]> git.saurik.com Git - redis.git/blob - src/redis.c
Added the SRANDMEMBER key <count> variant.
[redis.git] / src / redis.c
1 /*
2 * Copyright (c) 2009-2010, Salvatore Sanfilippo <antirez at gmail dot com>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are met:
7 *
8 * * Redistributions of source code must retain the above copyright notice,
9 * this list of conditions and the following disclaimer.
10 * * Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * * Neither the name of Redis nor the names of its contributors may be used
14 * to endorse or promote products derived from this software without
15 * specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27 * POSSIBILITY OF SUCH DAMAGE.
28 */
29
30 #include "redis.h"
31 #include "slowlog.h"
32 #include "bio.h"
33
34 #include <time.h>
35 #include <signal.h>
36 #include <sys/wait.h>
37 #include <errno.h>
38 #include <assert.h>
39 #include <ctype.h>
40 #include <stdarg.h>
41 #include <arpa/inet.h>
42 #include <sys/stat.h>
43 #include <fcntl.h>
44 #include <sys/time.h>
45 #include <sys/resource.h>
46 #include <sys/uio.h>
47 #include <limits.h>
48 #include <float.h>
49 #include <math.h>
50 #include <sys/resource.h>
51 #include <sys/utsname.h>
52
53 /* Our shared "common" objects */
54
55 struct sharedObjectsStruct shared;
56
57 /* Global vars that are actually used as constants. The following double
58 * values are used for double on-disk serialization, and are initialized
59 * at runtime to avoid strange compiler optimizations. */
60
61 double R_Zero, R_PosInf, R_NegInf, R_Nan;
62
63 /*================================= Globals ================================= */
64
65 /* Global vars */
66 struct redisServer server; /* server global state */
67 struct redisCommand *commandTable;
68
69 /* Our command table.
70 *
71 * Every entry is composed of the following fields:
72 *
73 * name: a string representing the command name.
74 * function: pointer to the C function implementing the command.
75 * arity: number of arguments, it is possible to use -N to say >= N
76 * sflags: command flags as string. See below for a table of flags.
77 * flags: flags as bitmask. Computed by Redis using the 'sflags' field.
78 * get_keys_proc: an optional function to get key arguments from a command.
79 * This is only used when the following three fields are not
80 * enough to specify what arguments are keys.
81 * first_key_index: first argument that is a key
82 * last_key_index: last argument that is a key
83 * key_step: step to get all the keys from first to last argument. For instance
84 * in MSET the step is two since arguments are key,val,key,val,...
85 * microseconds: microseconds of total execution time for this command.
86 * calls: total number of calls of this command.
87 *
88 * The flags, microseconds and calls fields are computed by Redis and should
89 * always be set to zero.
90 *
91 * Command flags are expressed using strings where every character represents
92 * a flag. Later the populateCommandTable() function will take care of
93 * populating the real 'flags' field using this characters.
94 *
95 * This is the meaning of the flags:
96 *
97 * w: write command (may modify the key space).
98 * r: read command (will never modify the key space).
99 * m: may increase memory usage once called. Don't allow if out of memory.
100 * a: admin command, like SAVE or SHUTDOWN.
101 * p: Pub/Sub related command.
102 * f: force replication of this command, regarless of server.dirty.
103 * s: command not allowed in scripts.
104 * R: random command. Command is not deterministic, that is, the same command
105 * with the same arguments, with the same key space, may have different
106 * results. For instance SPOP and RANDOMKEY are two random commands.
107 * S: Sort command output array if called from script, so that the output
108 * is deterministic.
109 * l: Allow command while loading the database.
110 * t: Allow command while a slave has stale data but is not allowed to
111 * server this data. Normally no command is accepted in this condition
112 * but just a few.
113 */
114 struct redisCommand redisCommandTable[] = {
115 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
116 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
117 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
118 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
119 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
120 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
121 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
122 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
123 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
124 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
125 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
126 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
127 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
128 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
129 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
130 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
131 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
132 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
133 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
134 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
135 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
136 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
137 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
138 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
139 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
140 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
141 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
142 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
143 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
144 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
145 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
146 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
147 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
148 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
149 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
150 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
151 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
152 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
153 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
154 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
155 {"srandmember",srandmemberCommand,-2,"rR",0,NULL,1,1,1,0,0},
156 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
157 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
158 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
159 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
160 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
161 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
162 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
163 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
164 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
165 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
166 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
167 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
168 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
169 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
170 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
171 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
172 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
173 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
174 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
175 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
176 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
177 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
178 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
179 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
180 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
182 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
183 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
184 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
185 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
186 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
187 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
188 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
189 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
190 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
191 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
192 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
193 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
194 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
195 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
196 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
197 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
198 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
199 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
200 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
201 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
202 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
203 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
204 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
205 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
206 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
207 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
208 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
209 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
210 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
211 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
212 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
213 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
214 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
215 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
216 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
217 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
218 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
219 {"exec",execCommand,1,"s",0,NULL,0,0,0,0,0},
220 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
221 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
222 {"replconf",replconfCommand,-1,"ars",0,NULL,0,0,0,0,0},
223 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
224 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
225 {"sort",sortCommand,-2,"wm",0,NULL,1,1,1,0,0},
226 {"info",infoCommand,-1,"rlt",0,NULL,0,0,0,0,0},
227 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
228 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
229 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
230 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
231 {"slaveof",slaveofCommand,3,"ast",0,NULL,0,0,0,0,0},
232 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
233 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
234 {"subscribe",subscribeCommand,-2,"rpslt",0,NULL,0,0,0,0,0},
235 {"unsubscribe",unsubscribeCommand,-1,"rpslt",0,NULL,0,0,0,0,0},
236 {"psubscribe",psubscribeCommand,-2,"rpslt",0,NULL,0,0,0,0,0},
237 {"punsubscribe",punsubscribeCommand,-1,"rpslt",0,NULL,0,0,0,0,0},
238 {"publish",publishCommand,3,"pflt",0,NULL,0,0,0,0,0},
239 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
240 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
241 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
242 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
243 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
244 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
245 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
246 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
247 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
248 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
249 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
250 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0},
251 {"bitop",bitopCommand,-4,"wm",0,NULL,2,-1,1,0,0},
252 {"bitcount",bitcountCommand,-2,"r",0,NULL,1,1,1,0,0}
253 };
254
255 /*============================ Utility functions ============================ */
256
257 /* Low level logging. To use only for very big messages, otherwise
258 * redisLog() is to prefer. */
259 void redisLogRaw(int level, const char *msg) {
260 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
261 const char *c = ".-*#";
262 FILE *fp;
263 char buf[64];
264 int rawmode = (level & REDIS_LOG_RAW);
265
266 level &= 0xff; /* clear flags */
267 if (level < server.verbosity) return;
268
269 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
270 if (!fp) return;
271
272 if (rawmode) {
273 fprintf(fp,"%s",msg);
274 } else {
275 int off;
276 struct timeval tv;
277
278 gettimeofday(&tv,NULL);
279 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
280 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
281 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
282 }
283 fflush(fp);
284
285 if (server.logfile) fclose(fp);
286
287 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
288 }
289
290 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
291 * is used across the code. The raw version is only used in order to dump
292 * the INFO output on crash. */
293 void redisLog(int level, const char *fmt, ...) {
294 va_list ap;
295 char msg[REDIS_MAX_LOGMSG_LEN];
296
297 if ((level&0xff) < server.verbosity) return;
298
299 va_start(ap, fmt);
300 vsnprintf(msg, sizeof(msg), fmt, ap);
301 va_end(ap);
302
303 redisLogRaw(level,msg);
304 }
305
306 /* Log a fixed message without printf-alike capabilities, in a way that is
307 * safe to call from a signal handler.
308 *
309 * We actually use this only for signals that are not fatal from the point
310 * of view of Redis. Signals that are going to kill the server anyway and
311 * where we need printf-alike features are served by redisLog(). */
312 void redisLogFromHandler(int level, const char *msg) {
313 int fd;
314 char buf[64];
315
316 if ((level&0xff) < server.verbosity ||
317 (server.logfile == NULL && server.daemonize)) return;
318 fd = server.logfile ?
