]> git.saurik.com Git - redis.git/blob - src/redis.c
Stop access to global vars. Not configurable.
[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
52 /* Our shared "common" objects */
53
54 struct sharedObjectsStruct shared;
55
56 /* Global vars that are actually used as constants. The following double
57 * values are used for double on-disk serialization, and are initialized
58 * at runtime to avoid strange compiler optimizations. */
59
60 double R_Zero, R_PosInf, R_NegInf, R_Nan;
61
62 /*================================= Globals ================================= */
63
64 /* Global vars */
65 struct redisServer server; /* server global state */
66 struct redisCommand *commandTable;
67
68 /* Our command table.
69 *
70 * Every entry is composed of the following fields:
71 *
72 * name: a string representing the command name.
73 * function: pointer to the C function implementing the command.
74 * arity: number of arguments, it is possible to use -N to say >= N
75 * sflags: command flags as string. See below for a table of flags.
76 * flags: flags as bitmask. Computed by Redis using the 'sflags' field.
77 * get_keys_proc: an optional function to get key arguments from a command.
78 * This is only used when the following three fields are not
79 * enough to specify what arguments are keys.
80 * first_key_index: first argument that is a key
81 * last_key_index: last argument that is a key
82 * key_step: step to get all the keys from first to last argument. For instance
83 * in MSET the step is two since arguments are key,val,key,val,...
84 * microseconds: microseconds of total execution time for this command.
85 * calls: total number of calls of this command.
86 *
87 * The flags, microseconds and calls fields are computed by Redis and should
88 * always be set to zero.
89 *
90 * Command flags are expressed using strings where every character represents
91 * a flag. Later the populateCommandTable() function will take care of
92 * populating the real 'flags' field using this characters.
93 *
94 * This is the meaning of the flags:
95 *
96 * w: write command (may modify the key space).
97 * r: read command (will never modify the key space).
98 * m: may increase memory usage once called. Don't allow if out of memory.
99 * a: admin command, like SAVE or SHUTDOWN.
100 * p: Pub/Sub related command.
101 * f: force replication of this command, regarless of server.dirty.
102 * s: command not allowed in scripts.
103 * R: random command. Command is not deterministic, that is, the same command
104 * with the same arguments, with the same key space, may have different
105 * results. For instance SPOP and RANDOMKEY are two random commands.
106 * S: Sort command output array if called from script, so that the output
107 * is deterministic.
108 */
109 struct redisCommand redisCommandTable[] = {
110 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
111 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
112 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
113 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
114 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
115 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
116 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
117 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
118 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
119 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
120 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
121 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
122 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
123 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
124 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
125 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
126 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
127 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
128 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
129 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
130 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
131 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
132 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
133 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
134 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
135 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
136 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
137 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
138 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
139 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
140 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
141 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
142 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
143 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
144 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
145 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
146 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
147 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
148 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
149 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
150 {"srandmember",srandmemberCommand,2,"rR",0,NULL,1,1,1,0,0},
151 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
152 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
153 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
154 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
155 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
156 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
157 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
158 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
159 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
160 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
161 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
162 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
163 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
164 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
165 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
166 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
167 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
168 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
169 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
170 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
171 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
172 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
173 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
174 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
175 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
176 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
177 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
178 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
179 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
180 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
182 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
183 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
184 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
185 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
186 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
187 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
188 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
189 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
190 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
191 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
192 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
193 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
194 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
195 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
196 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
197 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
198 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
199 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
200 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
201 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
202 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
203 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
204 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
205 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
206 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
207 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
208 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
209 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
210 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
211 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
212 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
213 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
214 {"exec",execCommand,1,"s",0,NULL,0,0,0,0,0},
215 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
216 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
217 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
218 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
219 {"sort",sortCommand,-2,"wmS",0,NULL,1,1,1,0,0},
220 {"info",infoCommand,-1,"r",0,NULL,0,0,0,0,0},
221 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
222 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
223 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
224 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
225 {"slaveof",slaveofCommand,3,"aws",0,NULL,0,0,0,0,0},
226 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
227 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
228 {"subscribe",subscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
229 {"unsubscribe",unsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
230 {"psubscribe",psubscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
231 {"punsubscribe",punsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
232 {"publish",publishCommand,3,"pf",0,NULL,0,0,0,0,0},
233 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
234 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
235 {"cluster",clusterCommand,-2,"ar",0,NULL,0,0,0,0,0},
236 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
237 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
238 {"asking",askingCommand,1,"r",0,NULL,0,0,0,0,0},
239 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
240 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
241 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
242 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
243 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
244 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
245 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
246 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0}
247 };
248
249 /*============================ Utility functions ============================ */
250
251 /* Low level logging. To use only for very big messages, otherwise
252 * redisLog() is to prefer. */
253 void redisLogRaw(int level, const char *msg) {
254 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
255 const char *c = ".-*#";
256 time_t now = time(NULL);
257 FILE *fp;
258 char buf[64];
259 int rawmode = (level & REDIS_LOG_RAW);
260
261 level &= 0xff; /* clear flags */
262 if (level < server.verbosity) return;
263
264 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
265 if (!fp) return;
266
267 if (rawmode) {
268 fprintf(fp,"%s",msg);
269 } else {
270 strftime(buf,sizeof(buf),"%d %b %H:%M:%S",localtime(&now));
271 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
272 }
273 fflush(fp);
274
275 if (server.logfile) fclose(fp);
276
277 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
278 }
279
280 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
281 * is used across the code. The raw version is only used in order to dump
282 * the INFO output on crash. */
283 void redisLog(int level, const char *fmt, ...) {
284 va_list ap;
285 char msg[REDIS_MAX_LOGMSG_LEN];
286
287 if ((level&0xff) < server.verbosity) return;
288
289 va_start(ap, fmt);
290 vsnprintf(msg, sizeof(msg), fmt, ap);
291 va_end(ap);
292
293 redisLogRaw(level,msg);
294 }
295
296 /* Log a fixed message without printf-alike capabilities, in a way that is
297 * safe to call from a signal handler.
298 *
299 * We actually use this only for signals that are not fatal from the point
300 * of view of Redis. Signals that are going to kill the server anyway and
301 * where we need printf-alike features are served by redisLog(). */
302 void redisLogFromHandler(int level, const char *msg) {
303 int fd;
304 char buf[64];
305
306 if ((level&0xff) < server.verbosity ||
307 (server.logfile == NULL && server.daemonize)) return;
308 fd = server.logfile ?
309 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
310 STDOUT_FILENO;
311 if (fd == -1) return;
312 ll2string(buf,sizeof(buf),getpid());
313 write(fd,"[",1);
314 write(fd,buf,strlen(buf));
315 write(fd," | signal handler] (",20);
316 ll2string(buf,sizeof(buf),time(NULL));
317 write(fd,buf,strlen(buf));
318 write(fd,") ",2);
319 write(fd,msg,strlen(msg));
320 write(fd,"\n",1);
321 if (server.logfile) close(fd);
322 }
323
324 /* Redis generally does not try to recover from out of memory conditions
325 * when allocating objects or strings, it is not clear if it will be possible
326 * to report this condition to the client since the networking layer itself
327 * is based on heap allocation for send buffers, so we simply abort.
