 }
1735 }
1736 if (server.daemonize) {
1737 redisLog(REDIS_NOTICE,"Removing the pid file.");
1738 unlink(server.pidfile);
1739 }
1740 /* Close the listening sockets. Apparently this allows faster restarts. */
1741 if (server.ipfd != -1) close(server.ipfd);
1742 if (server.sofd != -1) close(server.sofd);
1743 if (server.unixsocket) {
1744 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1745 unlink(server.unixsocket); /* don't care if this fails */
1746 }
1747
1748 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1749 return REDIS_OK;
1750 }
1751
1752 /*================================== Commands =============================== */
1753
1754 /* Return zero if strings are the same, non-zero if they are not.
1755 * The comparison is performed in a way that prevents an attacker to obtain
1756 * information about the nature of the strings just monitoring the execution
1757 * time of the function.
1758 *
1759 * Note that limiting the comparison length to strings up to 512 bytes we
1760 * can avoid leaking any information about the password length and any
1761 * possible branch misprediction related leak.
1762 */
1763 int time_independent_strcmp(char *a, char *b) {
1764 char bufa[REDIS_AUTHPASS_MAX_LEN], bufb[REDIS_AUTHPASS_MAX_LEN];
1765 /* The above two strlen perform len(a) + len(b) operations where either
1766 * a or b are fixed (our password) length, and the difference is only
1767 * relative to the length of the user provided string, so no information
1768 * leak is possible in the following two lines of code. */
1769 int alen = strlen(a);
1770 int blen = strlen(b);
1771 int j;
1772 int diff = 0;
1773
1774 /* We can't compare strings longer than our static buffers.
1775 * Note that this will never pass the first test in practical circumstances
1776 * so there is no info leak. */
1777 if (alen > sizeof(bufa) || blen > sizeof(bufb)) return 1;
1778
1779 memset(bufa,0,sizeof(bufa)); /* Constant time. */
1780 memset(bufb,0,sizeof(bufb)); /* Constant time. */
1781 /* Again the time of the following two copies is proportional to
1782 * len(a) + len(b) so no info is leaked. */
1783 memcpy(bufa,a,alen);
1784 memcpy(bufb,b,blen);
1785
1786 /* Always compare all the chars in the two buffers without
1787 * conditional expressions. */
1788 for (j = 0; j < sizeof(bufa); j++) {
1789 diff |= (bufa[j] ^ bufb[j]);
1790 }
1791 /* Length must be equal as well. */
1792 diff |= alen ^ blen;
1793 return diff; /* If zero strings are the same. */
1794 }
1795
1796 void authCommand(redisClient *c) {
1797 if (!server.requirepass) {
1798 addReplyError(c,"Client sent AUTH, but no password is set");
1799 } else if (!time_independent_strcmp(c->argv[1]->ptr, server.requirepass)) {
1800 c->authenticated = 1;
1801 addReply(c,shared.ok);
1802 } else {
1803 c->authenticated = 0;
1804 addReplyError(c,"invalid password");
1805 }
1806 }
1807
1808 void pingCommand(redisClient *c) {
1809 addReply(c,shared.pong);
1810 }
1811
1812 void echoCommand(redisClient *c) {
1813 addReplyBulk(c,c->argv[1]);
1814 }
1815
1816 void timeCommand(redisClient *c) {
1817 struct timeval tv;
1818
1819 /* gettimeofday() can only fail if &tv is a bad addresss so we
1820 * don't check for errors. */
1821 gettimeofday(&tv,NULL);
1822 addReplyMultiBulkLen(c,2);