319 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
320 STDOUT_FILENO;
321 if (fd == -1) return;
322 ll2string(buf,sizeof(buf),getpid());
323 if (write(fd,"[",1) == -1) goto err;
324 if (write(fd,buf,strlen(buf)) == -1) goto err;
325 if (write(fd," | signal handler] (",20) == -1) goto err;
326 ll2string(buf,sizeof(buf),time(NULL));
327 if (write(fd,buf,strlen(buf)) == -1) goto err;
328 if (write(fd,") ",2) == -1) goto err;
329 if (write(fd,msg,strlen(msg)) == -1) goto err;
330 if (write(fd,"\n",1) == -1) goto err;
331 err:
332 if (server.logfile) close(fd);
333 }
334
335 /* Return the UNIX time in microseconds */
336 long long ustime(void) {
337 struct timeval tv;
338 long long ust;
339
340 gettimeofday(&tv, NULL);
341 ust = ((long long)tv.tv_sec)*1000000;
342 ust += tv.tv_usec;
343 return ust;
344 }
345
346 /* Return the UNIX time in milliseconds */
347 long long mstime(void) {
348 return ustime()/1000;
349 }
350
351 /* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
352 * exit(), because the latter may interact with the same file objects used by
353 * the parent process. However if we are testing the coverage normal exit() is
354 * used in order to obtain the right coverage information. */
355 void exitFromChild(int retcode) {
356 #ifdef COVERAGE_TEST
357 exit(retcode);
358 #else
359 _exit(retcode);
360 #endif
361 }
362
363 /*====================== Hash table type implementation ==================== */
364
365 /* This is an hash table type that uses the SDS dynamic strings libary as
366 * keys and radis objects as values (objects can hold SDS strings,
367 * lists, sets). */
368
369 void dictVanillaFree(void *privdata, void *val)
370 {
371 DICT_NOTUSED(privdata);
372 zfree(val);
373 }
374
375 void dictListDestructor(void *privdata, void *val)
376 {
377 DICT_NOTUSED(privdata);
378 listRelease((list*)val);
379 }
380
381 int dictSdsKeyCompare(void *privdata, const void *key1,
382 const void *key2)
383 {
384 int l1,l2;
385 DICT_NOTUSED(privdata);
386
387 l1 = sdslen((sds)key1);
388 l2 = sdslen((sds)key2);
389 if (l1 != l2) return 0;
390 return memcmp(key1, key2, l1) == 0;
391 }
392
393 /* A case insensitive version used for the command lookup table. */
394 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
395 const void *key2)
396 {
397 DICT_NOTUSED(privdata);
398
399 return strcasecmp(key1, key2) == 0;
400 }
401
402 void dictRedisObjectDestructor(void *privdata, void *val)
403 {
404 DICT_NOTUSED(privdata);
405
406 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
407 decrRefCount(val);
408 }
409
410 void dictSdsDestructor(void *privdata, void *val)
411 {
412 DICT_NOTUSED(privdata);
413
414 sdsfree(val);
415 }
416
417 int dictObjKeyCompare(void *privdata, const void *key1,
418 const void *key2)
419 {
420 const robj *o1 = key1, *o2 = key2;
421 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
422 }
423
424 unsigned int dictObjHash(const void *key) {
425 const robj *o = key;
426 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
427 }
428
429 unsigned int dictSdsHash(const void *key) {
430 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
431 }
432
433 unsigned int dictSdsCaseHash(const void *key) {
434 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
435 }
436
437 int dictEncObjKeyCompare(void *privdata, const void *key1,
438 const void *key2)
439 {
440 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
441 int cmp;
442
443 if (o1->encoding == REDIS_ENCODING_INT &&
444 o2->encoding == REDIS_ENCODING_INT)
445 return o1->ptr == o2->ptr;
446
447 o1 = getDecodedObject(o1);
448 o2 = getDecodedObject(o2);
449 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
450 decrRefCount(o1);
451 decrRefCount(o2);
452 return cmp;
453 }
454
455 unsigned int dictEncObjHash(const void *key) {
456 robj *o = (robj*) key;
457
458 if (o->encoding == REDIS_ENCODING_RAW) {
459 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
460 } else {
461 if (o->encoding == REDIS_ENCODING_INT) {
462 char buf[32];
463 int len;
464
465 len = ll2string(buf,32,(long)o->ptr);
466 return dictGenHashFunction((unsigned char*)buf, len);
467 } else {
468 unsigned int hash;
469
470 o = getDecodedObject(o);
471 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
472 decrRefCount(o);
473 return hash;
474 }
475 }
476 }
477
478 /* Sets type hash table */
479 dictType setDictType = {
480 dictEncObjHash, /* hash function */
481 NULL, /* key dup */
482 NULL, /* val dup */
483 dictEncObjKeyCompare, /* key compare */
484 dictRedisObjectDestructor, /* key destructor */
485 NULL /* val destructor */
486 };
487
488 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
489 dictType zsetDictType = {
490 dictEncObjHash, /* hash function */
491 NULL, /* key dup */
492 NULL, /* val dup */
493 dictEncObjKeyCompare, /* key compare */
494 dictRedisObjectDestructor, /* key destructor */
495 NULL /* val destructor */
496 };
497
498 /* Db->dict, keys are sds strings, vals are Redis objects. */
499 dictType dbDictType = {
500 dictSdsHash, /* hash function */
501 NULL, /* key dup */
502 NULL, /* val dup */
503 dictSdsKeyCompare, /* key compare */
504 dictSdsDestructor, /* key destructor */
505 dictRedisObjectDestructor /* val destructor */
506 };
507
508 /* Db->expires */
509 dictType keyptrDictType = {
510 dictSdsHash, /* hash function */
511 NULL, /* key dup */
512 NULL, /* val dup */
513 dictSdsKeyCompare, /* key compare */
514 NULL, /* key destructor */
515 NULL /* val destructor */
516 };
517
518 /* Command table. sds string -> command struct pointer. */
519 dictType commandTableDictType = {
520 dictSdsCaseHash, /* hash function */
521 NULL, /* key dup */
522 NULL, /* val dup */
523 dictSdsKeyCaseCompare, /* key compare */
524 dictSdsDestructor, /* key destructor */
525 NULL /* val destructor */
526 };
527
528 /* Hash type hash table (note that small hashes are represented with zimpaps) */
529 dictType hashDictType = {
530 dictEncObjHash, /* hash function */
531 NULL, /* key dup */
532 NULL, /* val dup */
533 dictEncObjKeyCompare, /* key compare */
534 dictRedisObjectDestructor, /* key destructor */
535 dictRedisObjectDestructor /* val destructor */
536 };
537
538 /* Keylist hash table type has unencoded redis objects as keys and
539 * lists as values. It's used for blocking operations (BLPOP) and to
540 * map swapped keys to a list of clients waiting for this keys to be loaded. */
541 dictType keylistDictType = {
542 dictObjHash, /* hash function */
543 NULL, /* key dup */
544 NULL, /* val dup */
545 dictObjKeyCompare, /* key compare */
546 dictRedisObjectDestructor, /* key destructor */
547 dictListDestructor /* val destructor */
548 };
549
550 int htNeedsResize(dict *dict) {
551 long long size, used;
552
553 size = dictSlots(dict);
554 used = dictSize(dict);
555 return (size && used && size > DICT_HT_INITIAL_SIZE &&
556 (used*100/size < REDIS_HT_MINFILL));
557 }
558
559 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
560 * we resize the hash table to save memory */
561 void tryResizeHashTables(void) {
562 int j;
563
564 for (j = 0; j < server.dbnum; j++) {
565 if (htNeedsResize(server.db[j].dict))
566 dictResize(server.db[j].dict);
567 if (htNeedsResize(server.db[j].expires))
568 dictResize(server.db[j].expires);
569 }
570 }
571
572 /* Our hash table implementation performs rehashing incrementally while
573 * we write/read from the hash table. Still if the server is idle, the hash
574 * table will use two tables for a long time. So we try to use 1 millisecond
575 * of CPU time at every serverCron() loop in order to rehash some key. */
576 void incrementallyRehash(void) {
577 int j;
578
579 for (j = 0; j < server.dbnum; j++) {
580 /* Keys dictionary */
581 if (dictIsRehashing(server.db[j].dict)) {
582 dictRehashMilliseconds(server.db[j].dict,1);
583 break; /* already used our millisecond for this loop... */
584 }
585 /* Expires */
586 if (dictIsRehashing(server.db[j].expires)) {
587 dictRehashMilliseconds(server.db[j].expires,1);
588 break; /* already used our millisecond for this loop... */
589 }
590 }
591 }
592
593 /* This function is called once a background process of some kind terminates,
594 * as we want to avoid resizing the hash tables when there is a child in order
595 * to play well with copy-on-write (otherwise when a resize happens lots of
596 * memory pages are copied). The goal of this function is to update the ability
597 * for dict.c to resize the hash tables accordingly to the fact we have o not
598 * running childs. */
599 void updateDictResizePolicy(void) {
600 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
601 dictEnableResize();
602 else
603 dictDisableResize();
604 }
605
606 /* ======================= Cron: called every 100 ms ======================== */
607
608 /* Try to expire a few timed out keys. The algorithm used is adaptive and
609 * will use few CPU cycles if there are few expiring keys, otherwise
610 * it will get more aggressive to avoid that too much memory is used by
611 * keys that can be removed from the keyspace. */
612 void activeExpireCycle(void) {
613 int j, iteration = 0;
614 long long start = ustime(), timelimit;
615
616 /* We can use at max REDIS_EXPIRELOOKUPS_TIME_PERC percentage of CPU time
617 * per iteration. Since this function gets called with a frequency of
618 * REDIS_HZ times per second, the following is the max amount of
619 * microseconds we can spend in this function. */
620 timelimit = 1000000*REDIS_EXPIRELOOKUPS_TIME_PERC/REDIS_HZ/100;
621 if (timelimit <= 0) timelimit = 1;
622
623 for (j = 0; j < server.dbnum; j++) {
624 int expired;
625 redisDb *db = server.db+j;
626
627 /* Continue to expire if at the end of the cycle more than 25%
628 * of the keys were expired. */
629 do {
630 unsigned long num = dictSize(db->expires);
631 unsigned long slots = dictSlots(db->expires);
632 long long now = mstime();
633
634 /* When there are less than 1% filled slots getting random
635 * keys is expensive, so stop here waiting for better times...
636 * The dictionary will be resized asap. */
637 if (num && slots > DICT_HT_INITIAL_SIZE &&
638 (num*100/slots < 1)) break;
639
640 /* The main collection cycle. Sample random keys among keys
641 * with an expire set, checking for expired ones. */
642 expired = 0;
643 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
644 num = REDIS_EXPIRELOOKUPS_PER_CRON;
645 while (num--) {
646 dictEntry *de;
647 long long t;
648
649 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
650 t = dictGetSignedIntegerVal(de);
651 if (now > t) {
652 sds key = dictGetKey(de);
653 robj *keyobj = createStringObject(key,sdslen(key));
654
655 propagateExpire(db,keyobj);
656 dbDelete(db,keyobj);
657 decrRefCount(keyobj);
658 expired++;
659 server.stat_expiredkeys++;
660 }
661 }
662 /* We can't block forever here even if there are many keys to
663 * expire. So after a given amount of milliseconds return to the
664 * caller waiting for the other active expire cycle. */
665 iteration++;
666 if ((iteration & 0xf) == 0 && /* check once every 16 cycles. */
667 (ustime()-start) > timelimit) return;
668 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
669 }
670 }
671
672 void updateLRUClock(void) {
673 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
674 REDIS_LRU_CLOCK_MAX;
675 }
676
677
678 /* Add a sample to the operations per second array of samples. */
679 void trackOperationsPerSecond(void) {
680 long long t = mstime() - server.ops_sec_last_sample_time;
681 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
682 long long ops_sec;
683
684 ops_sec = t > 0 ? (ops*1000/t) : 0;
685
686 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
687 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
688 server.ops_sec_last_sample_time = mstime();
689 server.ops_sec_last_sample_ops = server.stat_numcommands;
690 }
691
692 /* Return the mean of all the samples. */
693 long long getOperationsPerSecond(void) {
694 int j;
695 long long sum = 0;
696
697 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
698 sum += server.ops_sec_samples[j];
699 return sum / REDIS_OPS_SEC_SAMPLES;
700 }
701
702 /* Check for timeouts. Returns non-zero if the client was terminated */
703 int clientsCronHandleTimeout(redisClient *c) {
704 time_t now = server.unixtime;
705
706 if (server.maxidletime &&
707 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
708 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
709 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
710 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
711 listLength(c->pubsub_patterns) == 0 &&
712 (now - c->lastinteraction > server.maxidletime))
713 {
714 redisLog(REDIS_VERBOSE,"Closing idle client");
715 freeClient(c);
716 return 1;
717 } else if (c->flags & REDIS_BLOCKED) {
718 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
719 addReply(c,shared.nullmultibulk);
720 unblockClientWaitingData(c);
721 }
722 }
723 return 0;
724 }
725
726 /* The client query buffer is an sds.c string that can end with a lot of
727 * free space not used, this function reclaims space if needed.