328 * At least the code will be simpler to read... */
329 void oom(const char *msg) {
330 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
331 sleep(1);
332 abort();
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 /*====================== Hash table type implementation ==================== */
352
353 /* This is an hash table type that uses the SDS dynamic strings libary as
354 * keys and radis objects as values (objects can hold SDS strings,
355 * lists, sets). */
356
357 void dictVanillaFree(void *privdata, void *val)
358 {
359 DICT_NOTUSED(privdata);
360 zfree(val);
361 }
362
363 void dictListDestructor(void *privdata, void *val)
364 {
365 DICT_NOTUSED(privdata);
366 listRelease((list*)val);
367 }
368
369 int dictSdsKeyCompare(void *privdata, const void *key1,
370 const void *key2)
371 {
372 int l1,l2;
373 DICT_NOTUSED(privdata);
374
375 l1 = sdslen((sds)key1);
376 l2 = sdslen((sds)key2);
377 if (l1 != l2) return 0;
378 return memcmp(key1, key2, l1) == 0;
379 }
380
381 /* A case insensitive version used for the command lookup table. */
382 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
383 const void *key2)
384 {
385 DICT_NOTUSED(privdata);
386
387 return strcasecmp(key1, key2) == 0;
388 }
389
390 void dictRedisObjectDestructor(void *privdata, void *val)
391 {
392 DICT_NOTUSED(privdata);
393
394 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
395 decrRefCount(val);
396 }
397
398 void dictSdsDestructor(void *privdata, void *val)
399 {
400 DICT_NOTUSED(privdata);
401
402 sdsfree(val);
403 }
404
405 int dictObjKeyCompare(void *privdata, const void *key1,
406 const void *key2)
407 {
408 const robj *o1 = key1, *o2 = key2;
409 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
410 }
411
412 unsigned int dictObjHash(const void *key) {
413 const robj *o = key;
414 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
415 }
416
417 unsigned int dictSdsHash(const void *key) {
418 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
419 }
420
421 unsigned int dictSdsCaseHash(const void *key) {
422 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
423 }
424
425 int dictEncObjKeyCompare(void *privdata, const void *key1,
426 const void *key2)
427 {
428 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
429 int cmp;
430
431 if (o1->encoding == REDIS_ENCODING_INT &&
432 o2->encoding == REDIS_ENCODING_INT)
433 return o1->ptr == o2->ptr;
434
435 o1 = getDecodedObject(o1);
436 o2 = getDecodedObject(o2);
437 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
438 decrRefCount(o1);
439 decrRefCount(o2);
440 return cmp;
441 }
442
443 unsigned int dictEncObjHash(const void *key) {
444 robj *o = (robj*) key;
445
446 if (o->encoding == REDIS_ENCODING_RAW) {
447 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
448 } else {
449 if (o->encoding == REDIS_ENCODING_INT) {
450 char buf[32];
451 int len;
452
453 len = ll2string(buf,32,(long)o->ptr);
454 return dictGenHashFunction((unsigned char*)buf, len);
455 } else {
456 unsigned int hash;
457
458 o = getDecodedObject(o);
459 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
460 decrRefCount(o);
461 return hash;
462 }
463 }
464 }
465
466 /* Sets type hash table */
467 dictType setDictType = {
468 dictEncObjHash, /* hash function */
469 NULL, /* key dup */
470 NULL, /* val dup */
471 dictEncObjKeyCompare, /* key compare */
472 dictRedisObjectDestructor, /* key destructor */
473 NULL /* val destructor */
474 };
475
476 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
477 dictType zsetDictType = {
478 dictEncObjHash, /* hash function */
479 NULL, /* key dup */
480 NULL, /* val dup */
481 dictEncObjKeyCompare, /* key compare */
482 dictRedisObjectDestructor, /* key destructor */
483 NULL /* val destructor */
484 };
485
486 /* Db->dict, keys are sds strings, vals are Redis objects. */
487 dictType dbDictType = {
488 dictSdsHash, /* hash function */
489 NULL, /* key dup */
490 NULL, /* val dup */
491 dictSdsKeyCompare, /* key compare */
492 dictSdsDestructor, /* key destructor */
493 dictRedisObjectDestructor /* val destructor */
494 };
495
496 /* Db->expires */
497 dictType keyptrDictType = {
498 dictSdsHash, /* hash function */
499 NULL, /* key dup */
500 NULL, /* val dup */
501 dictSdsKeyCompare, /* key compare */
502 NULL, /* key destructor */
503 NULL /* val destructor */
504 };
505
506 /* Command table. sds string -> command struct pointer. */
507 dictType commandTableDictType = {
508 dictSdsCaseHash, /* hash function */
509 NULL, /* key dup */
510 NULL, /* val dup */
511 dictSdsKeyCaseCompare, /* key compare */
512 dictSdsDestructor, /* key destructor */
513 NULL /* val destructor */
514 };
515
516 /* Hash type hash table (note that small hashes are represented with zimpaps) */
517 dictType hashDictType = {
518 dictEncObjHash, /* hash function */
519 NULL, /* key dup */
520 NULL, /* val dup */
521 dictEncObjKeyCompare, /* key compare */
522 dictRedisObjectDestructor, /* key destructor */
523 dictRedisObjectDestructor /* val destructor */
524 };
525
526 /* Keylist hash table type has unencoded redis objects as keys and
527 * lists as values. It's used for blocking operations (BLPOP) and to
528 * map swapped keys to a list of clients waiting for this keys to be loaded. */
529 dictType keylistDictType = {
530 dictObjHash, /* hash function */
531 NULL, /* key dup */
532 NULL, /* val dup */
533 dictObjKeyCompare, /* key compare */
534 dictRedisObjectDestructor, /* key destructor */
535 dictListDestructor /* val destructor */
536 };
537
538 /* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
539 * clusterNode structures. */
540 dictType clusterNodesDictType = {
541 dictSdsHash, /* hash function */
542 NULL, /* key dup */
543 NULL, /* val dup */
544 dictSdsKeyCompare, /* key compare */
545 dictSdsDestructor, /* key destructor */
546 NULL /* val destructor */
547 };
548
549 int htNeedsResize(dict *dict) {
550 long long size, used;
551
552 size = dictSlots(dict);
553 used = dictSize(dict);
554 return (size && used && size > DICT_HT_INITIAL_SIZE &&
555 (used*100/size < REDIS_HT_MINFILL));
556 }
557
558 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
559 * we resize the hash table to save memory */
560 void tryResizeHashTables(void) {
561 int j;
562
563 for (j = 0; j < server.dbnum; j++) {
564 if (htNeedsResize(server.db[j].dict))
565 dictResize(server.db[j].dict);
566 if (htNeedsResize(server.db[j].expires))
567 dictResize(server.db[j].expires);
568 }
569 }
570
571 /* Our hash table implementation performs rehashing incrementally while
572 * we write/read from the hash table. Still if the server is idle, the hash
573 * table will use two tables for a long time. So we try to use 1 millisecond
574 * of CPU time at every serverCron() loop in order to rehash some key. */
575 void incrementallyRehash(void) {
576 int j;
577
578 for (j = 0; j < server.dbnum; j++) {
579 if (dictIsRehashing(server.db[j].dict)) {
580 dictRehashMilliseconds(server.db[j].dict,1);
581 break; /* already used our millisecond for this loop... */
582 }
583 }
584 }
585
586 /* This function is called once a background process of some kind terminates,
587 * as we want to avoid resizing the hash tables when there is a child in order
588 * to play well with copy-on-write (otherwise when a resize happens lots of
589 * memory pages are copied). The goal of this function is to update the ability
590 * for dict.c to resize the hash tables accordingly to the fact we have o not
591 * running childs. */
592 void updateDictResizePolicy(void) {
593 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
594 dictEnableResize();
595 else
596 dictDisableResize();
597 }
598
599 /* ======================= Cron: called every 100 ms ======================== */
600
601 /* Try to expire a few timed out keys. The algorithm used is adaptive and
602 * will use few CPU cycles if there are few expiring keys, otherwise
603 * it will get more aggressive to avoid that too much memory is used by
604 * keys that can be removed from the keyspace. */
605 void activeExpireCycle(void) {
606 int j;
607
608 for (j = 0; j < server.dbnum; j++) {
609 int expired;
610 redisDb *db = server.db+j;
611
612 /* Continue to expire if at the end of the cycle more than 25%
613 * of the keys were expired. */
614 do {
615 long num = dictSize(db->expires);
616 long long now = mstime();
617
618 expired = 0;
619 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
620 num = REDIS_EXPIRELOOKUPS_PER_CRON;
621 while (num--) {
622 dictEntry *de;
623 long long t;
624
625 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
626 t = dictGetSignedIntegerVal(de);
627 if (now > t) {
628 sds key = dictGetKey(de);
629 robj *keyobj = createStringObject(key,sdslen(key));
630
631 propagateExpire(db,keyobj);
632 dbDelete(db,keyobj);
633 decrRefCount(keyobj);
634 expired++;
635 server.stat_expiredkeys++;
636 }
637 }
638 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
639 }
640 }
641
642 void updateLRUClock(void) {
643 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
644 REDIS_LRU_CLOCK_MAX;
645 }
646
647
648 /* Add a sample to the operations per second array of samples. */
649 void trackOperationsPerSecond(void) {
650 long long t = mstime() - server.ops_sec_last_sample_time;
651 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
652 long long ops_sec;
653
654 ops_sec = t > 0 ? (ops*1000/t) : 0;
655
656 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
657 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
658 server.ops_sec_last_sample_time = mstime();
659 server.ops_sec_last_sample_ops = server.stat_numcommands;
660 }
661
662 /* Return the mean of all the samples. */
663 long long getOperationsPerSecond(void) {
664 int j;
665 long long sum = 0;
666
667 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
668 sum += server.ops_sec_samples[j];
669 return sum / REDIS_OPS_SEC_SAMPLES;
670 }
671
672 /* Check for timeouts. Returns non-zero if the client was terminated */
673 int clientsCronHandleTimeout(redisClient *c) {
674 time_t now = server.unixtime;
675
676 if (server.maxidletime &&
677 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
678 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
679 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
680 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
681 listLength(c->pubsub_patterns) == 0 &&
682 (now - c->lastinteraction > server.maxidletime))
683 {
684 redisLog(REDIS_VERBOSE,"Closing idle client");
685 freeClient(c);
686 return 1;
687 } else if (c->flags & REDIS_BLOCKED) {
688 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
689 addReply(c,shared.nullmultibulk);
690 unblockClientWaitingData(c);
691 }
692 }
693 return 0;
694 }
695
696 /* The client query buffer is an sds.c string that can end with a lot of
697 * free space not used, this function reclaims space if needed.
698 *
699 * The funciton always returns 0 as it never terminates the client. */
700 int clientsCronResizeQueryBuffer(redisClient *c) {
701 size_t querybuf_size = sdsAllocSize(c->querybuf);
702 time_t idletime = server.unixtime - c->lastinteraction;
703
704 /* There are two conditions to resize the query buffer:
705 * 1) Query buffer is > BIG_ARG and too big for latest peak.
706 * 2) Client is inactive and the buffer is bigger than 1k. */
707 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
708 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
709 (querybuf_size > 1024 && idletime > 2))
710 {
711 /* Only resize the query buffer if it is actually wasting space. */
712 if (sdsavail(c->querybuf) > 1024) {
713 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
714 }
715 }
716 /* Reset the peak again to capture the peak memory usage in the next
717 * cycle. */
718 c->querybuf_peak = 0;
719 return 0;
720 }
721
722 void clientsCron(void) {
723 /* Make sure to process at least 1/100 of clients per call.
724 * Since this function is called 10 times per second we are sure that
725 * in the worst case we process all the clients in 10 seconds.