1823 addReplyBulkLongLong(c,tv.tv_sec);
1824 addReplyBulkLongLong(c,tv.tv_usec);
1825 }
1826
1827 /* Convert an amount of bytes into a human readable string in the form
1828 * of 100B, 2G, 100M, 4K, and so forth. */
1829 void bytesToHuman(char *s, unsigned long long n) {
1830 double d;
1831
1832 if (n < 1024) {
1833 /* Bytes */
1834 sprintf(s,"%lluB",n);
1835 return;
1836 } else if (n < (1024*1024)) {
1837 d = (double)n/(1024);
1838 sprintf(s,"%.2fK",d);
1839 } else if (n < (1024LL*1024*1024)) {
1840 d = (double)n/(1024*1024);
1841 sprintf(s,"%.2fM",d);
1842 } else if (n < (1024LL*1024*1024*1024)) {
1843 d = (double)n/(1024LL*1024*1024);
1844 sprintf(s,"%.2fG",d);
1845 }
1846 }
1847
1848 /* Create the string returned by the INFO command. This is decoupled
1849 * by the INFO command itself as we need to report the same information
1850 * on memory corruption problems. */
1851 sds genRedisInfoString(char *section) {
1852 sds info = sdsempty();
1853 time_t uptime = server.unixtime-server.stat_starttime;
1854 int j, numcommands;
1855 struct rusage self_ru, c_ru;
1856 unsigned long lol, bib;
1857 int allsections = 0, defsections = 0;
1858 int sections = 0;
1859
1860 if (section) {
1861 allsections = strcasecmp(section,"all") == 0;
1862 defsections = strcasecmp(section,"default") == 0;
1863 }
1864
1865 getrusage(RUSAGE_SELF, &self_ru);
1866 getrusage(RUSAGE_CHILDREN, &c_ru);
1867 getClientsMaxBuffers(&lol,&bib);
1868
1869 /* Server */
1870 if (allsections || defsections || !strcasecmp(section,"server")) {
1871 struct utsname name;
1872 char *mode;
1873
1874 if (server.sentinel_mode) mode = "sentinel";
1875 else mode = "standalone";
1876
1877 if (sections++) info = sdscat(info,"\r\n");
1878 uname(&name);
1879 info = sdscatprintf(info,
1880 "# Server\r\n"
1881 "redis_version:%s\r\n"
1882 "redis_git_sha1:%s\r\n"
1883 "redis_git_dirty:%d\r\n"
1884 "redis_mode:%s\r\n"
1885 "os:%s %s %s\r\n"
1886 "arch_bits:%d\r\n"
1887 "multiplexing_api:%s\r\n"
1888 "gcc_version:%d.%d.%d\r\n"
1889 "process_id:%ld\r\n"
1890 "run_id:%s\r\n"
1891 "tcp_port:%d\r\n"
1892 "uptime_in_seconds:%ld\r\n"
1893 "uptime_in_days:%ld\r\n"
1894 "lru_clock:%ld\r\n",
1895 REDIS_VERSION,
1896 redisGitSHA1(),
1897 strtol(redisGitDirty(),NULL,10) > 0,
1898 mode,
1899 name.sysname, name.release, name.machine,
1900 server.arch_bits,
1901 aeGetApiName(),
1902 #ifdef __GNUC__
1903 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1904 #else
1905 0,0,0,
1906 #endif
1907 (long) getpid(),
1908 server.runid,
1909 server.port,
1910 uptime,
1911 uptime/(3600*24),
1912 (unsigned long) server.lruclock);
1913 }
1914
1915 /* Clients */
1916 if (allsections || defsections || !strcasecmp(section,"clients")) {
1917 if (sections++) info = sdscat(info,"\r\n");
1918 info = sdscatprintf(info,
1919 "# Clients\r\n"
1920 "connected_clients:%lu\r\n"
1921 "client_longest_output_list:%lu\r\n"
1922 "client_biggest_input_buf:%lu\r\n"
1923 "blocked_clients:%d\r\n",
1924 listLength(server.clients)-listLength(server.slaves),
1925 lol, bib,
1926 server.bpop_blocked_clients);
1927 }
1928
1929 /* Memory */
1930 if (allsections || defsections || !strcasecmp(section,"memory")) {
1931 char hmem[64];
1932 char peak_hmem[64];
1933
1934 bytesToHuman(hmem,zmalloc_used_memory());
1935 bytesToHuman(peak_hmem,server.stat_peak_memory);
