]> git.saurik.com Git - redis.git/blob - src/redis.c
5ee44a31932165257153f24974e6c0bdde086eea
[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 /* Redis generally does not try to recover from out of memory conditions
297 * when allocating objects or strings, it is not clear if it will be possible
298 * to report this condition to the client since the networking layer itself
299 * is based on heap allocation for send buffers, so we simply abort.
300 * At least the code will be simpler to read... */
301 void oom(const char *msg) {
302 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
303 sleep(1);
304 abort();
305 }
306
307 /* Return the UNIX time in microseconds */
308 long long ustime(void) {
309 struct timeval tv;
310 long long ust;
311
312 gettimeofday(&tv, NULL);
313 ust = ((long long)tv.tv_sec)*1000000;
314 ust += tv.tv_usec;
315 return ust;
316 }
317
318 /* Return the UNIX time in milliseconds */
319 long long mstime(void) {
320 return ustime()/1000;
321 }
322
323 /*====================== Hash table type implementation ==================== */
324
325 /* This is an hash table type that uses the SDS dynamic strings libary as
326 * keys and radis objects as values (objects can hold SDS strings,
327 * lists, sets). */
328
329 void dictVanillaFree(void *privdata, void *val)
330 {
331 DICT_NOTUSED(privdata);
332 zfree(val);
333 }
334
335 void dictListDestructor(void *privdata, void *val)
336 {
337 DICT_NOTUSED(privdata);
338 listRelease((list*)val);
339 }
340
341 int dictSdsKeyCompare(void *privdata, const void *key1,
342 const void *key2)
343 {
344 int l1,l2;
345 DICT_NOTUSED(privdata);
346
347 l1 = sdslen((sds)key1);
348 l2 = sdslen((sds)key2);
349 if (l1 != l2) return 0;
350 return memcmp(key1, key2, l1) == 0;
351 }
352
353 /* A case insensitive version used for the command lookup table. */
354 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
355 const void *key2)
356 {
357 DICT_NOTUSED(privdata);
358
359 return strcasecmp(key1, key2) == 0;
360 }
361
362 void dictRedisObjectDestructor(void *privdata, void *val)
363 {
364 DICT_NOTUSED(privdata);
365
366 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
367 decrRefCount(val);
368 }
369
370 void dictSdsDestructor(void *privdata, void *val)
371 {
372 DICT_NOTUSED(privdata);
373
374 sdsfree(val);
375 }
376
377 int dictObjKeyCompare(void *privdata, const void *key1,
378 const void *key2)
379 {
380 const robj *o1 = key1, *o2 = key2;
381 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
382 }
383
384 unsigned int dictObjHash(const void *key) {
385 const robj *o = key;
386 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
387 }
388
389 unsigned int dictSdsHash(const void *key) {
390 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
391 }
392
393 unsigned int dictSdsCaseHash(const void *key) {
394 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
395 }
396
397 int dictEncObjKeyCompare(void *privdata, const void *key1,
398 const void *key2)
399 {
400 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
401 int cmp;
402
403 if (o1->encoding == REDIS_ENCODING_INT &&
404 o2->encoding == REDIS_ENCODING_INT)
405 return o1->ptr == o2->ptr;
406
407 o1 = getDecodedObject(o1);
408 o2 = getDecodedObject(o2);
409 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
410 decrRefCount(o1);
411 decrRefCount(o2);
412 return cmp;
413 }
414
415 unsigned int dictEncObjHash(const void *key) {
416 robj *o = (robj*) key;
417
418 if (o->encoding == REDIS_ENCODING_RAW) {
419 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
420 } else {
421 if (o->encoding == REDIS_ENCODING_INT) {
422 char buf[32];
423 int len;
424
425 len = ll2string(buf,32,(long)o->ptr);
426 return dictGenHashFunction((unsigned char*)buf, len);
427 } else {
428 unsigned int hash;
429
430 o = getDecodedObject(o);
431 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
432 decrRefCount(o);
433 return hash;
434 }
435 }
436 }
437
438 /* Sets type hash table */
439 dictType setDictType = {
440 dictEncObjHash, /* hash function */
441 NULL, /* key dup */
442 NULL, /* val dup */
443 dictEncObjKeyCompare, /* key compare */
444 dictRedisObjectDestructor, /* key destructor */
445 NULL /* val destructor */
446 };
447
448 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