726 * In normal conditions (a reasonable number of clients) we process
727 * all the clients in a shorter time. */
728 int numclients = listLength(server.clients);
729 int iterations = numclients/100;
730
731 if (iterations < 50)
732 iterations = (numclients < 50) ? numclients : 50;
733 while(listLength(server.clients) && iterations--) {
734 redisClient *c;
735 listNode *head;
736
737 /* Rotate the list, take the current head, process.
738 * This way if the client must be removed from the list it's the
739 * first element and we don't incur into O(N) computation. */
740 listRotate(server.clients);
741 head = listFirst(server.clients);
742 c = listNodeValue(head);
743 /* The following functions do different service checks on the client.
744 * The protocol is that they return non-zero if the client was
745 * terminated. */
746 if (clientsCronHandleTimeout(c)) continue;
747 if (clientsCronResizeQueryBuffer(c)) continue;
748 }
749 }
750
751 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
752 int j, loops = server.cronloops;
753 REDIS_NOTUSED(eventLoop);
754 REDIS_NOTUSED(id);
755 REDIS_NOTUSED(clientData);
756
757 /* Software watchdog: deliver the SIGALRM that will reach the signal
758 * handler if we don't return here fast enough. */
759 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
760
761 /* We take a cached value of the unix time in the global state because
762 * with virtual memory and aging there is to store the current time
763 * in objects at every object access, and accuracy is not needed.
764 * To access a global var is faster than calling time(NULL) */
765 server.unixtime = time(NULL);
766
767 trackOperationsPerSecond();
768
769 /* We have just 22 bits per object for LRU information.
770 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
771 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
772 *
773 * Note that even if this will wrap after 1.5 years it's not a problem,
774 * everything will still work but just some object will appear younger
775 * to Redis. But for this to happen a given object should never be touched
776 * for 1.5 years.
777 *
778 * Note that you can change the resolution altering the
779 * REDIS_LRU_CLOCK_RESOLUTION define.
780 */
781 updateLRUClock();
782
783 /* Record the max memory used since the server was started. */
784 if (zmalloc_used_memory() > server.stat_peak_memory)
785 server.stat_peak_memory = zmalloc_used_memory();
786
787 /* We received a SIGTERM, shutting down here in a safe way, as it is
788 * not ok doing so inside the signal handler. */
789 if (server.shutdown_asap) {
790 if (prepareForShutdown(0) == REDIS_OK) exit(0);
791 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
792 }
793
794 /* Show some info about non-empty databases */
795 for (j = 0; j < server.dbnum; j++) {
796 long long size, used, vkeys;
797
798 size = dictSlots(server.db[j].dict);
799 used = dictSize(server.db[j].dict);
800 vkeys = dictSize(server.db[j].expires);
801 if (!(loops % 50) && (used || vkeys)) {
802 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
803 /* dictPrintStats(server.dict); */
804 }
805 }
806
807 /* We don't want to resize the hash tables while a bacground saving
808 * is in progress: the saving child is created using fork() that is
809 * implemented with a copy-on-write semantic in most modern systems, so
810 * if we resize the HT while there is the saving child at work actually
811 * a lot of memory movements in the parent will cause a lot of pages
812 * copied. */
813 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
814 if (!(loops % 10)) tryResizeHashTables();
815 if (server.activerehashing) incrementallyRehash();
816 }
817
818 /* Show information about connected clients */
819 if (!(loops % 50)) {
820 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
821 listLength(server.clients)-listLength(server.slaves),
822 listLength(server.slaves),
823 zmalloc_used_memory());
824 }
825
826 /* We need to do a few operations on clients asynchronously. */
827 clientsCron();
828
829 /* Start a scheduled AOF rewrite if this was requested by the user while
830 * a BGSAVE was in progress. */
831 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
832 server.aof_rewrite_scheduled)
833 {
834 rewriteAppendOnlyFileBackground();
835 }
836
837 /* Check if a background saving or AOF rewrite in progress terminated. */
838 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
839 int statloc;
840 pid_t pid;
841
842 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
843 int exitcode = WEXITSTATUS(statloc);
844 int bysignal = 0;
845
846 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
847
848 if (pid == server.rdb_child_pid) {
849 backgroundSaveDoneHandler(exitcode,bysignal);
850 } else {
851 backgroundRewriteDoneHandler(exitcode,bysignal);
852 }
853 updateDictResizePolicy();
854 }
855 } else {
856 /* If there is not a background saving/rewrite in progress check if
857 * we have to save/rewrite now */
858 for (j = 0; j < server.saveparamslen; j++) {
859 struct saveparam *sp = server.saveparams+j;
860
861 if (server.dirty >= sp->changes &&
862 server.unixtime-server.lastsave > sp->seconds) {
863 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
864 sp->changes, sp->seconds);
865 rdbSaveBackground(server.rdb_filename);
866 break;
867 }
868 }
869
870 /* Trigger an AOF rewrite if needed */
871 if (server.rdb_child_pid == -1 &&
872 server.aof_child_pid == -1 &&
873 server.aof_rewrite_perc &&
874 server.aof_current_size > server.aof_rewrite_min_size)
875 {
876 long long base = server.aof_rewrite_base_size ?
877 server.aof_rewrite_base_size : 1;
878 long long growth = (server.aof_current_size*100/base) - 100;
879 if (growth >= server.aof_rewrite_perc) {
880 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
881 rewriteAppendOnlyFileBackground();
882 }
883 }
884 }
885
886
887 /* If we postponed an AOF buffer flush, let's try to do it every time the
888 * cron function is called. */
889 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
890
891 /* Expire a few keys per cycle, only if this is a master.
892 * On slaves we wait for DEL operations synthesized by the master
893 * in order to guarantee a strict consistency. */
894 if (server.masterhost == NULL) activeExpireCycle();
895
896 /* Close clients that need to be closed asynchronous */
897 freeClientsInAsyncFreeQueue();
898
899 /* Replication cron function -- used to reconnect to master and
900 * to detect transfer failures. */
901 if (!(loops % 10)) replicationCron();
902
903 /* Run other sub-systems specific cron jobs */
904 if (server.cluster_enabled && !(loops % 10)) clusterCron();
905
906 server.cronloops++;
907 return 100;
908 }
909
910 /* This function gets called every time Redis is entering the
911 * main loop of the event driven library, that is, before to sleep
912 * for ready file descriptors. */
913 void beforeSleep(struct aeEventLoop *eventLoop) {
914 REDIS_NOTUSED(eventLoop);
915 listNode *ln;
916 redisClient *c;
917
918 /* Try to process pending commands for clients that were just unblocked. */
919 while (listLength(server.unblocked_clients)) {
920 ln = listFirst(server.unblocked_clients);
921 redisAssert(ln != NULL);
922 c = ln->value;
923 listDelNode(server.unblocked_clients,ln);
924 c->flags &= ~REDIS_UNBLOCKED;
925
926 /* Process remaining data in the input buffer. */
927 if (c->querybuf && sdslen(c->querybuf) > 0) {
928 server.current_client = c;
929 processInputBuffer(c);
930 server.current_client = NULL;
931 }
932 }
933
934 /* Write the AOF buffer on disk */
935 flushAppendOnlyFile(0);
936 }
937
938 /* =========================== Server initialization ======================== */
939
940 void createSharedObjects(void) {
941 int j;
942
943 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
944 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
945 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
946 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
947 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
948 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
949 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
950 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
951 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
952 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
953 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
954 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
955 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
956 "-ERR Operation against a key holding the wrong kind of value\r\n"));
957 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
958 "-ERR no such key\r\n"));
959 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
960 "-ERR syntax error\r\n"));
961 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
962 "-ERR source and destination objects are the same\r\n"));
963 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
964 "-ERR index out of range\r\n"));
965 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
966 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
967 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
968 "-LOADING Redis is loading the dataset in memory\r\n"));
969 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
970 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
971 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
972 "-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"));
973 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
974 "-READONLY You can't write against a read only slave.\r\n"));
975 shared.oomerr = createObject(REDIS_STRING,sdsnew(
976 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
977 shared.space = createObject(REDIS_STRING,sdsnew(" "));
978 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
979 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
980 shared.select0 = createStringObject("select 0\r\n",10);
981 shared.select1 = createStringObject("select 1\r\n",10);
982 shared.select2 = createStringObject("select 2\r\n",10);
983 shared.select3 = createStringObject("select 3\r\n",10);
984 shared.select4 = createStringObject("select 4\r\n",10);
985 shared.select5 = createStringObject("select 5\r\n",10);
986 shared.select6 = createStringObject("select 6\r\n",10);
987 shared.select7 = createStringObject("select 7\r\n",10);
988 shared.select8 = createStringObject("select 8\r\n",10);
989 shared.select9 = createStringObject("select 9\r\n",10);
990 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
991 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
992 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
993 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
994 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
995 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
996 shared.del = createStringObject("DEL",3);
997 shared.rpop = createStringObject("RPOP",4);
998 shared.lpop = createStringObject("LPOP",4);
999 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1000 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1001 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1002 }
1003 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1004 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1005 sdscatprintf(sdsempty(),"*%d\r\n",j));
1006 shared.bulkhdr[j] = createObject(REDIS_STRING,
1007 sdscatprintf(sdsempty(),"$%d\r\n",j));
1008 }
1009 }
1010
1011 void initServerConfig() {
1012 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1013 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1014 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1015 server.port = REDIS_SERVERPORT;
1016 server.bindaddr = NULL;
1017 server.unixsocket = NULL;
1018 server.unixsocketperm = 0;
1019 server.ipfd = -1;
1020 server.sofd = -1;
1021 server.dbnum = REDIS_DEFAULT_DBNUM;
1022 server.verbosity = REDIS_NOTICE;
1023 server.maxidletime = REDIS_MAXIDLETIME;
1024 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1025 server.saveparams = NULL;
1026 server.loading = 0;
1027 server.logfile = NULL; /* NULL = log on standard output */
1028 server.syslog_enabled = 0;
1029 server.syslog_ident = zstrdup("redis");
1030 server.syslog_facility = LOG_LOCAL0;