1936 if (sections++) info = sdscat(info,"\r\n");
1937 info = sdscatprintf(info,
1938 "# Memory\r\n"
1939 "used_memory:%zu\r\n"
1940 "used_memory_human:%s\r\n"
1941 "used_memory_rss:%zu\r\n"
1942 "used_memory_peak:%zu\r\n"
1943 "used_memory_peak_human:%s\r\n"
1944 "used_memory_lua:%lld\r\n"
1945 "mem_fragmentation_ratio:%.2f\r\n"
1946 "mem_allocator:%s\r\n",
1947 zmalloc_used_memory(),
1948 hmem,
1949 zmalloc_get_rss(),
1950 server.stat_peak_memory,
1951 peak_hmem,
1952 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1953 zmalloc_get_fragmentation_ratio(),
1954 ZMALLOC_LIB
1955 );
1956 }
1957
1958 /* Persistence */
1959 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1960 if (sections++) info = sdscat(info,"\r\n");
1961 info = sdscatprintf(info,
1962 "# Persistence\r\n"
1963 "loading:%d\r\n"
1964 "rdb_changes_since_last_save:%lld\r\n"
1965 "rdb_bgsave_in_progress:%d\r\n"
1966 "rdb_last_save_time:%ld\r\n"
1967 "rdb_last_bgsave_status:%s\r\n"
1968 "rdb_last_bgsave_time_sec:%ld\r\n"
1969 "rdb_current_bgsave_time_sec:%ld\r\n"
1970 "aof_enabled:%d\r\n"
1971 "aof_rewrite_in_progress:%d\r\n"
1972 "aof_rewrite_scheduled:%d\r\n"
1973 "aof_last_rewrite_time_sec:%ld\r\n"
1974 "aof_current_rewrite_time_sec:%ld\r\n"
1975 "aof_last_bgrewrite_status:%s\r\n",
1976 server.loading,
1977 server.dirty,
1978 server.rdb_child_pid != -1,
1979 server.lastsave,
1980 (server.lastbgsave_status == REDIS_OK) ? "ok" : "err",
1981 server.rdb_save_time_last,
1982 (server.rdb_child_pid == -1) ?
1983 -1 : time(NULL)-server.rdb_save_time_start,
1984 server.aof_state != REDIS_AOF_OFF,
1985 server.aof_child_pid != -1,
1986 server.aof_rewrite_scheduled,
1987 server.aof_rewrite_time_last,
1988 (server.aof_child_pid == -1) ?
1989 -1 : time(NULL)-server.aof_rewrite_time_start,
1990 (server.aof_lastbgrewrite_status == REDIS_OK) ? "ok" : "err");
1991
1992 if (server.aof_state != REDIS_AOF_OFF) {
1993 info = sdscatprintf(info,
1994 "aof_current_size:%lld\r\n"
1995 "aof_base_size:%lld\r\n"
1996 "aof_pending_rewrite:%d\r\n"
1997 "aof_buffer_length:%zu\r\n"
1998 "aof_rewrite_buffer_length:%lu\r\n"
1999 "aof_pending_bio_fsync:%llu\r\n"
2000 "aof_delayed_fsync:%lu\r\n",
2001 (long long) server.aof_current_size,
2002 (long long) server.aof_rewrite_base_size,
2003 server.aof_rewrite_scheduled,
2004 sdslen(server.aof_buf),
2005 aofRewriteBufferSize(),
2006 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
2007 server.aof_delayed_fsync);
2008 }
2009
2010 if (server.loading) {
2011 double perc;
2012 time_t eta, elapsed;
2013 off_t remaining_bytes = server.loading_total_bytes-
2014 server.loading_loaded_bytes;
2015
2016 perc = ((double)server.loading_loaded_bytes /
2017 server.loading_total_bytes) * 100;
2018
2019 elapsed = server.unixtime-server.loading_start_time;
2020 if (elapsed == 0) {
2021 eta = 1; /* A fake 1 second figure if we don't have
2022 enough info */
2023 } else {
2024 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
2025 }
2026
2027 info = sdscatprintf(info,
2028 "loading_start_time:%ld\r\n"
2029 "loading_total_bytes:%llu\r\n"
2030 "loading_loaded_bytes:%llu\r\n"
2031 "loading_loaded_perc:%.2f\r\n"
2032 "loading_eta_seconds:%ld\r\n"
2033 ,(unsigned long) server.loading_start_time,
2034 (unsigned long long) server.loading_total_bytes,