449 dictType zsetDictType = {
450 dictEncObjHash, /* hash function */
451 NULL, /* key dup */
452 NULL, /* val dup */
453 dictEncObjKeyCompare, /* key compare */
454 dictRedisObjectDestructor, /* key destructor */
455 NULL /* val destructor */
456 };
457
458 /* Db->dict, keys are sds strings, vals are Redis objects. */
459 dictType dbDictType = {
460 dictSdsHash, /* hash function */
461 NULL, /* key dup */
462 NULL, /* val dup */
463 dictSdsKeyCompare, /* key compare */
464 dictSdsDestructor, /* key destructor */
465 dictRedisObjectDestructor /* val destructor */
466 };
467
468 /* Db->expires */
469 dictType keyptrDictType = {
470 dictSdsHash, /* hash function */
471 NULL, /* key dup */
472 NULL, /* val dup */
473 dictSdsKeyCompare, /* key compare */
474 NULL, /* key destructor */
475 NULL /* val destructor */
476 };
477
478 /* Command table. sds string -> command struct pointer. */
479 dictType commandTableDictType = {
480 dictSdsCaseHash, /* hash function */
481 NULL, /* key dup */
482 NULL, /* val dup */
483 dictSdsKeyCaseCompare, /* key compare */
484 dictSdsDestructor, /* key destructor */
485 NULL /* val destructor */
486 };
487
488 /* Hash type hash table (note that small hashes are represented with zimpaps) */
489 dictType hashDictType = {
490 dictEncObjHash, /* hash function */
491 NULL, /* key dup */
492 NULL, /* val dup */
493 dictEncObjKeyCompare, /* key compare */
494 dictRedisObjectDestructor, /* key destructor */
495 dictRedisObjectDestructor /* val destructor */
496 };
497
498 /* Keylist hash table type has unencoded redis objects as keys and
499 * lists as values. It's used for blocking operations (BLPOP) and to
500 * map swapped keys to a list of clients waiting for this keys to be loaded. */
501 dictType keylistDictType = {
502 dictObjHash, /* hash function */
503 NULL, /* key dup */
504 NULL, /* val dup */
505 dictObjKeyCompare, /* key compare */
506 dictRedisObjectDestructor, /* key destructor */
507 dictListDestructor /* val destructor */
508 };
509
510 /* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
511 * clusterNode structures. */
512 dictType clusterNodesDictType = {
513 dictSdsHash, /* hash function */
514 NULL, /* key dup */
515 NULL, /* val dup */
516 dictSdsKeyCompare, /* key compare */
517 dictSdsDestructor, /* key destructor */
518 NULL /* val destructor */
519 };
520
521 int htNeedsResize(dict *dict) {
522 long long size, used;
523
524 size = dictSlots(dict);
525 used = dictSize(dict);
526 return (size && used && size > DICT_HT_INITIAL_SIZE &&
527 (used*100/size < REDIS_HT_MINFILL));
528 }
529
530 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
531 * we resize the hash table to save memory */
532 void tryResizeHashTables(void) {
533 int j;
534
535 for (j = 0; j < server.dbnum; j++) {
536 if (htNeedsResize(server.db[j].dict))
537 dictResize(server.db[j].dict);
538 if (htNeedsResize(server.db[j].expires))
539 dictResize(server.db[j].expires);
540 }
541 }
542
543 /* Our hash table implementation performs rehashing incrementally while
544 * we write/read from the hash table. Still if the server is idle, the hash
545 * table will use two tables for a long time. So we try to use 1 millisecond
546 * of CPU time at every serverCron() loop in order to rehash some key. */
547 void incrementallyRehash(void) {
548 int j;
549
550 for (j = 0; j < server.dbnum; j++) {
551 if (dictIsRehashing(server.db[j].dict)) {
552 dictRehashMilliseconds(server.db[j].dict,1);
553 break; /* already used our millisecond for this loop... */
554 }
555 }
556 }
557
558 /* This function is called once a background process of some kind terminates,
559 * as we want to avoid resizing the hash tables when there is a child in order
560 * to play well with copy-on-write (otherwise when a resize happens lots of
561 * memory pages are copied). The goal of this function is to update the ability
562 * for dict.c to resize the hash tables accordingly to the fact we have o not
563 * running childs. */
564 void updateDictResizePolicy(void) {
565 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
566 dictEnableResize();
567 else
568 dictDisableResize();
569 }
570
571 /* ======================= Cron: called every 100 ms ======================== */
572