1031 server.daemonize = 0;
1032 server.aof_state = REDIS_AOF_OFF;
1033 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1034 server.aof_no_fsync_on_rewrite = 0;
1035 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1036 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1037 server.aof_rewrite_base_size = 0;
1038 server.aof_rewrite_scheduled = 0;
1039 server.aof_last_fsync = time(NULL);
1040 server.aof_delayed_fsync = 0;
1041 server.aof_fd = -1;
1042 server.aof_selected_db = -1; /* Make sure the first time will not match */
1043 server.aof_flush_postponed_start = 0;
1044 server.pidfile = zstrdup("/var/run/redis.pid");
1045 server.rdb_filename = zstrdup("dump.rdb");
1046 server.aof_filename = zstrdup("appendonly.aof");
1047 server.requirepass = NULL;
1048 server.rdb_compression = 1;
1049 server.activerehashing = 1;
1050 server.maxclients = REDIS_MAX_CLIENTS;
1051 server.bpop_blocked_clients = 0;
1052 server.maxmemory = 0;
1053 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1054 server.maxmemory_samples = 3;
1055 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1056 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1057 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1058 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1059 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1060 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1061 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1062 server.shutdown_asap = 0;
1063 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1064 server.repl_timeout = REDIS_REPL_TIMEOUT;
1065 server.cluster_enabled = 0;
1066 server.cluster.configfile = zstrdup("nodes.conf");
1067 server.lua_caller = NULL;
1068 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1069 server.lua_client = NULL;
1070 server.lua_timedout = 0;
1071
1072 updateLRUClock();
1073 resetServerSaveParams();
1074
1075 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1076 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1077 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1078 /* Replication related */
1079 server.masterauth = NULL;
1080 server.masterhost = NULL;
1081 server.masterport = 6379;
1082 server.master = NULL;
1083 server.repl_state = REDIS_REPL_NONE;
1084 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1085 server.repl_serve_stale_data = 1;
1086 server.repl_slave_ro = 1;
1087 server.repl_down_since = -1;
1088
1089 /* Client output buffer limits */
1090 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1091 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1092 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1093 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1094 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1095 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1096 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1097 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1098 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1099
1100 /* Double constants initialization */
1101 R_Zero = 0.0;
1102 R_PosInf = 1.0/R_Zero;
1103 R_NegInf = -1.0/R_Zero;
1104 R_Nan = R_Zero/R_Zero;
1105
1106 /* Command table -- we intiialize it here as it is part of the
1107 * initial configuration, since command names may be changed via
1108 * redis.conf using the rename-command directive. */
1109 server.commands = dictCreate(&commandTableDictType,NULL);
1110 populateCommandTable();
1111 server.delCommand = lookupCommandByCString("del");
1112 server.multiCommand = lookupCommandByCString("multi");
1113 server.lpushCommand = lookupCommandByCString("lpush");
1114
1115 /* Slow log */
1116 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1117 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1118
1119 /* Debugging */
1120 server.assert_failed = "<no assertion failed>";
1121 server.assert_file = "<no file>";
1122 server.assert_line = 0;
1123 server.bug_report_start = 0;
1124 server.watchdog_period = 0;
1125 }
1126
1127 /* This function will try to raise the max number of open files accordingly to
1128 * the configured max number of clients. It will also account for 32 additional
1129 * file descriptors as we need a few more for persistence, listening
1130 * sockets, log files and so forth.
1131 *
1132 * If it will not be possible to set the limit accordingly to the configured
1133 * max number of clients, the function will do the reverse setting
1134 * server.maxclients to the value that we can actually handle. */
1135 void adjustOpenFilesLimit(void) {
1136 rlim_t maxfiles = server.maxclients+32;
1137 struct rlimit limit;
1138
1139 if (maxfiles < 1024) maxfiles = 1024;
1140 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1141 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1142 strerror(errno));
1143 server.maxclients = 1024-32;
1144 } else {
1145 rlim_t oldlimit = limit.rlim_cur;
1146
1147 /* Set the max number of files if the current limit is not enough
1148 * for our needs. */
1149 if (oldlimit < maxfiles) {
1150 limit.rlim_cur = maxfiles;
1151 limit.rlim_max = maxfiles;
1152 if (setrlimit(RLIMIT_NOFILE,&limit) == -1) {
1153 server.maxclients = oldlimit-32;
1154 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1155 (int) maxfiles, strerror(errno), (int) server.maxclients);
1156 } else {
1157 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1158 (int) maxfiles);
1159 }
1160 }
1161 }
1162 }
1163
1164 void initServer() {
1165 int j;
1166
1167 signal(SIGHUP, SIG_IGN);
1168 signal(SIGPIPE, SIG_IGN);
1169 setupSignalHandlers();
1170
1171 if (server.syslog_enabled) {
1172 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1173 server.syslog_facility);
1174 }
1175
1176 server.current_client = NULL;
1177 server.clients = listCreate();
1178 server.clients_to_close = listCreate();
1179 server.slaves = listCreate();
1180 server.monitors = listCreate();
1181 server.unblocked_clients = listCreate();
1182
1183 createSharedObjects();
1184 adjustOpenFilesLimit();
1185 server.el = aeCreateEventLoop(server.maxclients+1024);
1186 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1187
1188 if (server.port != 0) {
1189 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1190 if (server.ipfd == ANET_ERR) {
1191 redisLog(REDIS_WARNING, "Opening port %d: %s",
1192 server.port, server.neterr);
1193 exit(1);
1194 }
1195 }
1196 if (server.unixsocket != NULL) {
1197 unlink(server.unixsocket); /* don't care if this fails */
1198 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1199 if (server.sofd == ANET_ERR) {
1200 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1201 exit(1);
1202 }
1203 }
1204 if (server.ipfd < 0 && server.sofd < 0) {
1205 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1206 exit(1);
1207 }
1208 for (j = 0; j < server.dbnum; j++) {
1209 server.db[j].dict = dictCreate(&dbDictType,NULL);
1210 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1211 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1212 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1213 server.db[j].id = j;
1214 }
1215 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1216 server.pubsub_patterns = listCreate();
1217 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1218 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1219 server.cronloops = 0;
1220 server.rdb_child_pid = -1;
1221 server.aof_child_pid = -1;
1222 server.aof_rewrite_buf = sdsempty();
1223 server.aof_buf = sdsempty();
1224 server.lastsave = time(NULL);
1225 server.dirty = 0;
1226 server.stat_numcommands = 0;
1227 server.stat_numconnections = 0;
1228 server.stat_expiredkeys = 0;
1229 server.stat_evictedkeys = 0;
1230 server.stat_starttime = time(NULL);
1231 server.stat_keyspace_misses = 0;
1232 server.stat_keyspace_hits = 0;
1233 server.stat_peak_memory = 0;
1234 server.stat_fork_time = 0;
1235 server.stat_rejected_conn = 0;
1236 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1237 server.ops_sec_idx = 0;
1238 server.ops_sec_last_sample_time = mstime();
1239 server.ops_sec_last_sample_ops = 0;
1240 server.unixtime = time(NULL);
1241 server.lastbgsave_status = REDIS_OK;
1242 server.stop_writes_on_bgsave_err = 1;
1243 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1244 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1245 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1246 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1247 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1248
1249 if (server.aof_state == REDIS_AOF_ON) {
1250 server.aof_fd = open(server.aof_filename,
1251 O_WRONLY|O_APPEND|O_CREAT,0644);
1252 if (server.aof_fd == -1) {
1253 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1254 strerror(errno));
1255 exit(1);
1256 }
1257 }
1258
1259 /* 32 bit instances are limited to 4GB of address space, so if there is
1260 * no explicit limit in the user provided configuration we set a limit
1261 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1262 * useless crashes of the Redis instance. */
1263 if (server.arch_bits == 32 && server.maxmemory == 0) {
1264 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1265 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1266 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1267 }
1268
1269 if (server.cluster_enabled) clusterInit();
1270 scriptingInit();
1271 slowlogInit();
1272 bioInit();
1273 }
1274
1275 /* Populates the Redis Command Table starting from the hard coded list
1276 * we have on top of redis.c file. */
1277 void populateCommandTable(void) {
1278 int j;
1279 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1280
1281 for (j = 0; j < numcommands; j++) {
1282 struct redisCommand *c = redisCommandTable+j;
1283 char *f = c->sflags;
1284 int retval;
1285
1286 while(*f != '\0') {
1287 switch(*f) {
1288 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1289 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1290 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1291 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1292 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1293 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1294 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1295 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1296 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1297 default: redisPanic("Unsupported command flag"); break;
1298 }
1299 f++;
1300 }
1301
1302 retval = dictAdd(server.commands, sdsnew(c->name), c);
1303 assert(retval == DICT_OK);
1304 }
1305 }
1306
1307 void resetCommandTableStats(void) {
1308 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1309 int j;
1310
1311 for (j = 0; j < numcommands; j++) {
1312 struct redisCommand *c = redisCommandTable+j;
1313
1314 c->microseconds = 0;
1315 c->calls = 0;
1316 }
1317 }
1318
1319 /* ========================== Redis OP Array API ============================ */
1320
1321 void redisOpArrayInit(redisOpArray *oa) {
1322 oa->ops = NULL;
1323 oa->numops = 0;
1324 }
1325
1326 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1327 robj **argv, int argc, int target)
1328 {
1329 redisOp *op;
1330
1331 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1332 op = oa->ops+oa->numops;
1333 op->cmd = cmd;
1334 op->dbid = dbid;
1335 op->argv = argv;
1336 op->argc = argc;
1337 op->target = target;
1338 oa->numops++;
1339 return oa->numops;
1340 }
1341
1342 void redisOpArrayFree(redisOpArray *oa) {
1343 while(oa->numops) {
1344 int j;
1345 redisOp *op;
1346
1347 oa->numops--;
1348 op = oa->ops+oa->numops;
1349 for (j = 0; j < op->argc; j++)
1350 decrRefCount(op->argv[j]);
1351 zfree(op->argv);
1352 }
1353 zfree(oa->ops);
1354 }
1355
1356 /* ====================== Commands lookup and execution ===================== */
1357
1358 struct redisCommand *lookupCommand(sds name) {
1359 return dictFetchValue(server.commands, name);
1360 }
1361
1362 struct redisCommand *lookupCommandByCString(char *s) {
1363 struct redisCommand *cmd;
1364 sds name = sdsnew(s);
1365
1366 cmd = dictFetchValue(server.commands, name);
1367 sdsfree(name);
1368 return cmd;
1369 }
1370
1371 /* Propagate the specified command (in the context of the specified database id)
1372 * to AOF, Slaves and Monitors.