2035 (unsigned long long) server.loading_loaded_bytes,
2036 perc,
2037 eta
2038 );
2039 }
2040 }
2041
2042 /* Stats */
2043 if (allsections || defsections || !strcasecmp(section,"stats")) {
2044 if (sections++) info = sdscat(info,"\r\n");
2045 info = sdscatprintf(info,
2046 "# Stats\r\n"
2047 "total_connections_received:%lld\r\n"
2048 "total_commands_processed:%lld\r\n"
2049 "instantaneous_ops_per_sec:%lld\r\n"
2050 "rejected_connections:%lld\r\n"
2051 "expired_keys:%lld\r\n"
2052 "evicted_keys:%lld\r\n"
2053 "keyspace_hits:%lld\r\n"
2054 "keyspace_misses:%lld\r\n"
2055 "pubsub_channels:%ld\r\n"
2056 "pubsub_patterns:%lu\r\n"
2057 "latest_fork_usec:%lld\r\n",
2058 server.stat_numconnections,
2059 server.stat_numcommands,
2060 getOperationsPerSecond(),
2061 server.stat_rejected_conn,
2062 server.stat_expiredkeys,
2063 server.stat_evictedkeys,
2064 server.stat_keyspace_hits,
2065 server.stat_keyspace_misses,
2066 dictSize(server.pubsub_channels),
2067 listLength(server.pubsub_patterns),
2068 server.stat_fork_time);
2069 }
2070
2071 /* Replication */
2072 if (allsections || defsections || !strcasecmp(section,"replication")) {
2073 if (sections++) info = sdscat(info,"\r\n");
2074 info = sdscatprintf(info,
2075 "# Replication\r\n"
2076 "role:%s\r\n",
2077 server.masterhost == NULL ? "master" : "slave");
2078 if (server.masterhost) {
2079 info = sdscatprintf(info,
2080 "master_host:%s\r\n"
2081 "master_port:%d\r\n"
2082 "master_link_status:%s\r\n"
2083 "master_last_io_seconds_ago:%d\r\n"
2084 "master_sync_in_progress:%d\r\n"
2085 ,server.masterhost,
2086 server.masterport,
2087 (server.repl_state == REDIS_REPL_CONNECTED) ?
2088 "up" : "down",
2089 server.master ?
2090 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
2091 server.repl_state == REDIS_REPL_TRANSFER
2092 );
2093
2094 if (server.repl_state == REDIS_REPL_TRANSFER) {
2095 info = sdscatprintf(info,
2096 "master_sync_left_bytes:%lld\r\n"
2097 "master_sync_last_io_seconds_ago:%d\r\n"
2098 , (long long)
2099 (server.repl_transfer_size - server.repl_transfer_read),
2100 (int)(server.unixtime-server.repl_transfer_lastio)
2101 );
2102 }
2103
2104 if (server.repl_state != REDIS_REPL_CONNECTED) {
2105 info = sdscatprintf(info,
2106 "master_link_down_since_seconds:%ld\r\n",
2107 (long)server.unixtime-server.repl_down_since);
2108 }
2109 info = sdscatprintf(info,
2110 "slave_priority:%d\r\n"
2111 "slave_read_only:%d\r\n",
2112 server.slave_priority,
2113 server.repl_slave_ro);
2114 }
2115 info = sdscatprintf(info,
2116 "connected_slaves:%lu\r\n",
2117 listLength(server.slaves));
2118 if (listLength(server.slaves)) {
2119 int slaveid = 0;
2120 listNode *ln;
2121 listIter li;
2122
2123 listRewind(server.slaves,&li);
2124 while((ln = listNext(&li))) {
2125 redisClient *slave = listNodeValue(ln);
2126 char *state = NULL;
2127 char ip[32];
2128 int port;
2129
2130 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
2131 switch(slave->replstate) {
2132 case REDIS_REPL_WAIT_BGSAVE_START:
2133 case REDIS_REPL_WAIT_BGSAVE_END:
2134 state = "wait_bgsave";
2135 break;
2136 case REDIS_REPL_SEND_BULK:
2137 state = "send_bulk";
2138 break;
2139 case REDIS_REPL_ONLINE:
2140 state = "online";
2141 break;
2142 }
2143 if (state == NULL) continue;