573 /* Try to expire a few timed out keys. The algorithm used is adaptive and
574 * will use few CPU cycles if there are few expiring keys, otherwise
575 * it will get more aggressive to avoid that too much memory is used by
576 * keys that can be removed from the keyspace. */
577 void activeExpireCycle(void) {
578 int j;
579
580 for (j = 0; j < server.dbnum; j++) {
581 int expired;
582 redisDb *db = server.db+j;
583
584 /* Continue to expire if at the end of the cycle more than 25%
585 * of the keys were expired. */
586 do {
587 long num = dictSize(db->expires);
588 long long now = mstime();
589
590 expired = 0;
591 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
592 num = REDIS_EXPIRELOOKUPS_PER_CRON;
593 while (num--) {
594 dictEntry *de;
595 long long t;
596
597 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
598 t = dictGetSignedIntegerVal(de);
599 if (now > t) {
600 sds key = dictGetKey(de);
601 robj *keyobj = createStringObject(key,sdslen(key));
602
603 propagateExpire(db,keyobj);
604 dbDelete(db,keyobj);
605 decrRefCount(keyobj);
606 expired++;
607 server.stat_expiredkeys++;
608 }
609 }
610 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
611 }
612 }
613
614 void updateLRUClock(void) {
615 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
616 REDIS_LRU_CLOCK_MAX;
617 }
618
619
620 /* Add a sample to the operations per second array of samples. */
621 void trackOperationsPerSecond(void) {
622 long long t = mstime() - server.ops_sec_last_sample_time;
623 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
624 long long ops_sec;
625
626 ops_sec = t > 0 ? (ops*1000/t) : 0;
627
628 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
629 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
630 server.ops_sec_last_sample_time = mstime();
631 server.ops_sec_last_sample_ops = server.stat_numcommands;
632 }
633
634 /* Return the mean of all the samples. */
635 long long getOperationsPerSecond(void) {
636 int j;
637 long long sum = 0;
638
639 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
640 sum += server.ops_sec_samples[j];
641 return sum / REDIS_OPS_SEC_SAMPLES;
642 }
643
644 /* Check for timeouts. Returns non-zero if the client was terminated */
645 int clientsCronHandleTimeout(redisClient *c) {
646 time_t now = server.unixtime;
647
648 if (server.maxidletime &&
649 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
650 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
651 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
652 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
653 listLength(c->pubsub_patterns) == 0 &&
654 (now - c->lastinteraction > server.maxidletime))
655 {
656 redisLog(REDIS_VERBOSE,"Closing idle client");
657 freeClient(c);
658 return 1;
659 } else if (c->flags & REDIS_BLOCKED) {
660 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
661 addReply(c,shared.nullmultibulk);
662 unblockClientWaitingData(c);
663 }
664 }
665 return 0;
666 }
667
668 /* The client query buffer is an sds.c string that can end with a lot of
669 * free space not used, this function reclaims space if needed.
670 *
671 * The funciton always returns 0 as it never terminates the client. */
672 int clientsCronResizeQueryBuffer(redisClient *c) {
673 size_t querybuf_size = sdsAllocSize(c->querybuf);
674 time_t idletime = server.unixtime - c->lastinteraction;
675
676 /* There are two conditions to resize the query buffer:
677 * 1) Query buffer is > BIG_ARG and too big for latest peak.
678 * 2) Client is inactive and the buffer is bigger than 1k. */
679 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
680 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
681 (querybuf_size > 1024 && idletime > 2))
682 {
683 /* Only resize the query buffer if it is actually wasting space. */
684 if (sdsavail(c->querybuf) > 1024) {
685 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
686 }
687 }
688 /* Reset the peak again to capture the peak memory usage in the next
689 * cycle. */
690 c->querybuf_peak = 0;
691 return 0;
692 }
693
694 void clientsCron(void) {
695 /* Make sure to process at least 1/100 of clients per call.
696 * Since this function is called 10 times per second we are sure that
697 * in the worst case we process all the clients in 10 seconds.