1373 *
1374 * flags are an xor between:
1375 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1376 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1377 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1378 */
1379 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1380 int flags)
1381 {
1382 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1383 feedAppendOnlyFile(cmd,dbid,argv,argc);
1384 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1385 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1386 }
1387
1388 /* Used inside commands to schedule the propagation of additional commands
1389 * after the current command is propagated to AOF / Replication. */
1390 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1391 int target)
1392 {
1393 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1394 }
1395
1396 /* Call() is the core of Redis execution of a command */
1397 void call(redisClient *c, int flags) {
1398 long long dirty, start = ustime(), duration;
1399
1400 /* Sent the command to clients in MONITOR mode, only if the commands are
1401 * not geneated from reading an AOF. */
1402 if (listLength(server.monitors) && !server.loading)
1403 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1404
1405 /* Call the command. */
1406 redisOpArrayInit(&server.also_propagate);
1407 dirty = server.dirty;
1408 c->cmd->proc(c);
1409 dirty = server.dirty-dirty;
1410 duration = ustime()-start;
1411
1412 /* When EVAL is called loading the AOF we don't want commands called
1413 * from Lua to go into the slowlog or to populate statistics. */
1414 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1415 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1416
1417 /* Log the command into the Slow log if needed, and populate the
1418 * per-command statistics that we show in INFO commandstats. */
1419 if (flags & REDIS_CALL_SLOWLOG)
1420 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1421 if (flags & REDIS_CALL_STATS) {
1422 c->cmd->microseconds += duration;
1423 c->cmd->calls++;
1424 }
1425
1426 /* Propagate the command into the AOF and replication link */
1427 if (flags & REDIS_CALL_PROPAGATE) {
1428 int flags = REDIS_PROPAGATE_NONE;
1429
1430 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1431 flags |= REDIS_PROPAGATE_REPL;
1432 if (dirty)
1433 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1434 if (flags != REDIS_PROPAGATE_NONE)
1435 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1436 }
1437 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1438 * PUSH command. */
1439 if (server.also_propagate.numops) {
1440 int j;
1441 redisOp *rop;
1442
1443 for (j = 0; j < server.also_propagate.numops; j++) {
1444 rop = &server.also_propagate.ops[j];
1445 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1446 }
1447 redisOpArrayFree(&server.also_propagate);
1448 }
1449 server.stat_numcommands++;
1450 }
1451
1452 /* If this function gets called we already read a whole
1453 * command, argments are in the client argv/argc fields.
1454 * processCommand() execute the command or prepare the
1455 * server for a bulk read from the client.
1456 *
1457 * If 1 is returned the client is still alive and valid and
1458 * and other operations can be performed by the caller. Otherwise
1459 * if 0 is returned the client was destroied (i.e. after QUIT). */
1460 int processCommand(redisClient *c) {
1461 /* The QUIT command is handled separately. Normal command procs will
1462 * go through checking for replication and QUIT will cause trouble
1463 * when FORCE_REPLICATION is enabled and would be implemented in
1464 * a regular command proc. */
1465 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1466 addReply(c,shared.ok);
1467 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1468 return REDIS_ERR;
1469 }
1470
1471 /* Now lookup the command and check ASAP about trivial error conditions
1472 * such as wrong arity, bad command name and so forth. */
1473 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1474 if (!c->cmd) {
1475 addReplyErrorFormat(c,"unknown command '%s'",
1476 (char*)c->argv[0]->ptr);
1477 return REDIS_OK;
1478 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1479 (c->argc < -c->cmd->arity)) {
1480 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1481 c->cmd->name);
1482 return REDIS_OK;
1483 }
1484
1485 /* Check if the user is authenticated */
1486 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1487 {
1488 addReplyError(c,"operation not permitted");
1489 return REDIS_OK;
1490 }
1491
1492 /* If cluster is enabled, redirect here */
1493 if (server.cluster_enabled &&
1494 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1495 int hashslot;
1496
1497 if (server.cluster.state != REDIS_CLUSTER_OK) {
1498 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1499 return REDIS_OK;
1500 } else {
1501 int ask;
1502 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1503 if (n == NULL) {
1504 addReplyError(c,"Multi keys request invalid in cluster");
1505 return REDIS_OK;
1506 } else if (n != server.cluster.myself) {
1507 addReplySds(c,sdscatprintf(sdsempty(),
1508 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1509 hashslot,n->ip,n->port));
1510 return REDIS_OK;
1511 }
1512 }
1513 }
1514
1515 /* Handle the maxmemory directive.
1516 *
1517 * First we try to free some memory if possible (if there are volatile
1518 * keys in the dataset). If there are not the only thing we can do
1519 * is returning an error. */
1520 if (server.maxmemory) {
1521 int retval = freeMemoryIfNeeded();
1522 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1523 addReply(c, shared.oomerr);
1524 return REDIS_OK;
1525 }
1526 }
1527
1528 /* Don't accept write commands if there are problems persisting on disk. */
1529 if (server.stop_writes_on_bgsave_err &&
1530 server.saveparamslen > 0
1531 && server.lastbgsave_status == REDIS_ERR &&
1532 c->cmd->flags & REDIS_CMD_WRITE)
1533 {
1534 addReply(c, shared.bgsaveerr);
1535 return REDIS_OK;
1536 }
1537
1538 /* Don't accept wirte commands if this is a read only slave. But
1539 * accept write commands if this is our master. */
1540 if (server.masterhost && server.repl_slave_ro &&
1541 !(c->flags & REDIS_MASTER) &&
1542 c->cmd->flags & REDIS_CMD_WRITE)
1543 {
1544 addReply(c, shared.roslaveerr);
1545 return REDIS_OK;
1546 }
1547
1548 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1549 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1550 &&
1551 c->cmd->proc != subscribeCommand &&
1552 c->cmd->proc != unsubscribeCommand &&
1553 c->cmd->proc != psubscribeCommand &&
1554 c->cmd->proc != punsubscribeCommand) {
1555 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1556 return REDIS_OK;
1557 }
1558
1559 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1560 * we are a slave with a broken link with master. */
1561 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1562 server.repl_serve_stale_data == 0 &&
1563 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1564 {
1565 addReplyError(c,
1566 "link with MASTER is down and slave-serve-stale-data is set to no");
1567 return REDIS_OK;
1568 }
1569
1570 /* Loading DB? Return an error if the command is not INFO */
1571 if (server.loading && c->cmd->proc != infoCommand) {
1572 addReply(c, shared.loadingerr);
1573 return REDIS_OK;
1574 }
1575
1576 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1577 if (server.lua_timedout &&
1578 !(c->cmd->proc != shutdownCommand &&
1579 c->argc == 2 &&
1580 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1581 !(c->cmd->proc == scriptCommand &&
1582 c->argc == 2 &&
1583 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1584 {
1585 addReply(c, shared.slowscripterr);
1586 return REDIS_OK;
1587 }
1588
1589 /* Exec the command */
1590 if (c->flags & REDIS_MULTI &&
1591 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1592 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1593 {
1594 queueMultiCommand(c);
1595 addReply(c,shared.queued);
1596 } else {
1597 call(c,REDIS_CALL_FULL);
1598 }
1599 return REDIS_OK;
1600 }
1601
1602 /*================================== Shutdown =============================== */
1603
1604 int prepareForShutdown(int flags) {
1605 int save = flags & REDIS_SHUTDOWN_SAVE;
1606 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1607
1608 redisLog(REDIS_WARNING,"User requested shutdown...");
1609 /* Kill the saving child if there is a background saving in progress.