728 *
729 * The funciton always returns 0 as it never terminates the client. */
730 int clientsCronResizeQueryBuffer(redisClient *c) {
731 size_t querybuf_size = sdsAllocSize(c->querybuf);
732 time_t idletime = server.unixtime - c->lastinteraction;
733
734 /* There are two conditions to resize the query buffer:
735 * 1) Query buffer is > BIG_ARG and too big for latest peak.
736 * 2) Client is inactive and the buffer is bigger than 1k. */
737 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
738 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
739 (querybuf_size > 1024 && idletime > 2))
740 {
741 /* Only resize the query buffer if it is actually wasting space. */
742 if (sdsavail(c->querybuf) > 1024) {
743 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
744 }
745 }
746 /* Reset the peak again to capture the peak memory usage in the next
747 * cycle. */
748 c->querybuf_peak = 0;
749 return 0;
750 }
751
752 void clientsCron(void) {
753 /* Make sure to process at least 1/(REDIS_HZ*10) of clients per call.
754 * Since this function is called REDIS_HZ times per second we are sure that
755 * in the worst case we process all the clients in 10 seconds.
756 * In normal conditions (a reasonable number of clients) we process
757 * all the clients in a shorter time. */
758 int numclients = listLength(server.clients);
759 int iterations = numclients/(REDIS_HZ*10);
760
761 if (iterations < 50)
762 iterations = (numclients < 50) ? numclients : 50;
763 while(listLength(server.clients) && iterations--) {
764 redisClient *c;
765 listNode *head;
766
767 /* Rotate the list, take the current head, process.
768 * This way if the client must be removed from the list it's the
769 * first element and we don't incur into O(N) computation. */
770 listRotate(server.clients);
771 head = listFirst(server.clients);
772 c = listNodeValue(head);
773 /* The following functions do different service checks on the client.
774 * The protocol is that they return non-zero if the client was
775 * terminated. */
776 if (clientsCronHandleTimeout(c)) continue;
777 if (clientsCronResizeQueryBuffer(c)) continue;
778 }
779 }
780
781 /* This is our timer interrupt, called REDIS_HZ times per second.
782 * Here is where we do a number of things that need to be done asynchronously.
783 * For instance:
784 *
785 * - Active expired keys collection (it is also performed in a lazy way on
786 * lookup).
787 * - Software watchdong.
788 * - Update some statistic.
789 * - Incremental rehashing of the DBs hash tables.
790 * - Triggering BGSAVE / AOF rewrite, and handling of terminated children.
791 * - Clients timeout of differnet kinds.
792 * - Replication reconnection.
793 * - Many more...
794 *
795 * Everything directly called here will be called REDIS_HZ times per second,
796 * so in order to throttle execution of things we want to do less frequently
797 * a macro is used: run_with_period(milliseconds) { .... }
798 */
799
800 /* Using the following macro you can run code inside serverCron() with the
801 * specified period, specified in milliseconds.
802 * The actual resolution depends on REDIS_HZ. */
803 #define run_with_period(_ms_) if (!(loops % ((_ms_)/(1000/REDIS_HZ))))
804
805 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
806 int j, loops = server.cronloops;
807 REDIS_NOTUSED(eventLoop);
808 REDIS_NOTUSED(id);
809 REDIS_NOTUSED(clientData);
810
811 /* Software watchdog: deliver the SIGALRM that will reach the signal
812 * handler if we don't return here fast enough. */
813 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
814
815 /* We take a cached value of the unix time in the global state because
816 * with virtual memory and aging there is to store the current time
817 * in objects at every object access, and accuracy is not needed.
818 * To access a global var is faster than calling time(NULL) */
819 server.unixtime = time(NULL);
820
821 run_with_period(100) trackOperationsPerSecond();
822
823 /* We have just 22 bits per object for LRU information.
824 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
825 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
826 *
827 * Note that even if this will wrap after 1.5 years it's not a problem,
828 * everything will still work but just some object will appear younger
829 * to Redis. But for this to happen a given object should never be touched
830 * for 1.5 years.
831 *
832 * Note that you can change the resolution altering the
833 * REDIS_LRU_CLOCK_RESOLUTION define.
834 */
835 updateLRUClock();
836
837 /* Record the max memory used since the server was started. */
838 if (zmalloc_used_memory() > server.stat_peak_memory)
839 server.stat_peak_memory = zmalloc_used_memory();
840
841 /* We received a SIGTERM, shutting down here in a safe way, as it is
842 * not ok doing so inside the signal handler. */
843 if (server.shutdown_asap) {
844 if (prepareForShutdown(0) == REDIS_OK) exit(0);
845 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
846 }
847
848 /* Show some info about non-empty databases */
849 run_with_period(5000) {
850 for (j = 0; j < server.dbnum; j++) {
851 long long size, used, vkeys;
852
853 size = dictSlots(server.db[j].dict);
854 used = dictSize(server.db[j].dict);
855 vkeys = dictSize(server.db[j].expires);
856 if (used || vkeys) {
857 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
858 /* dictPrintStats(server.dict); */
859 }
860 }
861 }
862
863 /* We don't want to resize the hash tables while a bacground saving
864 * is in progress: the saving child is created using fork() that is
865 * implemented with a copy-on-write semantic in most modern systems, so
866 * if we resize the HT while there is the saving child at work actually
867 * a lot of memory movements in the parent will cause a lot of pages
868 * copied. */
869 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
870 tryResizeHashTables();
871 if (server.activerehashing) incrementallyRehash();
872 }
873
874 /* Show information about connected clients */
875 run_with_period(5000) {
876 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
877 listLength(server.clients)-listLength(server.slaves),
878 listLength(server.slaves),
879 zmalloc_used_memory());
880 }
881
882 /* We need to do a few operations on clients asynchronously. */
883 clientsCron();
884
885 /* Start a scheduled AOF rewrite if this was requested by the user while
886 * a BGSAVE was in progress. */
887 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
888 server.aof_rewrite_scheduled)
889 {
890 rewriteAppendOnlyFileBackground();
891 }
892
893 /* Check if a background saving or AOF rewrite in progress terminated. */
894 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
895 int statloc;
896 pid_t pid;
897
898 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
899 int exitcode = WEXITSTATUS(statloc);
900 int bysignal = 0;
901
902 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
903
904 if (pid == server.rdb_child_pid) {
905 backgroundSaveDoneHandler(exitcode,bysignal);
906 } else {
907 backgroundRewriteDoneHandler(exitcode,bysignal);
908 }
909 updateDictResizePolicy();
910 }
911 } else {
912 /* If there is not a background saving/rewrite in progress check if
913 * we have to save/rewrite now */
914 for (j = 0; j < server.saveparamslen; j++) {
915 struct saveparam *sp = server.saveparams+j;
916
917 if (server.dirty >= sp->changes &&
918 server.unixtime-server.lastsave > sp->seconds) {
919 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
920 sp->changes, sp->seconds);
921 rdbSaveBackground(server.rdb_filename);
922 break;
923 }
924 }
925
926 /* Trigger an AOF rewrite if needed */
927 if (server.rdb_child_pid == -1 &&
928 server.aof_child_pid == -1 &&
929 server.aof_rewrite_perc &&
930 server.aof_current_size > server.aof_rewrite_min_size)
931 {
932 long long base = server.aof_rewrite_base_size ?
933 server.aof_rewrite_base_size : 1;
934 long long growth = (server.aof_current_size*100/base) - 100;
935 if (growth >= server.aof_rewrite_perc) {
936 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
937 rewriteAppendOnlyFileBackground();
938 }
939 }
940 }
941
942
943 /* If we postponed an AOF buffer flush, let's try to do it every time the
944 * cron function is called. */
945 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
946
947 /* Expire a few keys per cycle, only if this is a master.