698 * In normal conditions (a reasonable number of clients) we process
699 * all the clients in a shorter time. */
700 int numclients = listLength(server.clients);
701 int iterations = numclients/100;
702
703 if (iterations < 50)
704 iterations = (numclients < 50) ? numclients : 50;
705 while(listLength(server.clients) && iterations--) {
706 redisClient *c;
707 listNode *head;
708
709 /* Rotate the list, take the current head, process.
710 * This way if the client must be removed from the list it's the
711 * first element and we don't incur into O(N) computation. */
712 listRotate(server.clients);
713 head = listFirst(server.clients);
714 c = listNodeValue(head);
715 /* The following functions do different service checks on the client.
716 * The protocol is that they return non-zero if the client was
717 * terminated. */
718 if (clientsCronHandleTimeout(c)) continue;
719 if (clientsCronResizeQueryBuffer(c)) continue;
720 }
721 }
722
723 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
724 int j, loops = server.cronloops;
725 REDIS_NOTUSED(eventLoop);
726 REDIS_NOTUSED(id);
727 REDIS_NOTUSED(clientData);
728
729 /* Software watchdog: deliver the SIGALRM that will reach the signal
730 * handler if we don't return here fast enough. */
731 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
732
733 /* We take a cached value of the unix time in the global state because
734 * with virtual memory and aging there is to store the current time
735 * in objects at every object access, and accuracy is not needed.
736 * To access a global var is faster than calling time(NULL) */
737 server.unixtime = time(NULL);
738
739 trackOperationsPerSecond();
740
741 /* We have just 22 bits per object for LRU information.
742 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
743 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
744 *
745 * Note that even if this will wrap after 1.5 years it's not a problem,
746 * everything will still work but just some object will appear younger
747 * to Redis. But for this to happen a given object should never be touched
748 * for 1.5 years.
749 *
750 * Note that you can change the resolution altering the
751 * REDIS_LRU_CLOCK_RESOLUTION define.
752 */
753 updateLRUClock();
754
755 /* Record the max memory used since the server was started. */
756 if (zmalloc_used_memory() > server.stat_peak_memory)
757 server.stat_peak_memory = zmalloc_used_memory();
758
759 /* We received a SIGTERM, shutting down here in a safe way, as it is
760 * not ok doing so inside the signal handler. */
761 if (server.shutdown_asap) {
762 if (prepareForShutdown(0) == REDIS_OK) exit(0);
763 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
764 }
765
766 /* Show some info about non-empty databases */
767 for (j = 0; j < server.dbnum; j++) {
768 long long size, used, vkeys;
769
770 size = dictSlots(server.db[j].dict);
771 used = dictSize(server.db[j].dict);
772 vkeys = dictSize(server.db[j].expires);
773 if (!(loops % 50) && (used || vkeys)) {
774 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
775 /* dictPrintStats(server.dict); */
776 }
777 }
778
779 /* We don't want to resize the hash tables while a bacground saving
780 * is in progress: the saving child is created using fork() that is
781 * implemented with a copy-on-write semantic in most modern systems, so
782 * if we resize the HT while there is the saving child at work actually
783 * a lot of memory movements in the parent will cause a lot of pages
784 * copied. */
785 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
786 if (!(loops % 10)) tryResizeHashTables();
787 if (server.activerehashing) incrementallyRehash();
788 }
789
790 /* Show information about connected clients */
791 if (!(loops % 50)) {
792 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
793 listLength(server.clients)-listLength(server.slaves),
794 listLength(server.slaves),
795 zmalloc_used_memory());
796 }
797
798 /* We need to do a few operations on clients asynchronously. */
799 clientsCron();
800
801 /* Start a scheduled AOF rewrite if this was requested by the user while
802 * a BGSAVE was in progress. */
803 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
804 server.aof_rewrite_scheduled)
805 {
806 rewriteAppendOnlyFileBackground();
807 }
808
809 /* Check if a background saving or AOF rewrite in progress terminated. */
810 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
811 int statloc;
812 pid_t pid;
813
814 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
815 int exitcode = WEXITSTATUS(statloc);
816 int bysignal = 0;
817
818 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
819
820 if (pid == server.rdb_child_pid) {
821 backgroundSaveDoneHandler(exitcode,bysignal);
822 } else {
823 backgroundRewriteDoneHandler(exitcode,bysignal);
824 }
825 updateDictResizePolicy();
826 }
827 } else {
828 /* If there is not a background saving/rewrite in progress check if
829 * we have to save/rewrite now */
830 for (j = 0; j < server.saveparamslen; j++) {
831 struct saveparam *sp = server.saveparams+j;
832
833 if (server.dirty >= sp->changes &&
834 server.unixtime-server.lastsave > sp->seconds) {
835 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
836 sp->changes, sp->seconds);
837 rdbSaveBackground(server.rdb_filename);
838 break;
839 }
840 }
841
842 /* Trigger an AOF rewrite if needed */
843 if (server.rdb_child_pid == -1 &&
844 server.aof_child_pid == -1 &&
845 server.aof_rewrite_perc &&
846 server.aof_current_size > server.aof_rewrite_min_size)
847 {
848 long long base = server.aof_rewrite_base_size ?