1610 We want to avoid race conditions, for instance our saving child may
1611 overwrite the synchronous saving did by SHUTDOWN. */
1612 if (server.rdb_child_pid != -1) {
1613 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1614 kill(server.rdb_child_pid,SIGKILL);
1615 rdbRemoveTempFile(server.rdb_child_pid);
1616 }
1617 if (server.aof_state != REDIS_AOF_OFF) {
1618 /* Kill the AOF saving child as the AOF we already have may be longer
1619 * but contains the full dataset anyway. */
1620 if (server.aof_child_pid != -1) {
1621 redisLog(REDIS_WARNING,
1622 "There is a child rewriting the AOF. Killing it!");
1623 kill(server.aof_child_pid,SIGKILL);
1624 }
1625 /* Append only file: fsync() the AOF and exit */
1626 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1627 aof_fsync(server.aof_fd);
1628 }
1629 if ((server.saveparamslen > 0 && !nosave) || save) {
1630 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1631 /* Snapshotting. Perform a SYNC SAVE and exit */
1632 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1633 /* Ooops.. error saving! The best we can do is to continue
1634 * operating. Note that if there was a background saving process,
1635 * in the next cron() Redis will be notified that the background
1636 * saving aborted, handling special stuff like slaves pending for
1637 * synchronization... */
1638 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1639 return REDIS_ERR;
1640 }
1641 }
1642 if (server.daemonize) {
1643 redisLog(REDIS_NOTICE,"Removing the pid file.");
1644 unlink(server.pidfile);
1645 }
1646 /* Close the listening sockets. Apparently this allows faster restarts. */
1647 if (server.ipfd != -1) close(server.ipfd);
1648 if (server.sofd != -1) close(server.sofd);
1649 if (server.unixsocket) {
1650 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1651 unlink(server.unixsocket); /* don't care if this fails */
1652 }
1653
1654 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1655 return REDIS_OK;
1656 }
1657
1658 /*================================== Commands =============================== */
1659
1660 void authCommand(redisClient *c) {
1661 if (!server.requirepass) {
1662 addReplyError(c,"Client sent AUTH, but no password is set");
1663 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1664 c->authenticated = 1;
1665 addReply(c,shared.ok);
1666 } else {
1667 c->authenticated = 0;
1668 addReplyError(c,"invalid password");
1669 }
1670 }
1671
1672 void pingCommand(redisClient *c) {
1673 addReply(c,shared.pong);
1674 }
1675
1676 void echoCommand(redisClient *c) {
1677 addReplyBulk(c,c->argv[1]);
1678 }
1679
1680 void timeCommand(redisClient *c) {
1681 struct timeval tv;
1682
1683 /* gettimeofday() can only fail if &tv is a bad addresss so we
1684 * don't check for errors. */
1685 gettimeofday(&tv,NULL);
1686 addReplyMultiBulkLen(c,2);
1687 addReplyBulkLongLong(c,tv.tv_sec);
1688 addReplyBulkLongLong(c,tv.tv_usec);
1689 }
1690
1691 /* Convert an amount of bytes into a human readable string in the form
1692 * of 100B, 2G, 100M, 4K, and so forth. */
1693 void bytesToHuman(char *s, unsigned long long n) {
1694 double d;
1695
1696 if (n < 1024) {
1697 /* Bytes */
1698 sprintf(s,"%lluB",n);
1699 return;
1700 } else if (n < (1024*1024)) {
1701 d = (double)n/(1024);
1702 sprintf(s,"%.2fK",d);
1703 } else if (n < (1024LL*1024*1024)) {
1704 d = (double)n/(1024*1024);
1705 sprintf(s,"%.2fM",d);
1706 } else if (n < (1024LL*1024*1024*1024)) {
1707 d = (double)n/(1024LL*1024*1024);
1708 sprintf(s,"%.2fG",d);
1709 }
1710 }
1711
1712 /* Create the string returned by the INFO command. This is decoupled
1713 * by the INFO command itself as we need to report the same information
1714 * on memory corruption problems. */
1715 sds genRedisInfoString(char *section) {
1716 sds info = sdsempty();
1717 time_t uptime = server.unixtime-server.stat_starttime;
1718 int j, numcommands;
1719 struct rusage self_ru, c_ru;
1720 unsigned long lol, bib;
1721 int allsections = 0, defsections = 0;
1722 int sections = 0;
1723
1724 if (section) {
1725 allsections = strcasecmp(section,"all") == 0;
1726 defsections = strcasecmp(section,"default") == 0;
1727 }
1728
1729 getrusage(RUSAGE_SELF, &self_ru);
1730 getrusage(RUSAGE_CHILDREN, &c_ru);
1731 getClientsMaxBuffers(&lol,&bib);
1732
1733 /* Server */
1734 if (allsections || defsections || !strcasecmp(section,"server")) {
1735 if (sections++) info = sdscat(info,"\r\n");
1736 info = sdscatprintf(info,
1737 "# Server\r\n"
1738 "redis_version:%s\r\n"
1739 "redis_git_sha1:%s\r\n"
1740 "redis_git_dirty:%d\r\n"
1741 "arch_bits:%d\r\n"
1742 "multiplexing_api:%s\r\n"
1743 "gcc_version:%d.%d.%d\r\n"
1744 "process_id:%ld\r\n"
1745 "run_id:%s\r\n"
1746 "tcp_port:%d\r\n"
1747 "uptime_in_seconds:%ld\r\n"
1748 "uptime_in_days:%ld\r\n"
1749 "lru_clock:%ld\r\n",
1750 REDIS_VERSION,
1751 redisGitSHA1(),
1752 strtol(redisGitDirty(),NULL,10) > 0,
1753 server.arch_bits,
1754 aeGetApiName(),
1755 #ifdef __GNUC__
1756 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1757 #else
1758 0,0,0,
1759 #endif
1760 (long) getpid(),
1761 server.runid,
1762 server.port,
1763 uptime,
1764 uptime/(3600*24),
1765 (unsigned long) server.lruclock);
1766 }
1767
1768 /* Clients */
1769 if (allsections || defsections || !strcasecmp(section,"clients")) {
1770 if (sections++) info = sdscat(info,"\r\n");
1771 info = sdscatprintf(info,
1772 "# Clients\r\n"
1773 "connected_clients:%lu\r\n"
1774 "client_longest_output_list:%lu\r\n"
1775 "client_biggest_input_buf:%lu\r\n"
1776 "blocked_clients:%d\r\n",
1777 listLength(server.clients)-listLength(server.slaves),
1778 lol, bib,
1779 server.bpop_blocked_clients);
1780 }
1781
1782 /* Memory */
1783 if (allsections || defsections || !strcasecmp(section,"memory")) {
1784 char hmem[64];
1785 char peak_hmem[64];
1786
1787 bytesToHuman(hmem,zmalloc_used_memory());
1788 bytesToHuman(peak_hmem,server.stat_peak_memory);
1789 if (sections++) info = sdscat(info,"\r\n");
1790 info = sdscatprintf(info,
1791 "# Memory\r\n"
1792 "used_memory:%zu\r\n"
1793 "used_memory_human:%s\r\n"
1794 "used_memory_rss:%zu\r\n"
1795 "used_memory_peak:%zu\r\n"
1796 "used_memory_peak_human:%s\r\n"
1797 "used_memory_lua:%lld\r\n"
1798 "mem_fragmentation_ratio:%.2f\r\n"
1799 "mem_allocator:%s\r\n",
1800 zmalloc_used_memory(),
1801 hmem,
1802 zmalloc_get_rss(),
1803 server.stat_peak_memory,
1804 peak_hmem,
1805 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1806 zmalloc_get_fragmentation_ratio(),
1807 ZMALLOC_LIB
1808 );
1809 }
1810
1811 /* Persistence */
1812 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1813 if (sections++) info = sdscat(info,"\r\n");
1814 info = sdscatprintf(info,
1815 "# Persistence\r\n"
1816 "loading:%d\r\n"
1817 "aof_enabled:%d\r\n"
1818 "changes_since_last_save:%lld\r\n"
1819 "bgsave_in_progress:%d\r\n"
1820 "last_save_time:%ld\r\n"
1821 "last_bgsave_status:%s\r\n"
1822 "bgrewriteaof_in_progress:%d\r\n",
1823 server.loading,
1824 server.aof_state != REDIS_AOF_OFF,
1825 server.dirty,
1826 server.rdb_child_pid != -1,
1827 server.lastsave,
1828 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1829 server.aof_child_pid != -1);
1830
1831 if (server.aof_state != REDIS_AOF_OFF) {
1832 info = sdscatprintf(info,
1833 "aof_current_size:%lld\r\n"
1834 "aof_base_size:%lld\r\n"
1835 "aof_pending_rewrite:%d\r\n"
1836 "aof_buffer_length:%zu\r\n"
1837 "aof_pending_bio_fsync:%llu\r\n"
1838 "aof_delayed_fsync:%lu\r\n",
1839 (long long) server.aof_current_size,
1840 (long long) server.aof_rewrite_base_size,
1841 server.aof_rewrite_scheduled,
1842 sdslen(server.aof_buf),
1843 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1844 server.aof_delayed_fsync);
1845 }
1846
1847 if (server.loading) {
1848 double perc;
1849 time_t eta, elapsed;
1850 off_t remaining_bytes = server.loading_total_bytes-
1851 server.loading_loaded_bytes;
1852
1853 perc = ((double)server.loading_loaded_bytes /
1854 server.loading_total_bytes) * 100;
1855
1856 elapsed = server.unixtime-server.loading_start_time;
1857 if (elapsed == 0) {
1858 eta = 1; /* A fake 1 second figure if we don't have
1859 enough info */
1860 } else {
1861 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1862 }
1863
1864 info = sdscatprintf(info,
1865 "loading_start_time:%ld\r\n"
1866 "loading_total_bytes:%llu\r\n"
1867 "loading_loaded_bytes:%llu\r\n"
1868 "loading_loaded_perc:%.2f\r\n"
1869 "loading_eta_seconds:%ld\r\n"
1870 ,(unsigned long) server.loading_start_time,
1871 (unsigned long long) server.loading_total_bytes,
1872 (unsigned long long) server.loading_loaded_bytes,
1873 perc,
1874 eta
1875 );
1876 }
1877 }
1878
1879 /* Stats */
1880 if (allsections || defsections || !strcasecmp(section,"stats")) {
1881 if (sections++) info = sdscat(info,"\r\n");
1882 info = sdscatprintf(info,
1883 "# Stats\r\n"
1884 "total_connections_received:%lld\r\n"
1885 "total_commands_processed:%lld\r\n"
1886 "instantaneous_ops_per_sec:%lld\r\n"
1887 "rejected_connections:%lld\r\n"
1888 "expired_keys:%lld\r\n"
1889 "evicted_keys:%lld\r\n"
1890 "keyspace_hits:%lld\r\n"
1891 "keyspace_misses:%lld\r\n"
1892 "pubsub_channels:%ld\r\n"
1893 "pubsub_patterns:%lu\r\n"
1894 "latest_fork_usec:%lld\r\n",
1895 server.stat_numconnections,
1896 server.stat_numcommands,
1897 getOperationsPerSecond(),
1898 server.stat_rejected_conn,
1899 server.stat_expiredkeys,
1900 server.stat_evictedkeys,
1901 server.stat_keyspace_hits,
1902 server.stat_keyspace_misses,
1903 dictSize(server.pubsub_channels),
1904 listLength(server.pubsub_patterns),
1905 server.stat_fork_time);
1906 }
1907
1908 /* Replication */
1909 if (allsections || defsections || !strcasecmp(section,"replication")) {
1910 if (sections++) info = sdscat(info,"\r\n");
1911 info = sdscatprintf(info,
1912 "# Replication\r\n"
1913 "role:%s\r\n",
1914 server.masterhost == NULL ? "master" : "slave");
1915 if (server.masterhost) {
1916 info = sdscatprintf(info,
1917 "master_host:%s\r\n"
1918 "master_port:%d\r\n"
1919 "master_link_status:%s\r\n"
1920 "master_last_io_seconds_ago:%d\r\n"
1921 "master_sync_in_progress:%d\r\n"
1922 ,server.masterhost,
1923 server.masterport,
1924 (server.repl_state == REDIS_REPL_CONNECTED) ?