2144 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2145 slaveid,ip,slave->slave_listening_port,state);
2146 slaveid++;
2147 }
2148 }
2149 }
2150
2151 /* CPU */
2152 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2153 if (sections++) info = sdscat(info,"\r\n");
2154 info = sdscatprintf(info,
2155 "# CPU\r\n"
2156 "used_cpu_sys:%.2f\r\n"
2157 "used_cpu_user:%.2f\r\n"
2158 "used_cpu_sys_children:%.2f\r\n"
2159 "used_cpu_user_children:%.2f\r\n",
2160 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2161 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2162 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2163 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2164 }
2165
2166 /* cmdtime */
2167 if (allsections || !strcasecmp(section,"commandstats")) {
2168 if (sections++) info = sdscat(info,"\r\n");
2169 info = sdscatprintf(info, "# Commandstats\r\n");
2170 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2171 for (j = 0; j < numcommands; j++) {
2172 struct redisCommand *c = redisCommandTable+j;
2173
2174 if (!c->calls) continue;
2175 info = sdscatprintf(info,
2176 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2177 c->name, c->calls, c->microseconds,
2178 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2179 }
2180 }
2181
2182 /* Key space */
2183 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2184 if (sections++) info = sdscat(info,"\r\n");
2185 info = sdscatprintf(info, "# Keyspace\r\n");
2186 for (j = 0; j < server.dbnum; j++) {
2187 long long keys, vkeys;
2188
2189 keys = dictSize(server.db[j].dict);
2190 vkeys = dictSize(server.db[j].expires);
2191 if (keys || vkeys) {
2192 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2193 j, keys, vkeys);
2194 }
2195 }
2196 }
2197 return info;
2198 }
2199
2200 void infoCommand(redisClient *c) {
2201 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2202
2203 if (c->argc > 2) {
2204 addReply(c,shared.syntaxerr);
2205 return;
2206 }
2207 sds info = genRedisInfoString(section);
2208 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2209 (unsigned long)sdslen(info)));
2210 addReplySds(c,info);
2211 addReply(c,shared.crlf);
2212 }
2213
2214 void monitorCommand(redisClient *c) {
2215 /* ignore MONITOR if already slave or in monitor mode */
2216 if (c->flags & REDIS_SLAVE) return;
2217
2218 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2219 c->slaveseldb = 0;
2220 listAddNodeTail(server.monitors,c);
2221 addReply(c,shared.ok);
2222 }
2223
2224 /* ============================ Maxmemory directive ======================== */
2225
2226 /* This function gets called when 'maxmemory' is set on the config file to limit
2227 * the max memory used by the server, before processing a command.
2228 *
2229 * The goal of the function is to free enough memory to keep Redis under the
2230 * configured memory limit.
2231 *
2232 * The function starts calculating how many bytes should be freed to keep
2233 * Redis under the limit, and enters a loop selecting the best keys to
2234 * evict accordingly to the configured policy.
2235 *
2236 * If all the bytes needed to return back under the limit were freed the
2237 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2238 * should block the execution of commands that will result in more memory
2239 * used by the server.