849 server.aof_rewrite_base_size : 1;
850 long long growth = (server.aof_current_size*100/base) - 100;
851 if (growth >= server.aof_rewrite_perc) {
852 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
853 rewriteAppendOnlyFileBackground();
854 }
855 }
856 }
857
858
859 /* If we postponed an AOF buffer flush, let's try to do it every time the
860 * cron function is called. */
861 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
862
863 /* Expire a few keys per cycle, only if this is a master.
864 * On slaves we wait for DEL operations synthesized by the master
865 * in order to guarantee a strict consistency. */
866 if (server.masterhost == NULL) activeExpireCycle();
867
868 /* Close clients that need to be closed asynchronous */
869 freeClientsInAsyncFreeQueue();
870
871 /* Replication cron function -- used to reconnect to master and
872 * to detect transfer failures. */
873 if (!(loops % 10)) replicationCron();
874
875 /* Run other sub-systems specific cron jobs */
876 if (server.cluster_enabled && !(loops % 10)) clusterCron();
877
878 server.cronloops++;
879 return 100;
880 }
881
882 /* This function gets called every time Redis is entering the
883 * main loop of the event driven library, that is, before to sleep
884 * for ready file descriptors. */
885 void beforeSleep(struct aeEventLoop *eventLoop) {
886 REDIS_NOTUSED(eventLoop);
887 listNode *ln;
888 redisClient *c;
889
890 /* Try to process pending commands for clients that were just unblocked. */
891 while (listLength(server.unblocked_clients)) {
892 ln = listFirst(server.unblocked_clients);
893 redisAssert(ln != NULL);
894 c = ln->value;
895 listDelNode(server.unblocked_clients,ln);
896 c->flags &= ~REDIS_UNBLOCKED;
897
898 /* Process remaining data in the input buffer. */
899 if (c->querybuf && sdslen(c->querybuf) > 0) {
900 server.current_client = c;
901 processInputBuffer(c);
902 server.current_client = NULL;
903 }
904 }
905
906 /* Write the AOF buffer on disk */
907 flushAppendOnlyFile(0);
908 }
909
910 /* =========================== Server initialization ======================== */
911
912 void createSharedObjects(void) {
913 int j;
914
915 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
916 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
917 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
918 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
919 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
920 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
921 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
922 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
923 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
924 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
925 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
926 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
927 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
928 "-ERR Operation against a key holding the wrong kind of value\r\n"));
929 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
930 "-ERR no such key\r\n"));
931 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
932 "-ERR syntax error\r\n"));
933 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
934 "-ERR source and destination objects are the same\r\n"));
935 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
936 "-ERR index out of range\r\n"));
937 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
938 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
939 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
940 "-LOADING Redis is loading the dataset in memory\r\n"));
941 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
942 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
943 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
944 "-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"));
945 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
946 "-READONLY You can't write against a read only slave.\r\n"));
947 shared.oomerr = createObject(REDIS_STRING,sdsnew(
948 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
949 shared.space = createObject(REDIS_STRING,sdsnew(" "));
950 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
951 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
952 shared.select0 = createStringObject("select 0\r\n",10);
953 shared.select1 = createStringObject("select 1\r\n",10);
954 shared.select2 = createStringObject("select 2\r\n",10);
955 shared.select3 = createStringObject("select 3\r\n",10);
956 shared.select4 = createStringObject("select 4\r\n",10);
957 shared.select5 = createStringObject("select 5\r\n",10);
958 shared.select6 = createStringObject("select 6\r\n",10);
959 shared.select7 = createStringObject("select 7\r\n",10);
960 shared.select8 = createStringObject("select 8\r\n",10);
961 shared.select9 = createStringObject("select 9\r\n",10);
962 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
963 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
964 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
965 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
966 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
967 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
968 shared.del = createStringObject("DEL",3);
969 shared.rpop = createStringObject("RPOP",4);
970 shared.lpop = createStringObject("LPOP",4);
971 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