1925 "up" : "down",
1926 server.master ?
1927 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
1928 server.repl_state == REDIS_REPL_TRANSFER
1929 );
1930
1931 if (server.repl_state == REDIS_REPL_TRANSFER) {
1932 info = sdscatprintf(info,
1933 "master_sync_left_bytes:%ld\r\n"
1934 "master_sync_last_io_seconds_ago:%d\r\n"
1935 ,(long)server.repl_transfer_left,
1936 (int)(server.unixtime-server.repl_transfer_lastio)
1937 );
1938 }
1939
1940 if (server.repl_state != REDIS_REPL_CONNECTED) {
1941 info = sdscatprintf(info,
1942 "master_link_down_since_seconds:%ld\r\n",
1943 (long)server.unixtime-server.repl_down_since);
1944 }
1945 }
1946 info = sdscatprintf(info,
1947 "connected_slaves:%lu\r\n",
1948 listLength(server.slaves));
1949 if (listLength(server.slaves)) {
1950 int slaveid = 0;
1951 listNode *ln;
1952 listIter li;
1953
1954 listRewind(server.slaves,&li);
1955 while((ln = listNext(&li))) {
1956 redisClient *slave = listNodeValue(ln);
1957 char *state = NULL;
1958 char ip[32];
1959 int port;
1960
1961 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
1962 switch(slave->replstate) {
1963 case REDIS_REPL_WAIT_BGSAVE_START:
1964 case REDIS_REPL_WAIT_BGSAVE_END:
1965 state = "wait_bgsave";
1966 break;
1967 case REDIS_REPL_SEND_BULK:
1968 state = "send_bulk";
1969 break;
1970 case REDIS_REPL_ONLINE:
1971 state = "online";
1972 break;
1973 }
1974 if (state == NULL) continue;
1975 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
1976 slaveid,ip,port,state);
1977 slaveid++;
1978 }
1979 }
1980 }
1981
1982 /* CPU */
1983 if (allsections || defsections || !strcasecmp(section,"cpu")) {
1984 if (sections++) info = sdscat(info,"\r\n");
1985 info = sdscatprintf(info,
1986 "# CPU\r\n"
1987 "used_cpu_sys:%.2f\r\n"
1988 "used_cpu_user:%.2f\r\n"
1989 "used_cpu_sys_children:%.2f\r\n"
1990 "used_cpu_user_children:%.2f\r\n",
1991 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
1992 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
1993 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
1994 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
1995 }
1996
1997 /* cmdtime */
1998 if (allsections || !strcasecmp(section,"commandstats")) {
1999 if (sections++) info = sdscat(info,"\r\n");
2000 info = sdscatprintf(info, "# Commandstats\r\n");
2001 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2002 for (j = 0; j < numcommands; j++) {
2003 struct redisCommand *c = redisCommandTable+j;
2004
2005 if (!c->calls) continue;
2006 info = sdscatprintf(info,
2007 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2008 c->name, c->calls, c->microseconds,
2009 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2010 }
2011 }
2012
2013 /* Clusetr */
2014 if (allsections || defsections || !strcasecmp(section,"cluster")) {
2015 if (sections++) info = sdscat(info,"\r\n");
2016 info = sdscatprintf(info,
2017 "# Cluster\r\n"
2018 "cluster_enabled:%d\r\n",
2019 server.cluster_enabled);
2020 }
2021
2022 /* Key space */
2023 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2024 if (sections++) info = sdscat(info,"\r\n");
2025 info = sdscatprintf(info, "# Keyspace\r\n");
2026 for (j = 0; j < server.dbnum; j++) {
2027 long long keys, vkeys;
2028
2029 keys = dictSize(server.db[j].dict);
2030 vkeys = dictSize(server.db[j].expires);
2031 if (keys || vkeys) {
2032 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2033 j, keys, vkeys);
2034 }
2035 }
2036 }
2037 return info;
2038 }
2039
2040 void infoCommand(redisClient *c) {
2041 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2042
2043 if (c->argc > 2) {
2044 addReply(c,shared.syntaxerr);
2045 return;
2046 }
2047 sds info = genRedisInfoString(section);
2048 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2049 (unsigned long)sdslen(info)));
2050 addReplySds(c,info);
2051 addReply(c,shared.crlf);
2052 }
2053
2054 void monitorCommand(redisClient *c) {
2055 /* ignore MONITOR if aleady slave or in monitor mode */
2056 if (c->flags & REDIS_SLAVE) return;
2057
2058 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2059 c->slaveseldb = 0;
2060 listAddNodeTail(server.monitors,c);
2061 addReply(c,shared.ok);
2062 }
2063
2064 /* ============================ Maxmemory directive ======================== */
2065
2066 /* This function gets called when 'maxmemory' is set on the config file to limit
2067 * the max memory used by the server, before processing a command.
2068 *
2069 * The goal of the function is to free enough memory to keep Redis under the
2070 * configured memory limit.
2071 *
2072 * The function starts calculating how many bytes should be freed to keep
2073 * Redis under the limit, and enters a loop selecting the best keys to
2074 * evict accordingly to the configured policy.
2075 *
2076 * If all the bytes needed to return back under the limit were freed the
2077 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2078 * should block the execution of commands that will result in more memory
2079 * used by the server.