948 * On slaves we wait for DEL operations synthesized by the master
949 * in order to guarantee a strict consistency. */
950 if (server.masterhost == NULL) activeExpireCycle();
951
952 /* Close clients that need to be closed asynchronous */
953 freeClientsInAsyncFreeQueue();
954
955 /* Replication cron function -- used to reconnect to master and
956 * to detect transfer failures. */
957 run_with_period(1000) replicationCron();
958
959 server.cronloops++;
960 return 1000/REDIS_HZ;
961 }
962
963 /* This function gets called every time Redis is entering the
964 * main loop of the event driven library, that is, before to sleep
965 * for ready file descriptors. */
966 void beforeSleep(struct aeEventLoop *eventLoop) {
967 REDIS_NOTUSED(eventLoop);
968 listNode *ln;
969 redisClient *c;
970
971 /* Try to process pending commands for clients that were just unblocked. */
972 while (listLength(server.unblocked_clients)) {
973 ln = listFirst(server.unblocked_clients);
974 redisAssert(ln != NULL);
975 c = ln->value;
976 listDelNode(server.unblocked_clients,ln);
977 c->flags &= ~REDIS_UNBLOCKED;
978
979 /* Process remaining data in the input buffer. */
980 if (c->querybuf && sdslen(c->querybuf) > 0) {
981 server.current_client = c;
982 processInputBuffer(c);
983 server.current_client = NULL;
984 }
985 }
986
987 /* Write the AOF buffer on disk */
988 flushAppendOnlyFile(0);
989 }
990
991 /* =========================== Server initialization ======================== */
992
993 void createSharedObjects(void) {
994 int j;
995
996 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
997 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
998 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
999 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
1000 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
1001 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
1002 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
1003 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
1004 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
1005 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
1006 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
1007 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
1008 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
1009 "-ERR Operation against a key holding the wrong kind of value\r\n"));
1010 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
1011 "-ERR no such key\r\n"));
1012 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
1013 "-ERR syntax error\r\n"));
1014 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
1015 "-ERR source and destination objects are the same\r\n"));
1016 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
1017 "-ERR index out of range\r\n"));
1018 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
1019 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
1020 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
1021 "-LOADING Redis is loading the dataset in memory\r\n"));
1022 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
1023 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
1024 shared.masterdownerr = createObject(REDIS_STRING,sdsnew(
1025 "-MASTERDOWN Link with MASTER is down and slave-serve-stale-data is set to 'no'.\r\n"));
1026 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
1027 "-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"));
1028 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
1029 "-READONLY You can't write against a read only slave.\r\n"));
1030 shared.oomerr = createObject(REDIS_STRING,sdsnew(
1031 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
1032 shared.space = createObject(REDIS_STRING,sdsnew(" "));
1033 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1034 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1035
1036 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1037 shared.select[j] = createObject(REDIS_STRING,
1038 sdscatprintf(sdsempty(),"select %d\r\n", j));
1039 }
1040 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1041 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1042 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1043 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1044 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1045 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1046 shared.del = createStringObject("DEL",3);
1047 shared.rpop = createStringObject("RPOP",4);
1048 shared.lpop = createStringObject("LPOP",4);
1049 shared.lpush = createStringObject("LPUSH",5);
1050 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1051 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1052 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1053 }
1054 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1055 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1056 sdscatprintf(sdsempty(),"*%d\r\n",j));
1057 shared.bulkhdr[j] = createObject(REDIS_STRING,
1058 sdscatprintf(sdsempty(),"$%d\r\n",j));
1059 }
1060 }
1061
1062 void initServerConfig() {
1063 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1064 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1065 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1066 server.port = REDIS_SERVERPORT;
1067 server.bindaddr = NULL;
1068 server.unixsocket = NULL;
1069 server.unixsocketperm = 0;
1070 server.ipfd = -1;
1071 server.sofd = -1;
1072 server.dbnum = REDIS_DEFAULT_DBNUM;
1073 server.verbosity = REDIS_NOTICE;
1074 server.maxidletime = REDIS_MAXIDLETIME;
1075 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1076 server.saveparams = NULL;
1077 server.loading = 0;
1078 server.logfile = NULL; /* NULL = log on standard output */
1079 server.syslog_enabled = 0;
1080 server.syslog_ident = zstrdup("redis");
1081 server.syslog_facility = LOG_LOCAL0;
1082 server.daemonize = 0;
1083 server.aof_state = REDIS_AOF_OFF;
1084 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1085 server.aof_no_fsync_on_rewrite = 0;
1086 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1087 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1088 server.aof_rewrite_base_size = 0;
1089 server.aof_rewrite_scheduled = 0;
1090 server.aof_last_fsync = time(NULL);
1091 server.aof_rewrite_time_last = -1;
1092 server.aof_rewrite_time_start = -1;
1093 server.aof_lastbgrewrite_status = REDIS_OK;
1094 server.aof_delayed_fsync = 0;
1095 server.aof_fd = -1;
1096 server.aof_selected_db = -1; /* Make sure the first time will not match */
1097 server.aof_flush_postponed_start = 0;
1098 server.pidfile = zstrdup("/var/run/redis.pid");
1099 server.rdb_filename = zstrdup("dump.rdb");
1100 server.aof_filename = zstrdup("appendonly.aof");
1101 server.requirepass = NULL;
1102 server.rdb_compression = 1;
1103 server.rdb_checksum = 1;
1104 server.activerehashing = 1;
1105 server.maxclients = REDIS_MAX_CLIENTS;
1106 server.bpop_blocked_clients = 0;
1107 server.maxmemory = 0;
1108 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1109 server.maxmemory_samples = 3;
1110 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1111 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1112 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1113 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1114 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1115 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1116 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1117 server.shutdown_asap = 0;
1118 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1119 server.repl_timeout = REDIS_REPL_TIMEOUT;
1120 server.lua_caller = NULL;
1121 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1122 server.lua_client = NULL;
1123 server.lua_timedout = 0;
1124
1125 updateLRUClock();
1126 resetServerSaveParams();
1127
1128 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1129 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1130 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1131 /* Replication related */
1132 server.masterauth = NULL;
1133 server.masterhost = NULL;
1134 server.masterport = 6379;
1135 server.master = NULL;
1136 server.repl_state = REDIS_REPL_NONE;
1137 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1138 server.repl_serve_stale_data = 1;
1139 server.repl_slave_ro = 1;
1140 server.repl_down_since = time(NULL);
1141 server.slave_priority = REDIS_DEFAULT_SLAVE_PRIORITY;
1142
1143 /* Client output buffer limits */
1144 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1145 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1146 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1147 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1148 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1149 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1150 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1151 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1152 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1153
1154 /* Double constants initialization */
1155 R_Zero = 0.0;
1156 R_PosInf = 1.0/R_Zero;
1157 R_NegInf = -1.0/R_Zero;
1158 R_Nan = R_Zero/R_Zero;
1159
1160 /* Command table -- we intiialize it here as it is part of the
1161 * initial configuration, since command names may be changed via
1162 * redis.conf using the rename-command directive. */
1163 server.commands = dictCreate(&commandTableDictType,NULL);
1164 populateCommandTable();
1165 server.delCommand = lookupCommandByCString("del");
1166 server.multiCommand = lookupCommandByCString("multi");
1167 server.lpushCommand = lookupCommandByCString("lpush");
1168 server.lpopCommand = lookupCommandByCString("lpop");
1169 server.rpopCommand = lookupCommandByCString("rpop");
1170
1171 /* Slow log */
1172 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1173 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1174
1175 /* Debugging */
1176 server.assert_failed = "<no assertion failed>";
1177 server.assert_file = "<no file>";
1178 server.assert_line = 0;
1179 server.bug_report_start = 0;
1180 server.watchdog_period = 0;
1181 }
1182
1183 /* This function will try to raise the max number of open files accordingly to
1184 * the configured max number of clients. It will also account for 32 additional
1185 * file descriptors as we need a few more for persistence, listening
1186 * sockets, log files and so forth.
1187 *
1188 * If it will not be possible to set the limit accordingly to the configured
1189 * max number of clients, the function will do the reverse setting
1190 * server.maxclients to the value that we can actually handle. */
1191 void adjustOpenFilesLimit(void) {
1192 rlim_t maxfiles = server.maxclients+32;
1193 struct rlimit limit;
1194
1195 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1196 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1197 strerror(errno));
1198 server.maxclients = 1024-32;
1199 } else {
1200 rlim_t oldlimit = limit.rlim_cur;
1201
1202 /* Set the max number of files if the current limit is not enough
1203 * for our needs. */
1204 if (oldlimit < maxfiles) {
1205 rlim_t f;
1206
1207 f = maxfiles;
1208 while(f > oldlimit) {
1209 limit.rlim_cur = f;
1210 limit.rlim_max = f;
1211 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1212 f -= 128;
1213 }
1214 if (f < oldlimit) f = oldlimit;
1215 if (f != maxfiles) {
1216 server.maxclients = f-32;
1217 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1218 (int) maxfiles, strerror(errno), (int) server.maxclients);
1219 } else {
1220 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1221 (int) maxfiles);
1222 }
1223 }
1224 }
1225 }
1226
1227 void initServer() {
1228 int j;
1229
1230 signal(SIGHUP, SIG_IGN);
1231 signal(SIGPIPE, SIG_IGN);
1232 setupSignalHandlers();
1233
1234 if (server.syslog_enabled) {
1235 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1236 server.syslog_facility);
1237 }
1238
1239 server.current_client = NULL;
1240 server.clients = listCreate();
1241 server.clients_to_close = listCreate();
1242 server.slaves = listCreate();
1243 server.monitors = listCreate();
1244 server.unblocked_clients = listCreate();
1245 server.ready_keys = listCreate();
1246
1247 createSharedObjects();
1248 adjustOpenFilesLimit();
1249 server.el = aeCreateEventLoop(server.maxclients+1024);
1250 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1251
1252 if (server.port != 0) {
1253 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1254 if (server.ipfd == ANET_ERR) {
1255 redisLog(REDIS_WARNING, "Opening port %d: %s",
1256 server.port, server.neterr);
1257 exit(1);
1258 }
1259 }
1260 if (server.unixsocket != NULL) {
1261 unlink(server.unixsocket); /* don't care if this fails */
1262 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1263 if (server.sofd == ANET_ERR) {
1264 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1265 exit(1);
1266 }
1267 }
1268 if (server.ipfd < 0 && server.sofd < 0) {
1269 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1270 exit(1);
1271 }
1272 for (j = 0; j < server.dbnum; j++) {
1273 server.db[j].dict = dictCreate(&dbDictType,NULL);
1274 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1275 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1276 server.db[j].ready_keys = dictCreate(&setDictType,NULL);
1277 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1278 server.db[j].id = j;
1279 }
1280 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1281 server.pubsub_patterns = listCreate();
1282 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1283 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1284 server.cronloops = 0;
1285 server.rdb_child_pid = -1;
1286 server.aof_child_pid = -1;
1287 aofRewriteBufferReset();
1288 server.aof_buf = sdsempty();
1289 server.lastsave = time(NULL);
1290 server.rdb_save_time_last = -1;
1291 server.rdb_save_time_start = -1;
1292 server.dirty = 0;
1293 server.stat_numcommands = 0;
1294 server.stat_numconnections = 0;
1295 server.stat_expiredkeys = 0;
1296 server.stat_evictedkeys = 0;
1297 server.stat_starttime = time(NULL);
1298 server.stat_keyspace_misses = 0;
1299 server.stat_keyspace_hits = 0;
1300 server.stat_peak_memory = 0;
1301 server.stat_fork_time = 0;
1302 server.stat_rejected_conn = 0;
1303 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1304 server.ops_sec_idx = 0;
1305 server.ops_sec_last_sample_time = mstime();
1306 server.ops_sec_last_sample_ops = 0;
1307 server.unixtime = time(NULL);
1308 server.lastbgsave_status = REDIS_OK;
1309 server.stop_writes_on_bgsave_err = 1;
1310 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1311 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1312 acceptTcpHandler,NULL) == AE_ERR) redisPanic("Unrecoverable error creating server.ipfd file event.");
1313 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1314 acceptUnixHandler,NULL) == AE_ERR) redisPanic("Unrecoverable error creating server.sofd file event.");
1315
1316 if (server.aof_state == REDIS_AOF_ON) {
1317 server.aof_fd = open(server.aof_filename,
1318 O_WRONLY|O_APPEND|O_CREAT,0644);
1319 if (server.aof_fd == -1) {
1320 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1321 strerror(errno));
1322 exit(1);
1323 }
1324 }
1325
1326 /* 32 bit instances are limited to 4GB of address space, so if there is
1327 * no explicit limit in the user provided configuration we set a limit
1328 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1329 * useless crashes of the Redis instance. */
1330 if (server.arch_bits == 32 && server.maxmemory == 0) {
1331 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1332 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1333 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1334 }
1335
1336 scriptingInit();
1337 slowlogInit();
1338 bioInit();
1339 }
1340
1341 /* Populates the Redis Command Table starting from the hard coded list
1342 * we have on top of redis.c file. */
1343 void populateCommandTable(void) {
1344 int j;
1345 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1346
1347 for (j = 0; j < numcommands; j++) {
1348 struct redisCommand *c = redisCommandTable+j;
1349 char *f = c->sflags;
1350 int retval;
1351
1352 while(*f != '\0') {
1353 switch(*f) {
1354 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1355 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1356 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1357 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1358 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1359 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1360 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1361 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1362 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1363 case 'l': c->flags |= REDIS_CMD_LOADING; break;
1364 case 't': c->flags |= REDIS_CMD_STALE; break;
1365 default: redisPanic("Unsupported command flag"); break;
1366 }
1367 f++;
1368 }
1369
1370 retval = dictAdd(server.commands, sdsnew(c->name), c);
1371 assert(retval == DICT_OK);
1372 }
1373 }
1374
1375 void resetCommandTableStats(void) {
1376 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1377 int j;
1378
1379 for (j = 0; j < numcommands; j++) {
1380 struct redisCommand *c = redisCommandTable+j;
1381
1382 c->microseconds = 0;
1383 c->calls = 0;
1384 }
1385 }
1386
1387 /* ========================== Redis OP Array API ============================ */
1388
1389 void redisOpArrayInit(redisOpArray *oa) {
1390 oa->ops = NULL;
1391 oa->numops = 0;
1392 }
1393
1394 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1395 robj **argv, int argc, int target)
1396 {
1397 redisOp *op;
1398
1399 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1400 op = oa->ops+oa->numops;
1401 op->cmd = cmd;
1402 op->dbid = dbid;
1403 op->argv = argv;
1404 op->argc = argc;
1405 op->target = target;
1406 oa->numops++;
1407 return oa->numops;
1408 }
1409
1410 void redisOpArrayFree(redisOpArray *oa) {
1411 while(oa->numops) {
1412 int j;
1413 redisOp *op;
1414
1415 oa->numops--;
1416 op = oa->ops+oa->numops;
1417 for (j = 0; j < op->argc; j++)
1418 decrRefCount(op->argv[j]);
1419 zfree(op->argv);
1420 }
1421 zfree(oa->ops);
1422 }
1423
1424 /* ====================== Commands lookup and execution ===================== */
1425
1426 struct redisCommand *lookupCommand(sds name) {
1427 return dictFetchValue(server.commands, name);
1428 }
1429
1430 struct redisCommand *lookupCommandByCString(char *s) {
1431 struct redisCommand *cmd;
1432 sds name = sdsnew(s);
1433
1434 cmd = dictFetchValue(server.commands, name);
1435 sdsfree(name);
1436 return cmd;
1437 }
1438
1439 /* Propagate the specified command (in the context of the specified database id)
1440 * to AOF, Slaves and Monitors.