972 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
973 shared.integers[j]->encoding = REDIS_ENCODING_INT;
974 }
975 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
976 shared.mbulkhdr[j] = createObject(REDIS_STRING,
977 sdscatprintf(sdsempty(),"*%d\r\n",j));
978 shared.bulkhdr[j] = createObject(REDIS_STRING,
979 sdscatprintf(sdsempty(),"$%d\r\n",j));
980 }
981 }
982
983 void initServerConfig() {
984 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
985 server.runid[REDIS_RUN_ID_SIZE] = '\0';
986 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
987 server.port = REDIS_SERVERPORT;
988 server.bindaddr = NULL;
989 server.unixsocket = NULL;
990 server.unixsocketperm = 0;
991 server.ipfd = -1;
992 server.sofd = -1;
993 server.dbnum = REDIS_DEFAULT_DBNUM;
994 server.verbosity = REDIS_NOTICE;
995 server.maxidletime = REDIS_MAXIDLETIME;
996 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
997 server.saveparams = NULL;
998 server.loading = 0;
999 server.logfile = NULL; /* NULL = log on standard output */
1000 server.syslog_enabled = 0;
1001 server.syslog_ident = zstrdup("redis");
1002 server.syslog_facility = LOG_LOCAL0;
1003 server.daemonize = 0;
1004 server.aof_state = REDIS_AOF_OFF;
1005 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1006 server.aof_no_fsync_on_rewrite = 0;
1007 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1008 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1009 server.aof_rewrite_base_size = 0;
1010 server.aof_rewrite_scheduled = 0;
1011 server.aof_last_fsync = time(NULL);
1012 server.aof_delayed_fsync = 0;
1013 server.aof_fd = -1;
1014 server.aof_selected_db = -1; /* Make sure the first time will not match */
1015 server.aof_flush_postponed_start = 0;
1016 server.pidfile = zstrdup("/var/run/redis.pid");
1017 server.rdb_filename = zstrdup("dump.rdb");
1018 server.aof_filename = zstrdup("appendonly.aof");
1019 server.requirepass = NULL;
1020 server.rdb_compression = 1;
1021 server.activerehashing = 1;
1022 server.maxclients = REDIS_MAX_CLIENTS;
1023 server.bpop_blocked_clients = 0;
1024 server.maxmemory = 0;
1025 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1026 server.maxmemory_samples = 3;
1027 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1028 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1029 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1030 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1031 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1032 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1033 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1034 server.shutdown_asap = 0;
1035 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1036 server.repl_timeout = REDIS_REPL_TIMEOUT;
1037 server.cluster_enabled = 0;
1038 server.cluster.configfile = zstrdup("nodes.conf");
1039 server.lua_caller = NULL;
1040 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1041 server.lua_client = NULL;
1042 server.lua_timedout = 0;
1043
1044 updateLRUClock();
1045 resetServerSaveParams();
1046
1047 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1048 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1049 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1050 /* Replication related */
1051 server.masterauth = NULL;
1052 server.masterhost = NULL;
1053 server.masterport = 6379;
1054 server.master = NULL;
1055 server.repl_state = REDIS_REPL_NONE;
1056 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1057 server.repl_serve_stale_data = 1;
1058 server.repl_slave_ro = 1;
1059 server.repl_down_since = -1;
1060
1061 /* Client output buffer limits */
1062 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1063 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1064 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1065 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1066 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1067 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1068 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1069 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1070 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1071
1072 /* Double constants initialization */
1073 R_Zero = 0.0;
1074 R_PosInf = 1.0/R_Zero;
1075 R_NegInf = -1.0/R_Zero;
1076 R_Nan = R_Zero/R_Zero;
1077
1078 /* Command table -- we intiialize it here as it is part of the
1079 * initial configuration, since command names may be changed via
1080 * redis.conf using the rename-command directive. */
1081 server.commands = dictCreate(&commandTableDictType,NULL);
1082 populateCommandTable();
1083 server.delCommand = lookupCommandByCString("del");
1084 server.multiCommand = lookupCommandByCString("multi");
1085 server.lpushCommand = lookupCommandByCString("lpush");
1086
1087 /* Slow log */
1088 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1089 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1090
1091 /* Debugging */
1092 server.assert_failed = "<no assertion failed>";
1093 server.assert_file = "<no file>";
1094 server.assert_line = 0;
1095 server.bug_report_start = 0;
1096 server.watchdog_period = 0;
1097 }
1098
1099 /* This function will try to raise the max number of open files accordingly to
1100 * the configured max number of clients. It will also account for 32 additional
1101 * file descriptors as we need a few more for persistence, listening
1102 * sockets, log files and so forth.