2240 */
2241 int freeMemoryIfNeeded(void) {
2242 size_t mem_used, mem_tofree, mem_freed;
2243 int slaves = listLength(server.slaves);
2244
2245 /* Remove the size of slaves output buffers and AOF buffer from the
2246 * count of used memory. */
2247 mem_used = zmalloc_used_memory();
2248 if (slaves) {
2249 listIter li;
2250 listNode *ln;
2251
2252 listRewind(server.slaves,&li);
2253 while((ln = listNext(&li))) {
2254 redisClient *slave = listNodeValue(ln);
2255 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2256 if (obuf_bytes > mem_used)
2257 mem_used = 0;
2258 else
2259 mem_used -= obuf_bytes;
2260 }
2261 }
2262 if (server.aof_state != REDIS_AOF_OFF) {
2263 mem_used -= sdslen(server.aof_buf);
2264 mem_used -= aofRewriteBufferSize();
2265 }
2266
2267 /* Check if we are over the memory limit. */
2268 if (mem_used <= server.maxmemory) return REDIS_OK;
2269
2270 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2271 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2272
2273 /* Compute how much memory we need to free. */
2274 mem_tofree = mem_used - server.maxmemory;
2275 mem_freed = 0;
2276 while (mem_freed < mem_tofree) {
2277 int j, k, keys_freed = 0;
2278
2279 for (j = 0; j < server.dbnum; j++) {
2280 long bestval = 0; /* just to prevent warning */
2281 sds bestkey = NULL;
2282 struct dictEntry *de;
2283 redisDb *db = server.db+j;
2284 dict *dict;
2285
2286 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2287 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2288 {
2289 dict = server.db[j].dict;
2290 } else {
2291 dict = server.db[j].expires;
2292 }
2293 if (dictSize(dict) == 0) continue;
2294
2295 /* volatile-random and allkeys-random policy */
2296 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2297 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2298 {
2299 de = dictGetRandomKey(dict);
2300 bestkey = dictGetKey(de);
2301 }
2302
2303 /* volatile-lru and allkeys-lru policy */
2304 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2305 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2306 {
2307 for (k = 0; k < server.maxmemory_samples; k++) {
2308 sds thiskey;
2309 long thisval;
2310 robj *o;
2311
2312 de = dictGetRandomKey(dict);
2313 thiskey = dictGetKey(de);
2314 /* When policy is volatile-lru we need an additional lookup
2315 * to locate the real key, as dict is set to db->expires. */
2316 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2317 de = dictFind(db->dict, thiskey);
2318 o = dictGetVal(de);
2319 thisval = estimateObjectIdleTime(o);
2320
2321 /* Higher idle time is better candidate for deletion */
2322 if (bestkey == NULL || thisval > bestval) {
2323 bestkey = thiskey;
2324 bestval = thisval;
2325 }
2326 }
2327 }
2328
2329 /* volatile-ttl */
2330 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2331 for (k = 0; k < server.maxmemory_samples; k++) {
2332 sds thiskey;
2333 long thisval;
2334
2335 de = dictGetRandomKey(dict);
2336 thiskey = dictGetKey(de);
2337 thisval = (long) dictGetVal(de);
2338
2339 /* Expire sooner (minor expire unix timestamp) is better
2340 * candidate for deletion */
2341 if (bestkey == NULL || thisval < bestval) {
2342 bestkey = thiskey;
2343 bestval = thisval;
2344 }
2345 }
2346 }
2347
2348 /* Finally remove the selected key. */
2349 if (bestkey) {
2350 long long delta;
2351
2352 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2353 int archived = archive(db, keyobj);
2354 redisAssert(archived != 0);
2355 propagateExpire(db,keyobj);
2356 /* We compute the amount of memory freed by dbDelete() alone.
2357 * It is possible that actually the memory needed to propagate
2358 * the DEL in AOF and replication link is greater than the one
2359 * we are freeing removing the key, but we can't account for
2360 * that otherwise we would never exit the loop.
2361 *
2362 * AOF and Output buffer memory will be freed eventually so
2363 * we only care about memory used by the key space. */
2364 delta = (long long) zmalloc_used_memory();
2365 dbDelete(db,keyobj);
2366 delta -= (long long) zmalloc_used_memory();
2367 mem_freed += delta;
2368 server.stat_evictedkeys++;
2369 decrRefCount(keyobj);
2370 keys_freed++;
2371
2372 /* When the memory to free starts to be big enough, we may
2373 * start spending so much time here that is impossible to