1441 *
1442 * flags are an xor between:
1443 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1444 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1445 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1446 */
1447 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1448 int flags)
1449 {
1450 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1451 feedAppendOnlyFile(cmd,dbid,argv,argc);
1452 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1453 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1454 }
1455
1456 /* Used inside commands to schedule the propagation of additional commands
1457 * after the current command is propagated to AOF / Replication. */
1458 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1459 int target)
1460 {
1461 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1462 }
1463
1464 /* Call() is the core of Redis execution of a command */
1465 void call(redisClient *c, int flags) {
1466 long long dirty, start = ustime(), duration;
1467
1468 /* Sent the command to clients in MONITOR mode, only if the commands are
1469 * not geneated from reading an AOF. */
1470 if (listLength(server.monitors) && !server.loading)
1471 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1472
1473 /* Call the command. */
1474 redisOpArrayInit(&server.also_propagate);
1475 dirty = server.dirty;
1476 c->cmd->proc(c);
1477 dirty = server.dirty-dirty;
1478 duration = ustime()-start;
1479
1480 /* When EVAL is called loading the AOF we don't want commands called
1481 * from Lua to go into the slowlog or to populate statistics. */
1482 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1483 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1484
1485 /* Log the command into the Slow log if needed, and populate the
1486 * per-command statistics that we show in INFO commandstats. */
1487 if (flags & REDIS_CALL_SLOWLOG)
1488 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1489 if (flags & REDIS_CALL_STATS) {
1490 c->cmd->microseconds += duration;
1491 c->cmd->calls++;
1492 }
1493
1494 /* Propagate the command into the AOF and replication link */
1495 if (flags & REDIS_CALL_PROPAGATE) {
1496 int flags = REDIS_PROPAGATE_NONE;
1497
1498 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1499 flags |= REDIS_PROPAGATE_REPL;
1500 if (dirty)
1501 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1502 if (flags != REDIS_PROPAGATE_NONE)
1503 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1504 }
1505 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1506 * PUSH command. */
1507 if (server.also_propagate.numops) {
1508 int j;
1509 redisOp *rop;
1510
1511 for (j = 0; j < server.also_propagate.numops; j++) {
1512 rop = &server.also_propagate.ops[j];
1513 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1514 }
1515 redisOpArrayFree(&server.also_propagate);
1516 }
1517 server.stat_numcommands++;
1518 }
1519
1520 /* If this function gets called we already read a whole
1521 * command, argments are in the client argv/argc fields.
1522 * processCommand() execute the command or prepare the
1523 * server for a bulk read from the client.
1524 *
1525 * If 1 is returned the client is still alive and valid and
1526 * and other operations can be performed by the caller. Otherwise
1527 * if 0 is returned the client was destroied (i.e. after QUIT). */
1528 int processCommand(redisClient *c) {
1529 /* The QUIT command is handled separately. Normal command procs will
1530 * go through checking for replication and QUIT will cause trouble
1531 * when FORCE_REPLICATION is enabled and would be implemented in
1532 * a regular command proc. */
1533 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1534 addReply(c,shared.ok);
1535 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1536 return REDIS_ERR;
1537 }
1538
1539 /* Now lookup the command and check ASAP about trivial error conditions
1540 * such as wrong arity, bad command name and so forth. */
1541 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1542 if (!c->cmd) {
1543 addReplyErrorFormat(c,"unknown command '%s'",
1544 (char*)c->argv[0]->ptr);
1545 return REDIS_OK;
1546 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1547 (c->argc < -c->cmd->arity)) {
1548 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1549 c->cmd->name);
1550 return REDIS_OK;
1551 }
1552
1553 /* Check if the user is authenticated */
1554 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1555 {
1556 addReplyError(c,"operation not permitted");
1557 return REDIS_OK;
1558 }
1559
1560 /* Handle the maxmemory directive.
1561 *
1562 * First we try to free some memory if possible (if there are volatile
1563 * keys in the dataset). If there are not the only thing we can do
1564 * is returning an error. */
1565 if (server.maxmemory) {
1566 int retval = freeMemoryIfNeeded();
1567 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1568 addReply(c, shared.oomerr);
1569 return REDIS_OK;
1570 }
1571 }
1572
1573 /* Don't accept write commands if there are problems persisting on disk. */
1574 if (server.stop_writes_on_bgsave_err &&
1575 server.saveparamslen > 0
1576 && server.lastbgsave_status == REDIS_ERR &&
1577 c->cmd->flags & REDIS_CMD_WRITE)
1578 {
1579 addReply(c, shared.bgsaveerr);
1580 return REDIS_OK;
1581 }
1582
1583 /* Don't accept write commands if this is a read only slave. But
1584 * accept write commands if this is our master. */
1585 if (server.masterhost && server.repl_slave_ro &&
1586 !(c->flags & REDIS_MASTER) &&
1587 c->cmd->flags & REDIS_CMD_WRITE)
1588 {
1589 addReply(c, shared.roslaveerr);
1590 return REDIS_OK;
1591 }
1592
1593 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1594 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1595 &&
1596 c->cmd->proc != subscribeCommand &&
1597 c->cmd->proc != unsubscribeCommand &&
1598 c->cmd->proc != psubscribeCommand &&
1599 c->cmd->proc != punsubscribeCommand) {
1600 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1601 return REDIS_OK;
1602 }
1603
1604 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1605 * we are a slave with a broken link with master. */
1606 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1607 server.repl_serve_stale_data == 0 &&
1608 !(c->cmd->flags & REDIS_CMD_STALE))
1609 {
1610 addReply(c, shared.masterdownerr);
1611 return REDIS_OK;
1612 }
1613
1614 /* Loading DB? Return an error if the command has not the
1615 * REDIS_CMD_LOADING flag. */
1616 if (server.loading && !(c->cmd->flags & REDIS_CMD_LOADING)) {
1617 addReply(c, shared.loadingerr);
1618 return REDIS_OK;
1619 }
1620
1621 /* Lua script too slow? Only allow commands with REDIS_CMD_STALE flag. */
1622 if (server.lua_timedout &&
1623 !(c->cmd->proc == shutdownCommand &&
1624 c->argc == 2 &&
1625 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1626 !(c->cmd->proc == scriptCommand &&
1627 c->argc == 2 &&
1628 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1629 {
1630 addReply(c, shared.slowscripterr);
1631 return REDIS_OK;
1632 }
1633
1634 /* Exec the command */
1635 if (c->flags & REDIS_MULTI &&
1636 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1637 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1638 {
1639 queueMultiCommand(c);
1640 addReply(c,shared.queued);
1641 } else {
1642 call(c,REDIS_CALL_FULL);
1643 if (listLength(server.ready_keys))
1644 handleClientsBlockedOnLists();
1645 }
1646 return REDIS_OK;
1647 }
1648
1649 /*================================== Shutdown =============================== */
1650
1651 int prepareForShutdown(int flags) {
1652 int save = flags & REDIS_SHUTDOWN_SAVE;
1653 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1654
1655 redisLog(REDIS_WARNING,"User requested shutdown...");
1656 /* Kill the saving child if there is a background saving in progress.