2080 */
2081 int freeMemoryIfNeeded(void) {
2082 size_t mem_used, mem_tofree, mem_freed;
2083 int slaves = listLength(server.slaves);
2084
2085 /* Remove the size of slaves output buffers and AOF buffer from the
2086 * count of used memory. */
2087 mem_used = zmalloc_used_memory();
2088 if (slaves) {
2089 listIter li;
2090 listNode *ln;
2091
2092 listRewind(server.slaves,&li);
2093 while((ln = listNext(&li))) {
2094 redisClient *slave = listNodeValue(ln);
2095 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2096 if (obuf_bytes > mem_used)
2097 mem_used = 0;
2098 else
2099 mem_used -= obuf_bytes;
2100 }
2101 }
2102 if (server.aof_state != REDIS_AOF_OFF) {
2103 mem_used -= sdslen(server.aof_buf);
2104 mem_used -= sdslen(server.aof_rewrite_buf);
2105 }
2106
2107 /* Check if we are over the memory limit. */
2108 if (mem_used <= server.maxmemory) return REDIS_OK;
2109
2110 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2111 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2112
2113 /* Compute how much memory we need to free. */
2114 mem_tofree = mem_used - server.maxmemory;
2115 mem_freed = 0;
2116 while (mem_freed < mem_tofree) {
2117 int j, k, keys_freed = 0;
2118
2119 for (j = 0; j < server.dbnum; j++) {
2120 long bestval = 0; /* just to prevent warning */
2121 sds bestkey = NULL;
2122 struct dictEntry *de;
2123 redisDb *db = server.db+j;
2124 dict *dict;
2125
2126 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2127 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2128 {
2129 dict = server.db[j].dict;
2130 } else {
2131 dict = server.db[j].expires;
2132 }
2133 if (dictSize(dict) == 0) continue;
2134
2135 /* volatile-random and allkeys-random policy */
2136 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2137 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2138 {
2139 de = dictGetRandomKey(dict);
2140 bestkey = dictGetKey(de);
2141 }
2142
2143 /* volatile-lru and allkeys-lru policy */
2144 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2145 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2146 {
2147 for (k = 0; k < server.maxmemory_samples; k++) {
2148 sds thiskey;
2149 long thisval;
2150 robj *o;
2151
2152 de = dictGetRandomKey(dict);
2153 thiskey = dictGetKey(de);
2154 /* When policy is volatile-lru we need an additonal lookup
2155 * to locate the real key, as dict is set to db->expires. */
2156 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2157 de = dictFind(db->dict, thiskey);
2158 o = dictGetVal(de);
2159 thisval = estimateObjectIdleTime(o);
2160
2161 /* Higher idle time is better candidate for deletion */
2162 if (bestkey == NULL || thisval > bestval) {
2163 bestkey = thiskey;
2164 bestval = thisval;
2165 }
2166 }
2167 }
2168
2169 /* volatile-ttl */
2170 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2171 for (k = 0; k < server.maxmemory_samples; k++) {
2172 sds thiskey;
2173 long thisval;
2174
2175 de = dictGetRandomKey(dict);
2176 thiskey = dictGetKey(de);
2177 thisval = (long) dictGetVal(de);
2178
2179 /* Expire sooner (minor expire unix timestamp) is better
2180 * candidate for deletion */
2181 if (bestkey == NULL || thisval < bestval) {
2182 bestkey = thiskey;
2183 bestval = thisval;
2184 }
2185 }
2186 }
2187
2188 /* Finally remove the selected key. */
2189 if (bestkey) {
2190 long long delta;
2191
2192 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2193 propagateExpire(db,keyobj);
2194 /* We compute the amount of memory freed by dbDelete() alone.
2195 * It is possible that actually the memory needed to propagate
2196 * the DEL in AOF and replication link is greater than the one
2197 * we are freeing removing the key, but we can't account for
2198 * that otherwise we would never exit the loop.
2199 *
2200 * AOF and Output buffer memory will be freed eventually so
2201 * we only care about memory used by the key space. */
2202 delta = (long long) zmalloc_used_memory();
2203 dbDelete(db,keyobj);
2204 delta -= (long long) zmalloc_used_memory();
2205 mem_freed += delta;
2206 server.stat_evictedkeys++;
2207 decrRefCount(keyobj);
2208 keys_freed++;
2209
2210 /* When the memory to free starts to be big enough, we may
2211 * start spending so much time here that is impossible to
2212 * deliver data to the slaves fast enough, so we force the
2213 * transmission here inside the loop. */
2214 if (slaves) flushSlavesOutputBuffers();
2215 }
2216 }
2217 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2218 }
2219 return REDIS_OK;
2220 }
2221
2222 /* =================================== Main! ================================ */
2223
2224 #ifdef __linux__
2225 int linuxOvercommitMemoryValue(void) {
2226 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2227 char buf[64];
2228
2229 if (!fp) return -1;
2230 if (fgets(buf,64,fp) == NULL) {
2231 fclose(fp);
2232 return -1;
2233 }
2234 fclose(fp);
2235
2236 return atoi(buf);
2237 }
2238
2239 void linuxOvercommitMemoryWarning(void) {
2240 if (linuxOvercommitMemoryValue() == 0) {
2241 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.");
2242 }
2243 }
2244 #endif /* __linux__ */
2245
2246 void createPidFile(void) {
2247 /* Try to write the pid file in a best-effort way. */
2248 FILE *fp = fopen(server.pidfile,"w");
2249 if (fp) {
2250 fprintf(fp,"%d\n",(int)getpid());
2251 fclose(fp);
2252 }
2253 }
2254
2255 void daemonize(void) {
2256 int fd;
2257
2258 if (fork() != 0) exit(0); /* parent exits */
2259 setsid(); /* create a new session */
2260
2261 /* Every output goes to /dev/null. If Redis is daemonized but
2262 * the 'logfile' is set to 'stdout' in the configuration file
2263 * it will not log at all. */
2264 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2265 dup2(fd, STDIN_FILENO);
2266 dup2(fd, STDOUT_FILENO);
2267 dup2(fd, STDERR_FILENO);
2268 if (fd > STDERR_FILENO) close(fd);
2269 }
2270 }
2271
2272 void version() {
2273 printf("Redis server v=%s sha=%s:%d malloc=%s\n", REDIS_VERSION,
2274 redisGitSHA1(), atoi(redisGitDirty()) > 0, ZMALLOC_LIB);
2275 exit(0);
2276 }
2277
2278 void usage() {
2279 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2280 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2281 fprintf(stderr," ./redis-server -v or --version\n");
2282 fprintf(stderr," ./redis-server -h or --help\n");
2283 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2284 fprintf(stderr,"Examples:\n");
2285 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2286 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2287 fprintf(stderr," ./redis-server --port 7777\n");
2288 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2289 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2290 exit(1);
2291 }
2292
2293 void redisAsciiArt(void) {
2294 #include "asciilogo.h"
2295 char *buf = zmalloc(1024*16);
2296
2297 snprintf(buf,1024*16,ascii_logo,
2298 REDIS_VERSION,
2299 redisGitSHA1(),
2300 strtol(redisGitDirty(),NULL,10) > 0,
2301 (sizeof(long) == 8) ? "64" : "32",
2302 server.cluster_enabled ? "cluster" : "stand alone",
2303 server.port,
2304 (long) getpid()
2305 );
2306 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2307 zfree(buf);
2308 }
2309
2310 static void sigtermHandler(int sig) {
2311 REDIS_NOTUSED(sig);
2312
2313 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2314 server.shutdown_asap = 1;
2315 }
2316
2317 void setupSignalHandlers(void) {
2318 struct sigaction act;
2319
2320 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2321 * Otherwise, sa_handler is used. */
2322 sigemptyset(&act.sa_mask);
2323 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
2324 act.sa_handler = sigtermHandler;
2325 sigaction(SIGTERM, &act, NULL);
2326
2327 #ifdef HAVE_BACKTRACE
2328 sigemptyset(&act.sa_mask);
2329 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
2330 act.sa_sigaction = sigsegvHandler;
2331 sigaction(SIGSEGV, &act, NULL);
2332 sigaction(SIGBUS, &act, NULL);
2333 sigaction(SIGFPE, &act, NULL);
2334 sigaction(SIGILL, &act, NULL);
2335 #endif
2336 return;
2337 }
2338
2339 void memtest(size_t megabytes, int passes);
2340
2341 int main(int argc, char **argv) {
2342 long long start;
2343 struct timeval tv;
2344
2345 /* We need to initialize our libraries, and the server configuration. */
2346 zmalloc_enable_thread_safeness();
2347 srand(time(NULL)^getpid());
2348 gettimeofday(&tv,NULL);
2349 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2350 initServerConfig();
2351
2352 if (argc >= 2) {
2353 int j = 1; /* First option to parse in argv[] */
2354 sds options = sdsempty();
2355 char *configfile = NULL;
2356
2357 /* Handle special options --help and --version */
2358 if (strcmp(argv[1], "-v") == 0 ||
2359 strcmp(argv[1], "--version") == 0) version();
2360 if (strcmp(argv[1], "--help") == 0 ||
2361 strcmp(argv[1], "-h") == 0) usage();
2362 if (strcmp(argv[1], "--test-memory") == 0) {
2363 if (argc == 3) {
2364 memtest(atoi(argv[2]),50);
2365 exit(0);
2366 } else {
2367 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2368 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2369 exit(1);
2370 }
2371 }
2372
2373 /* First argument is the config file name? */
2374 if (argv[j][0] != '-' || argv[j][1] != '-')
2375 configfile = argv[j++];
2376 /* All the other options are parsed and conceptually appended to the
2377 * configuration file. For instance --port 6380 will generate the
2378 * string "port 6380\n" to be parsed after the actual file name
2379 * is parsed, if any. */
2380 while(j != argc) {
2381 if (argv[j][0] == '-' && argv[j][1] == '-') {
2382 /* Option name */
2383 if (sdslen(options)) options = sdscat(options,"\n");
2384 options = sdscat(options,argv[j]+2);
2385 options = sdscat(options," ");
2386 } else {
2387 /* Option argument */
2388 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2389 options = sdscat(options," ");
2390 }
2391 j++;
2392 }
2393 resetServerSaveParams();
2394 loadServerConfig(configfile,options);
2395 sdsfree(options);
2396 } else {
2397 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'");
2398 }
2399 if (server.daemonize) daemonize();
2400 initServer();
2401 if (server.daemonize) createPidFile();
2402 redisAsciiArt();
2403 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2404 #ifdef __linux__
2405 linuxOvercommitMemoryWarning();
2406 #endif
2407 start = ustime();
2408 if (server.aof_state == REDIS_AOF_ON) {
2409 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2410 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2411 } else {
2412 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2413 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2414 (float)(ustime()-start)/1000000);
2415 } else if (errno != ENOENT) {
2416 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2417 exit(1);
2418 }
2419 }
2420 if (server.ipfd > 0)
2421 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2422 if (server.sofd > 0)
2423 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2424 aeSetBeforeSleepProc(server.el,beforeSleep);
2425 aeMain(server.el);
2426 aeDeleteEventLoop(server.el);
2427 return 0;
2428 }
2429
2430 /* The End */