2374 * deliver data to the slaves fast enough, so we force the
2375 * transmission here inside the loop. */
2376 if (slaves) flushSlavesOutputBuffers();
2377 }
2378 }
2379 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2380 }
2381 return REDIS_OK;
2382 }
2383
2384 /* =================================== Main! ================================ */
2385
2386 #ifdef __linux__
2387 int linuxOvercommitMemoryValue(void) {
2388 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2389 char buf[64];
2390
2391 if (!fp) return -1;
2392 if (fgets(buf,64,fp) == NULL) {
2393 fclose(fp);
2394 return -1;
2395 }
2396 fclose(fp);
2397
2398 return atoi(buf);
2399 }
2400
2401 void linuxOvercommitMemoryWarning(void) {
2402 if (linuxOvercommitMemoryValue() == 0) {
2403 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.");
2404 }
2405 }
2406 #endif /* __linux__ */
2407
2408 void createPidFile(void) {
2409 /* Try to write the pid file in a best-effort way. */
2410 FILE *fp = fopen(server.pidfile,"w");
2411 if (fp) {
2412 fprintf(fp,"%d\n",(int)getpid());
2413 fclose(fp);
2414 }
2415 }
2416
2417 void daemonize(void) {
2418 int fd;
2419
2420 if (fork() != 0) exit(0); /* parent exits */
2421 setsid(); /* create a new session */
2422
2423 /* Every output goes to /dev/null. If Redis is daemonized but
2424 * the 'logfile' is set to 'stdout' in the configuration file
2425 * it will not log at all. */
2426 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2427 dup2(fd, STDIN_FILENO);
2428 dup2(fd, STDOUT_FILENO);
2429 dup2(fd, STDERR_FILENO);
2430 if (fd > STDERR_FILENO) close(fd);
2431 }
2432 }
2433
2434 void version() {
2435 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2436 REDIS_VERSION,
2437 redisGitSHA1(),
2438 atoi(redisGitDirty()) > 0,
2439 ZMALLOC_LIB,
2440 sizeof(long) == 4 ? 32 : 64);
2441 exit(0);
2442 }
2443
2444 void usage() {
2445 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2446 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2447 fprintf(stderr," ./redis-server -v or --version\n");
2448 fprintf(stderr," ./redis-server -h or --help\n");
2449 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2450 fprintf(stderr,"Examples:\n");
2451 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2452 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2453 fprintf(stderr," ./redis-server --port 7777\n");
2454 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2455 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n\n");
2456 fprintf(stderr,"Sentinel mode:\n");
2457 fprintf(stderr," ./redis-server /etc/sentinel.conf --sentinel\n");
2458 exit(1);
2459 }
2460
2461 void redisAsciiArt(void) {
2462 #include "asciilogo.h"
2463 char *buf = zmalloc(1024*16);
2464 char *mode = "stand alone";
2465
2466 if (server.sentinel_mode) mode = "sentinel";
2467
2468 snprintf(buf,1024*16,ascii_logo,
2469 REDIS_VERSION,
2470 redisGitSHA1(),
2471 strtol(redisGitDirty(),NULL,10) > 0,
2472 (sizeof(long) == 8) ? "64" : "32",
2473 mode, server.port,
2474 (long) getpid()
2475 );
2476 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2477 zfree(buf);
2478 }
2479
2480 static void sigtermHandler(int sig) {
2481 REDIS_NOTUSED(sig);
2482
2483 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2484 server.shutdown_asap = 1;
2485 }
2486
2487 void setupSignalHandlers(void) {
2488 struct sigaction act;
2489
2490 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2491 * Otherwise, sa_handler is used. */
2492 sigemptyset(&act.sa_mask);
2493 act.sa_flags = 0;
2494 act.sa_handler = sigtermHandler;
2495 sigaction(SIGTERM, &act, NULL);
2496
2497 #ifdef HAVE_BACKTRACE
2498 sigemptyset(&act.sa_mask);
2499 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2500 act.sa_sigaction = sigsegvHandler;
2501 sigaction(SIGSEGV, &act, NULL);
2502 sigaction(SIGBUS, &act, NULL);
2503 sigaction(SIGFPE, &act, NULL);
2504 sigaction(SIGILL, &act, NULL);
2505 #endif
2506 return;
2507 }
2508
2509 void memtest(size_t megabytes, int passes);
2510
2511 /* Returns 1 if there is --sentinel among the arguments or if
2512 * argv[0] is exactly "redis-sentinel". */
2513 int checkForSentinelMode(int argc, char **argv) {
2514 int j;
2515
2516 if (strstr(argv[0],"redis-sentinel") != NULL) return 1;
2517 for (j = 1; j < argc; j++)
2518 if (!strcmp(argv[j],"--sentinel")) return 1;
2519 return 0;
2520 }
2521