1657 We want to avoid race conditions, for instance our saving child may
1658 overwrite the synchronous saving did by SHUTDOWN. */
1659 if (server.rdb_child_pid != -1) {
1660 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1661 kill(server.rdb_child_pid,SIGKILL);
1662 rdbRemoveTempFile(server.rdb_child_pid);
1663 }
1664 if (server.aof_state != REDIS_AOF_OFF) {
1665 /* Kill the AOF saving child as the AOF we already have may be longer
1666 * but contains the full dataset anyway. */
1667 if (server.aof_child_pid != -1) {
1668 redisLog(REDIS_WARNING,
1669 "There is a child rewriting the AOF. Killing it!");
1670 kill(server.aof_child_pid,SIGKILL);
1671 }
1672 /* Append only file: fsync() the AOF and exit */
1673 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1674 aof_fsync(server.aof_fd);
1675 }
1676 if ((server.saveparamslen > 0 && !nosave) || save) {
1677 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1678 /* Snapshotting. Perform a SYNC SAVE and exit */
1679 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1680 /* Ooops.. error saving! The best we can do is to continue
1681 * operating. Note that if there was a background saving process,
1682 * in the next cron() Redis will be notified that the background
1683 * saving aborted, handling special stuff like slaves pending for
1684 * synchronization... */
1685 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1686 return REDIS_ERR;
1687 }
1688 }
1689 if (server.daemonize) {
1690 redisLog(REDIS_NOTICE,"Removing the pid file.");
1691 unlink(server.pidfile);
1692 }
1693 /* Close the listening sockets. Apparently this allows faster restarts. */
1694 if (server.ipfd != -1) close(server.ipfd);
1695 if (server.sofd != -1) close(server.sofd);
1696 if (server.unixsocket) {
1697 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1698 unlink(server.unixsocket); /* don't care if this fails */
1699 }
1700
1701 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1702 return REDIS_OK;
1703 }
1704
1705 /*================================== Commands =============================== */
1706
1707 /* Return zero if strings are the same, non-zero if they are not.
1708 * The comparison is performed in a way that prevents an attacker to obtain
1709 * information about the nature of the strings just monitoring the execution
1710 * time of the function.
1711 *
1712 * Note that limiting the comparison length to strings up to 512 bytes we
1713 * can avoid leaking any information about the password length and any
1714 * possible branch misprediction related leak.
1715 */
1716 int time_independent_strcmp(char *a, char *b) {
1717 char bufa[REDIS_AUTHPASS_MAX_LEN], bufb[REDIS_AUTHPASS_MAX_LEN];
1718 /* The above two strlen perform len(a) + len(b) operations where either
1719 * a or b are fixed (our password) length, and the difference is only
1720 * relative to the length of the user provided string, so no information
1721 * leak is possible in the following two lines of code. */
1722 int alen = strlen(a);
1723 int blen = strlen(b);
1724 int j;
1725 int diff = 0;
1726
1727 /* We can't compare strings longer than our static buffers.
1728 * Note that this will never pass the first test in practical circumstances
1729 * so there is no info leak. */
1730 if (alen > sizeof(bufa) || blen > sizeof(bufb)) return 1;
1731
1732 memset(bufa,0,sizeof(bufa)); /* Constant time. */
1733 memset(bufb,0,sizeof(bufb)); /* Constant time. */
1734 /* Again the time of the following two copies is proportional to
1735 * len(a) + len(b) so no info is leaked. */
1736 memcpy(bufa,a,alen);
1737 memcpy(bufb,b,blen);
1738
1739 /* Always compare all the chars in the two buffers without
1740 * conditional expressions. */
1741 for (j = 0; j < sizeof(bufa); j++) {
1742 diff |= (bufa[j] ^ bufb[j]);
1743 }
1744 /* Length must be equal as well. */
1745 diff |= alen ^ blen;
1746 return diff; /* If zero strings are the same. */
1747 }
1748
1749 void authCommand(redisClient *c) {
1750 if (!server.requirepass) {
1751 addReplyError(c,"Client sent AUTH, but no password is set");
1752 } else if (!time_independent_strcmp(c->argv[1]->ptr, server.requirepass)) {
1753 c->authenticated = 1;
1754 addReply(c,shared.ok);
1755 } else {
1756 c->authenticated = 0;
1757 addReplyError(c,"invalid password");
1758 }
1759 }
1760
1761 void pingCommand(redisClient *c) {
1762 addReply(c,shared.pong);
1763 }
1764
1765 void echoCommand(redisClient *c) {
1766 addReplyBulk(c,c->argv[1]);
1767 }
1768
1769 void timeCommand(redisClient *c) {
1770 struct timeval tv;
1771
1772 /* gettimeofday() can only fail if &tv is a bad addresss so we
1773 * don't check for errors. */
1774 gettimeofday(&tv,NULL);
1775 addReplyMultiBulkLen(c,2);
1776 addReplyBulkLongLong(c,tv.tv_sec);
1777 addReplyBulkLongLong(c,tv.tv_usec);
1778 }
1779
1780 /* Convert an amount of bytes into a human readable string in the form
1781 * of 100B, 2G, 100M, 4K, and so forth. */
1782 void bytesToHuman(char *s, unsigned long long n) {
1783 double d;
1784
1785 if (n < 1024) {
1786 /* Bytes */
1787 sprintf(s,"%lluB",n);
1788 return;
1789 } else if (n < (1024*1024)) {
1790 d = (double)n/(1024);
1791 sprintf(s,"%.2fK",d);
1792 } else if (n < (1024LL*1024*1024)) {
1793 d = (double)n/(1024*1024);
1794 sprintf(s,"%.2fM",d);
1795 } else if (n < (1024LL*1024*1024*1024)) {
1796 d = (double)n/(1024LL*1024*1024);
1797 sprintf(s,"%.2fG",d);
1798 }
1799 }
1800
1801 /* Create the string returned by the INFO command. This is decoupled
1802 * by the INFO command itself as we need to report the same information
1803 * on memory corruption problems. */
1804 sds genRedisInfoString(char *section) {
1805 sds info = sdsempty();
1806 time_t uptime = server.unixtime-server.stat_starttime;
1807 int j, numcommands;
1808 struct rusage self_ru, c_ru;
1809 unsigned long lol, bib;
1810 int allsections = 0, defsections = 0;
1811 int sections = 0;
1812
1813 if (section) {
1814 allsections = strcasecmp(section,"all") == 0;
1815 defsections = strcasecmp(section,"default") == 0;
1816 }
1817
1818 getrusage(RUSAGE_SELF, &self_ru);
1819 getrusage(RUSAGE_CHILDREN, &c_ru);
1820 getClientsMaxBuffers(&lol,&bib);
1821
1822 /* Server */
1823 if (allsections || defsections || !strcasecmp(section,"server")) {
1824 struct utsname name;
1825
1826 if (sections++) info = sdscat(info,"\r\n");
1827 uname(&name);
1828 info = sdscatprintf(info,
1829 "# Server\r\n"
1830 "redis_version:%s\r\n"
1831 "redis_git_sha1:%s\r\n"
1832 "redis_git_dirty:%d\r\n"
1833 "os:%s %s %s\r\n"
1834 "arch_bits:%d\r\n"
1835 "multiplexing_api:%s\r\n"
1836 "gcc_version:%d.%d.%d\r\n"
1837 "process_id:%ld\r\n"
1838 "run_id:%s\r\n"
1839 "tcp_port:%d\r\n"
1840 "uptime_in_seconds:%ld\r\n"
1841 "uptime_in_days:%ld\r\n"
1842 "lru_clock:%ld\r\n",
1843 REDIS_VERSION,
1844 redisGitSHA1(),
1845 strtol(redisGitDirty(),NULL,10) > 0,
1846 name.sysname, name.release, name.machine,
1847 server.arch_bits,
1848 aeGetApiName(),
1849 #ifdef __GNUC__
1850 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1851 #else
1852 0,0,0,
1853 #endif
1854 (long) getpid(),
1855 server.runid,
1856 server.port,
1857 uptime,
1858 uptime/(3600*24),
1859 (unsigned long) server.lruclock);
1860 }
1861
1862 /* Clients */
1863 if (allsections || defsections || !strcasecmp(section,"clients")) {
1864 if (sections++) info = sdscat(info,"\r\n");
1865 info = sdscatprintf(info,
1866 "# Clients\r\n"
1867 "connected_clients:%lu\r\n"
1868 "client_longest_output_list:%lu\r\n"
1869 "client_biggest_input_buf:%lu\r\n"
1870 "blocked_clients:%d\r\n",
1871 listLength(server.clients)-listLength(server.slaves),
1872 lol, bib,
1873 server.bpop_blocked_clients);
1874 }
1875
1876 /* Memory */
1877 if (allsections || defsections || !strcasecmp(section,"memory")) {
1878 char hmem[64];
1879 char peak_hmem[64];
1880
1881 bytesToHuman(hmem,zmalloc_used_memory());
1882 bytesToHuman(peak_hmem,server.stat_peak_memory);
1883 if (sections++) info = sdscat(info,"\r\n");
1884 info = sdscatprintf(info,
1885 "# Memory\r\n"
1886 "used_memory:%zu\r\n"
1887 "used_memory_human:%s\r\n"
1888 "used_memory_rss:%zu\r\n"
1889 "used_memory_peak:%zu\r\n"
1890 "used_memory_peak_human:%s\r\n"
1891 "used_memory_lua:%lld\r\n"
1892 "mem_fragmentation_ratio:%.2f\r\n"
1893 "mem_allocator:%s\r\n",
1894 zmalloc_used_memory(),
1895 hmem,
1896 zmalloc_get_rss(),
1897 server.stat_peak_memory,
1898 peak_hmem,
1899 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1900 zmalloc_get_fragmentation_ratio(),
1901 ZMALLOC_LIB
1902 );
1903 }
1904
1905 /* Persistence */
1906 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1907 if (sections++) info = sdscat(info,"\r\n");
1908 info = sdscatprintf(info,
1909 "# Persistence\r\n"
1910 "loading:%d\r\n"
1911 "rdb_changes_since_last_save:%lld\r\n"
1912 "rdb_bgsave_in_progress:%d\r\n"
1913 "rdb_last_save_time:%ld\r\n"
1914 "rdb_last_bgsave_status:%s\r\n"
1915 "rdb_last_bgsave_time_sec:%ld\r\n"
1916 "rdb_current_bgsave_time_sec:%ld\r\n"
1917 "aof_enabled:%d\r\n"
1918 "aof_rewrite_in_progress:%d\r\n"
1919 "aof_rewrite_scheduled:%d\r\n"
1920 "aof_last_rewrite_time_sec:%ld\r\n"
1921 "aof_current_rewrite_time_sec:%ld\r\n"
1922 "aof_last_bgrewrite_status:%s\r\n",
1923 server.loading,
1924 server.dirty,
1925 server.rdb_child_pid != -1,
1926 server.lastsave,
1927 (server.lastbgsave_status == REDIS_OK) ? "ok" : "err",
1928 server.rdb_save_time_last,
1929 (server.rdb_child_pid == -1) ?
1930 -1 : time(NULL)-server.rdb_save_time_start,
1931 server.aof_state != REDIS_AOF_OFF,
1932 server.aof_child_pid != -1,
1933 server.aof_rewrite_scheduled,
1934 server.aof_rewrite_time_last,
1935 (server.aof_child_pid == -1) ?
1936 -1 : time(NULL)-server.aof_rewrite_time_start,
1937 (server.aof_lastbgrewrite_status == REDIS_OK) ? "ok" : "err");
1938
1939 if (server.aof_state != REDIS_AOF_OFF) {
1940 info = sdscatprintf(info,
1941 "aof_current_size:%lld\r\n"
1942 "aof_base_size:%lld\r\n"
1943 "aof_pending_rewrite:%d\r\n"
1944 "aof_buffer_length:%zu\r\n"
1945 "aof_rewrite_buffer_length:%lu\r\n"
1946 "aof_pending_bio_fsync:%llu\r\n"
1947 "aof_delayed_fsync:%lu\r\n",
1948 (long long) server.aof_current_size,
1949 (long long) server.aof_rewrite_base_size,
1950 server.aof_rewrite_scheduled,
1951 sdslen(server.aof_buf),
1952 aofRewriteBufferSize(),
1953 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1954 server.aof_delayed_fsync);
1955 }
1956
1957 if (server.loading) {
1958 double perc;
1959 time_t eta, elapsed;
1960 off_t remaining_bytes = server.loading_total_bytes-
1961 server.loading_loaded_bytes;
1962
1963 perc = ((double)server.loading_loaded_bytes /
1964 server.loading_total_bytes) * 100;
1965
1966 elapsed = server.unixtime-server.loading_start_time;
1967 if (elapsed == 0) {
1968 eta = 1; /* A fake 1 second figure if we don't have
1969 enough info */
1970 } else {
1971 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1972 }
1973
1974 info = sdscatprintf(info,
1975 "loading_start_time:%ld\r\n"
1976 "loading_total_bytes:%llu\r\n"
1977 "loading_loaded_bytes:%llu\r\n"
1978 "loading_loaded_perc:%.2f\r\n"
1979 "loading_eta_seconds:%ld\r\n"
1980 ,(unsigned long) server.loading_start_time,
1981 (unsigned long long) server.loading_total_bytes,
1982 (unsigned long long) server.loading_loaded_bytes,
1983 perc,
1984 eta
1985 );
1986 }
1987 }
1988
1989 /* Stats */
1990 if (allsections || defsections || !strcasecmp(section,"stats")) {
1991 if (sections++) info = sdscat(info,"\r\n");
1992 info = sdscatprintf(info,
1993 "# Stats\r\n"
1994 "total_connections_received:%lld\r\n"
1995 "total_commands_processed:%lld\r\n"
1996 "instantaneous_ops_per_sec:%lld\r\n"
1997 "rejected_connections:%lld\r\n"
1998 "expired_keys:%lld\r\n"
1999 "evicted_keys:%lld\r\n"
2000 "keyspace_hits:%lld\r\n"
2001 "keyspace_misses:%lld\r\n"
2002 "pubsub_channels:%ld\r\n"
2003 "pubsub_patterns:%lu\r\n"
2004 "latest_fork_usec:%lld\r\n",
2005 server.stat_numconnections,
2006 server.stat_numcommands,
2007 getOperationsPerSecond(),
2008 server.stat_rejected_conn,
2009 server.stat_expiredkeys,
2010 server.stat_evictedkeys,
2011 server.stat_keyspace_hits,
2012 server.stat_keyspace_misses,
2013 dictSize(server.pubsub_channels),
2014 listLength(server.pubsub_patterns),
2015 server.stat_fork_time);
2016 }
2017
2018 /* Replication */
2019 if (allsections || defsections || !strcasecmp(section,"replication")) {
2020 if (sections++) info = sdscat(info,"\r\n");
2021 info = sdscatprintf(info,
2022 "# Replication\r\n"
2023 "role:%s\r\n",
2024 server.masterhost == NULL ? "master" : "slave");
2025 if (server.masterhost) {
2026 info = sdscatprintf(info,
2027 "master_host:%s\r\n"
2028 "master_port:%d\r\n"
2029 "master_link_status:%s\r\n"
2030 "master_last_io_seconds_ago:%d\r\n"
2031 "master_sync_in_progress:%d\r\n"
2032 ,server.masterhost,
2033 server.masterport,
2034 (server.repl_state == REDIS_REPL_CONNECTED) ?
2035 "up" : "down",
2036 server.master ?
2037 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
2038 server.repl_state == REDIS_REPL_TRANSFER
2039 );
2040
2041 if (server.repl_state == REDIS_REPL_TRANSFER) {
2042 info = sdscatprintf(info,
2043 "master_sync_left_bytes:%lld\r\n"
2044 "master_sync_last_io_seconds_ago:%d\r\n"
2045 , (long long)
2046 (server.repl_transfer_size - server.repl_transfer_read),
2047 (int)(server.unixtime-server.repl_transfer_lastio)
2048 );
2049 }
2050
2051 if (server.repl_state != REDIS_REPL_CONNECTED) {
2052 info = sdscatprintf(info,
2053 "master_link_down_since_seconds:%ld\r\n",
2054 (long)server.unixtime-server.repl_down_since);
2055 }
2056 info = sdscatprintf(info,
2057 "slave_priority:%d\r\n", server.slave_priority);
2058 }
2059 info = sdscatprintf(info,
2060 "connected_slaves:%lu\r\n",
2061 listLength(server.slaves));
2062 if (listLength(server.slaves)) {
2063 int slaveid = 0;
2064 listNode *ln;
2065 listIter li;
2066
2067 listRewind(server.slaves,&li);
2068 while((ln = listNext(&li))) {
2069 redisClient *slave = listNodeValue(ln);
2070 char *state = NULL;
2071 char ip[32];
2072 int port;
2073
2074 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
2075 switch(slave->replstate) {
2076 case REDIS_REPL_WAIT_BGSAVE_START:
2077 case REDIS_REPL_WAIT_BGSAVE_END:
2078 state = "wait_bgsave";
2079 break;
2080 case REDIS_REPL_SEND_BULK:
2081 state = "send_bulk";
2082 break;
2083 case REDIS_REPL_ONLINE:
2084 state = "online";
2085 break;
2086 }
2087 if (state == NULL) continue;
2088 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2089 slaveid,ip,slave->slave_listening_port,state);
2090 slaveid++;
2091 }
2092 }
2093 }
2094
2095 /* CPU */
2096 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2097 if (sections++) info = sdscat(info,"\r\n");
2098 info = sdscatprintf(info,
2099 "# CPU\r\n"
2100 "used_cpu_sys:%.2f\r\n"
2101 "used_cpu_user:%.2f\r\n"
2102 "used_cpu_sys_children:%.2f\r\n"
2103 "used_cpu_user_children:%.2f\r\n",
2104 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2105 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2106 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2107 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2108 }
2109
2110 /* cmdtime */
2111 if (allsections || !strcasecmp(section,"commandstats")) {
2112 if (sections++) info = sdscat(info,"\r\n");
2113 info = sdscatprintf(info, "# Commandstats\r\n");
2114 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2115 for (j = 0; j < numcommands; j++) {
2116 struct redisCommand *c = redisCommandTable+j;
2117
2118 if (!c->calls) continue;
2119 info = sdscatprintf(info,
2120 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2121 c->name, c->calls, c->microseconds,
2122 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2123 }
2124 }
2125
2126 /* Key space */
2127 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2128 if (sections++) info = sdscat(info,"\r\n");
2129 info = sdscatprintf(info, "# Keyspace\r\n");
2130 for (j = 0; j < server.dbnum; j++) {
2131 long long keys, vkeys;
2132
2133 keys = dictSize(server.db[j].dict);
2134 vkeys = dictSize(server.db[j].expires);
2135 if (keys || vkeys) {
2136 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2137 j, keys, vkeys);
2138 }
2139 }
2140 }
2141 return info;
2142 }
2143
2144 void infoCommand(redisClient *c) {
2145 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2146
2147 if (c->argc > 2) {
2148 addReply(c,shared.syntaxerr);
2149 return;
2150 }
2151 sds info = genRedisInfoString(section);
2152 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2153 (unsigned long)sdslen(info)));
2154 addReplySds(c,info);
2155 addReply(c,shared.crlf);
2156 }
2157
2158 void monitorCommand(redisClient *c) {
2159 /* ignore MONITOR if aleady slave or in monitor mode */
2160 if (c->flags & REDIS_SLAVE) return;
2161
2162 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2163 c->slaveseldb = 0;
2164 listAddNodeTail(server.monitors,c);
2165 addReply(c,shared.ok);
2166 }
2167
2168 /* ============================ Maxmemory directive ======================== */
2169
2170 /* This function gets called when 'maxmemory' is set on the config file to limit
2171 * the max memory used by the server, before processing a command.
2172 *
2173 * The goal of the function is to free enough memory to keep Redis under the
2174 * configured memory limit.
2175 *
2176 * The function starts calculating how many bytes should be freed to keep
2177 * Redis under the limit, and enters a loop selecting the best keys to
2178 * evict accordingly to the configured policy.
2179 *
2180 * If all the bytes needed to return back under the limit were freed the
2181 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2182 * should block the execution of commands that will result in more memory
2183 * used by the server.