2522 /* Function called at startup to load RDB or AOF file in memory. */
2523 void loadDataFromDisk(void) {
2524 long long start = ustime();
2525 if (server.aof_state == REDIS_AOF_ON) {
2526 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2527 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2528 } else {
2529 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2530 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2531 (float)(ustime()-start)/1000000);
2532 } else if (errno != ENOENT) {
2533 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2534 exit(1);
2535 }
2536 }
2537 }
2538
2539 void redisOutOfMemoryHandler(size_t allocation_size) {
2540 redisLog(REDIS_WARNING,"Out Of Memory allocating %zu bytes!",
2541 allocation_size);
2542 redisPanic("OOM");
2543 }
2544
2545 int main(int argc, char **argv) {
2546 struct timeval tv;
2547
2548 /* We need to initialize our libraries, and the server configuration. */
2549 zmalloc_enable_thread_safeness();
2550 zmalloc_set_oom_handler(redisOutOfMemoryHandler);
2551 srand(time(NULL)^getpid());
2552 gettimeofday(&tv,NULL);
2553 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2554 server.sentinel_mode = checkForSentinelMode(argc,argv);
2555 initServerConfig();
2556
2557 /* We need to init sentinel right now as parsing the configuration file
2558 * in sentinel mode will have the effect of populating the sentinel
2559 * data structures with master nodes to monitor. */
2560 if (server.sentinel_mode) {
2561 initSentinelConfig();
2562 initSentinel();
2563 }
2564
2565 if (argc >= 2) {
2566 int j = 1; /* First option to parse in argv[] */
2567 sds options = sdsempty();
2568 char *configfile = NULL;
2569
2570 /* Handle special options --help and --version */
2571 if (strcmp(argv[1], "-v") == 0 ||
2572 strcmp(argv[1], "--version") == 0) version();
2573 if (strcmp(argv[1], "--help") == 0 ||
2574 strcmp(argv[1], "-h") == 0) usage();
2575 if (strcmp(argv[1], "--test-memory") == 0) {
2576 if (argc == 3) {
2577 memtest(atoi(argv[2]),50);
2578 exit(0);
2579 } else {
2580 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2581 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2582 exit(1);
2583 }
2584 }
2585
2586 /* First argument is the config file name? */
2587 if (argv[j][0] != '-' || argv[j][1] != '-')
2588 configfile = argv[j++];
2589 /* All the other options are parsed and conceptually appended to the
2590 * configuration file. For instance --port 6380 will generate the
2591 * string "port 6380\n" to be parsed after the actual file name
2592 * is parsed, if any. */
2593 while(j != argc) {
2594 if (argv[j][0] == '-' && argv[j][1] == '-') {
2595 /* Option name */
2596 if (sdslen(options)) options = sdscat(options,"\n");
2597 options = sdscat(options,argv[j]+2);
2598 options = sdscat(options," ");
2599 } else {
2600 /* Option argument */
2601 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2602 options = sdscat(options," ");
2603 }
2604 j++;
2605 }
2606 resetServerSaveParams();
2607 loadServerConfig(configfile,options);
2608 sdsfree(options);
2609 } else {
2610 redisLog(REDIS_WARNING, "Warning: no config file specified, using the default config. In order to specify a config file use %s /path/to/%s.conf", argv[0], server.sentinel_mode ? "sentinel" : "redis");
2611 }
2612 if (server.daemonize) daemonize();
2613 initServer();
2614 if (server.daemonize) createPidFile();
2615 redisAsciiArt();
2616
2617 if (!server.sentinel_mode) {
2618 /* Things only needed when not running in Sentinel mode. */
2619 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2620 #ifdef __linux__
2621 linuxOvercommitMemoryWarning();
2622 #endif
2623 loadDataFromDisk();
2624 if (server.ipfd > 0)
2625 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2626 if (server.sofd > 0)
2627 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2628 }
2629
2630 /* Warning the user about suspicious maxmemory setting. */
2631 if (server.maxmemory > 0 && server.maxmemory < 1024*1024) {
2632 redisLog(REDIS_WARNING,"WARNING: You specified a maxmemory value that is less than 1MB (current value is %llu bytes). Are you sure this is what you really want?", server.maxmemory);
2633 }
2634
2635 aeSetBeforeSleepProc(server.el,beforeSleep);
2636 aeMain(server.el);
2637 aeDeleteEventLoop(server.el);
2638 return 0;
2639 }
2640
2641 /* The End */