2184 */
2185 int freeMemoryIfNeeded(void) {
2186 size_t mem_used, mem_tofree, mem_freed;
2187 int slaves = listLength(server.slaves);
2188
2189 /* Remove the size of slaves output buffers and AOF buffer from the
2190 * count of used memory. */
2191 mem_used = zmalloc_used_memory();
2192 if (slaves) {
2193 listIter li;
2194 listNode *ln;
2195
2196 listRewind(server.slaves,&li);
2197 while((ln = listNext(&li))) {
2198 redisClient *slave = listNodeValue(ln);
2199 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2200 if (obuf_bytes > mem_used)
2201 mem_used = 0;
2202 else
2203 mem_used -= obuf_bytes;
2204 }
2205 }
2206 if (server.aof_state != REDIS_AOF_OFF) {
2207 mem_used -= sdslen(server.aof_buf);
2208 mem_used -= aofRewriteBufferSize();
2209 }
2210
2211 /* Check if we are over the memory limit. */
2212 if (mem_used <= server.maxmemory) return REDIS_OK;
2213
2214 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2215 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2216
2217 /* Compute how much memory we need to free. */
2218 mem_tofree = mem_used - server.maxmemory;
2219 mem_freed = 0;
2220 while (mem_freed < mem_tofree) {
2221 int j, k, keys_freed = 0;
2222
2223 for (j = 0; j < server.dbnum; j++) {
2224 long bestval = 0; /* just to prevent warning */
2225 sds bestkey = NULL;
2226 struct dictEntry *de;
2227 redisDb *db = server.db+j;
2228 dict *dict;
2229
2230 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2231 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2232 {
2233 dict = server.db[j].dict;
2234 } else {
2235 dict = server.db[j].expires;
2236 }
2237 if (dictSize(dict) == 0) continue;
2238
2239 /* volatile-random and allkeys-random policy */
2240 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2241 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2242 {
2243 de = dictGetRandomKey(dict);
2244 bestkey = dictGetKey(de);
2245 }
2246
2247 /* volatile-lru and allkeys-lru policy */
2248 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2249 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2250 {
2251 for (k = 0; k < server.maxmemory_samples; k++) {
2252 sds thiskey;
2253 long thisval;
2254 robj *o;
2255
2256 de = dictGetRandomKey(dict);
2257 thiskey = dictGetKey(de);
2258 /* When policy is volatile-lru we need an additonal lookup
2259 * to locate the real key, as dict is set to db->expires. */
2260 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2261 de = dictFind(db->dict, thiskey);
2262 o = dictGetVal(de);
2263 thisval = estimateObjectIdleTime(o);
2264
2265 /* Higher idle time is better candidate for deletion */
2266 if (bestkey == NULL || thisval > bestval) {
2267 bestkey = thiskey;
2268 bestval = thisval;
2269 }
2270 }
2271 }
2272
2273 /* volatile-ttl */
2274 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2275 for (k = 0; k < server.maxmemory_samples; k++) {
2276 sds thiskey;
2277 long thisval;
2278
2279 de = dictGetRandomKey(dict);
2280 thiskey = dictGetKey(de);
2281 thisval = (long) dictGetVal(de);
2282
2283 /* Expire sooner (minor expire unix timestamp) is better
2284 * candidate for deletion */
2285 if (bestkey == NULL || thisval < bestval) {
2286 bestkey = thiskey;
2287 bestval = thisval;
2288 }
2289 }
2290 }
2291
2292 /* Finally remove the selected key. */
2293 if (bestkey) {
2294 long long delta;
2295
2296 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2297 propagateExpire(db,keyobj);
2298 /* We compute the amount of memory freed by dbDelete() alone.
2299 * It is possible that actually the memory needed to propagate
2300 * the DEL in AOF and replication link is greater than the one
2301 * we are freeing removing the key, but we can't account for
2302 * that otherwise we would never exit the loop.
2303 *
2304 * AOF and Output buffer memory will be freed eventually so
2305 * we only care about memory used by the key space. */
2306 delta = (long long) zmalloc_used_memory();
2307 dbDelete(db,keyobj);
2308 delta -= (long long) zmalloc_used_memory();
2309 mem_freed += delta;
2310 server.stat_evictedkeys++;
2311 decrRefCount(keyobj);
2312 keys_freed++;
2313
2314 /* When the memory to free starts to be big enough, we may
2315 * start spending so much time here that is impossible to
2316 * deliver data to the slaves fast enough, so we force the
2317 * transmission here inside the loop. */
2318 if (slaves) flushSlavesOutputBuffers();
2319 }
2320 }
2321 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2322 }
2323 return REDIS_OK;
2324 }
2325
2326 /* =================================== Main! ================================ */
2327
2328 #ifdef __linux__
2329 int linuxOvercommitMemoryValue(void) {
2330 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2331 char buf[64];
2332
2333 if (!fp) return -1;
2334 if (fgets(buf,64,fp) == NULL) {
2335 fclose(fp);
2336 return -1;
2337 }
2338 fclose(fp);
2339
2340 return atoi(buf);
2341 }
2342
2343 void linuxOvercommitMemoryWarning(void) {
2344 if (linuxOvercommitMemoryValue() == 0) {
2345 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.");
2346 }
2347 }
2348 #endif /* __linux__ */
2349
2350 void createPidFile(void) {
2351 /* Try to write the pid file in a best-effort way. */
2352 FILE *fp = fopen(server.pidfile,"w");
2353 if (fp) {
2354 fprintf(fp,"%d\n",(int)getpid());
2355 fclose(fp);
2356 }
2357 }
2358
2359 void daemonize(void) {
2360 int fd;
2361
2362 if (fork() != 0) exit(0); /* parent exits */
2363 setsid(); /* create a new session */
2364
2365 /* Every output goes to /dev/null. If Redis is daemonized but
2366 * the 'logfile' is set to 'stdout' in the configuration file
2367 * it will not log at all. */
2368 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2369 dup2(fd, STDIN_FILENO);
2370 dup2(fd, STDOUT_FILENO);
2371 dup2(fd, STDERR_FILENO);
2372 if (fd > STDERR_FILENO) close(fd);
2373 }
2374 }
2375
2376 void version() {
2377 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2378 REDIS_VERSION,
2379 redisGitSHA1(),
2380 atoi(redisGitDirty()) > 0,
2381 ZMALLOC_LIB,
2382 sizeof(long) == 4 ? 32 : 64);
2383 exit(0);
2384 }
2385
2386 void usage() {
2387 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2388 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2389 fprintf(stderr," ./redis-server -v or --version\n");
2390 fprintf(stderr," ./redis-server -h or --help\n");
2391 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2392 fprintf(stderr,"Examples:\n");
2393 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2394 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2395 fprintf(stderr," ./redis-server --port 7777\n");
2396 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2397 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2398 exit(1);
2399 }
2400
2401 void redisAsciiArt(void) {
2402 #include "asciilogo.h"
2403 char *buf = zmalloc(1024*16);
2404
2405 snprintf(buf,1024*16,ascii_logo,
2406 REDIS_VERSION,
2407 redisGitSHA1(),
2408 strtol(redisGitDirty(),NULL,10) > 0,
2409 (sizeof(long) == 8) ? "64" : "32",
2410 "stand alone",
2411 server.port,
2412 (long) getpid()
2413 );
2414 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2415 zfree(buf);
2416 }
2417
2418 static void sigtermHandler(int sig) {
2419 REDIS_NOTUSED(sig);
2420
2421 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2422 server.shutdown_asap = 1;
2423 }
2424
2425 void setupSignalHandlers(void) {
2426 struct sigaction act;
2427
2428 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2429 * Otherwise, sa_handler is used. */
2430 sigemptyset(&act.sa_mask);
2431 act.sa_flags = 0;
2432 act.sa_handler = sigtermHandler;
2433 sigaction(SIGTERM, &act, NULL);
2434
2435 #ifdef HAVE_BACKTRACE
2436 sigemptyset(&act.sa_mask);
2437 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2438 act.sa_sigaction = sigsegvHandler;
2439 sigaction(SIGSEGV, &act, NULL);
2440 sigaction(SIGBUS, &act, NULL);
2441 sigaction(SIGFPE, &act, NULL);
2442 sigaction(SIGILL, &act, NULL);
2443 #endif
2444 return;
2445 }
2446
2447 void memtest(size_t megabytes, int passes);
2448
2449 void redisOutOfMemoryHandler(size_t allocation_size) {
2450 redisLog(REDIS_WARNING,"Out Of Memory allocating %zu bytes!",
2451 allocation_size);
2452 redisPanic("OOM");
2453 }
2454
2455 int main(int argc, char **argv) {
2456 long long start;
2457 struct timeval tv;
2458
2459 /* We need to initialize our libraries, and the server configuration. */
2460 zmalloc_enable_thread_safeness();
2461 zmalloc_set_oom_handler(redisOutOfMemoryHandler);
2462 srand(time(NULL)^getpid());
2463 gettimeofday(&tv,NULL);
2464 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2465 initServerConfig();
2466
2467 if (argc >= 2) {
2468 int j = 1; /* First option to parse in argv[] */
2469 sds options = sdsempty();
2470 char *configfile = NULL;
2471
2472 /* Handle special options --help and --version */
2473 if (strcmp(argv[1], "-v") == 0 ||
2474 strcmp(argv[1], "--version") == 0) version();
2475 if (strcmp(argv[1], "--help") == 0 ||
2476 strcmp(argv[1], "-h") == 0) usage();
2477 if (strcmp(argv[1], "--test-memory") == 0) {
2478 if (argc == 3) {
2479 memtest(atoi(argv[2]),50);
2480 exit(0);
2481 } else {
2482 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2483 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2484 exit(1);
2485 }
2486 }
2487
2488 /* First argument is the config file name? */
2489 if (argv[j][0] != '-' || argv[j][1] != '-')
2490 configfile = argv[j++];
2491 /* All the other options are parsed and conceptually appended to the
2492 * configuration file. For instance --port 6380 will generate the
2493 * string "port 6380\n" to be parsed after the actual file name
2494 * is parsed, if any. */
2495 while(j != argc) {
2496 if (argv[j][0] == '-' && argv[j][1] == '-') {
2497 /* Option name */
2498 if (sdslen(options)) options = sdscat(options,"\n");
2499 options = sdscat(options,argv[j]+2);
2500 options = sdscat(options," ");
2501 } else {
2502 /* Option argument */
2503 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2504 options = sdscat(options," ");
2505 }
2506 j++;
2507 }
2508 resetServerSaveParams();
2509 loadServerConfig(configfile,options);
2510 sdsfree(options);
2511 } else {
2512 redisLog(REDIS_WARNING,"Warning: no config file specified, using the default config. In order to specify a config file use 'redis-server /path/to/redis.conf'");
2513 }
2514 if (server.daemonize) daemonize();
2515 initServer();
2516 if (server.daemonize) createPidFile();
2517 redisAsciiArt();
2518 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2519 #ifdef __linux__
2520 linuxOvercommitMemoryWarning();
2521 #endif
2522 start = ustime();
2523 if (server.aof_state == REDIS_AOF_ON) {
2524 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2525 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2526 } else {
2527 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2528 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2529 (float)(ustime()-start)/1000000);
2530 } else if (errno != ENOENT) {
2531 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2532 exit(1);
2533 }
2534 }
2535 if (server.ipfd > 0)
2536 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2537 if (server.sofd > 0)
2538 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2539 aeSetBeforeSleepProc(server.el,beforeSleep);
2540 aeMain(server.el);
2541 aeDeleteEventLoop(server.el);
2542 return 0;
2543 }
2544
2545 /* The End */