1103 *
1104 * If it will not be possible to set the limit accordingly to the configured
1105 * max number of clients, the function will do the reverse setting
1106 * server.maxclients to the value that we can actually handle. */
1107 void adjustOpenFilesLimit(void) {
1108 rlim_t maxfiles = server.maxclients+32;
1109 struct rlimit limit;
1110
1111 if (maxfiles < 1024) maxfiles = 1024;
1112 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1113 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1114 strerror(errno));
1115 server.maxclients = 1024-32;
1116 } else {
1117 rlim_t oldlimit = limit.rlim_cur;
1118
1119 /* Set the max number of files if the current limit is not enough
1120 * for our needs. */
1121 if (oldlimit < maxfiles) {
1122 limit.rlim_cur = maxfiles;
1123 limit.rlim_max = maxfiles;
1124 if (setrlimit(RLIMIT_NOFILE,&limit) == -1) {
1125 server.maxclients = oldlimit-32;
1126 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1127 (int) maxfiles, strerror(errno), (int) server.maxclients);
1128 } else {
1129 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1130 (int) maxfiles);
1131 }
1132 }
1133 }
1134 }
1135
1136 void initServer() {
1137 int j;
1138
1139 signal(SIGHUP, SIG_IGN);
1140 signal(SIGPIPE, SIG_IGN);
1141 setupSignalHandlers();
1142
1143 if (server.syslog_enabled) {
1144 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1145 server.syslog_facility);
1146 }
1147
1148 server.current_client = NULL;
1149 server.clients = listCreate();
1150 server.clients_to_close = listCreate();
1151 server.slaves = listCreate();
1152 server.monitors = listCreate();
1153 server.unblocked_clients = listCreate();
1154
1155 createSharedObjects();
1156 adjustOpenFilesLimit();
1157 server.el = aeCreateEventLoop(server.maxclients+1024);
1158 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1159
1160 if (server.port != 0) {
1161 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1162 if (server.ipfd == ANET_ERR) {
1163 redisLog(REDIS_WARNING, "Opening port %d: %s",
1164 server.port, server.neterr);
1165 exit(1);
1166 }
1167 }
1168 if (server.unixsocket != NULL) {
1169 unlink(server.unixsocket); /* don't care if this fails */
1170 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1171 if (server.sofd == ANET_ERR) {
1172 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1173 exit(1);
1174 }
1175 }
1176 if (server.ipfd < 0 && server.sofd < 0) {
1177 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1178 exit(1);
1179 }
1180 for (j = 0; j < server.dbnum; j++) {
1181 server.db[j].dict = dictCreate(&dbDictType,NULL);
1182 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1183 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1184 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1185 server.db[j].id = j;
1186 }
1187 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1188 server.pubsub_patterns = listCreate();
1189 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1190 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1191 server.cronloops = 0;
1192 server.rdb_child_pid = -1;
1193 server.aof_child_pid = -1;
1194 server.aof_rewrite_buf = sdsempty();
1195 server.aof_buf = sdsempty();
1196 server.lastsave = time(NULL);
1197 server.dirty = 0;
1198 server.stat_numcommands = 0;
1199 server.stat_numconnections = 0;
1200 server.stat_expiredkeys = 0;
1201 server.stat_evictedkeys = 0;
1202 server.stat_starttime = time(NULL);
1203 server.stat_keyspace_misses = 0;
1204 server.stat_keyspace_hits = 0;
1205 server.stat_peak_memory = 0;
1206 server.stat_fork_time = 0;
1207 server.stat_rejected_conn = 0;
1208 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1209 server.ops_sec_idx = 0;
1210 server.ops_sec_last_sample_time = mstime();
1211 server.ops_sec_last_sample_ops = 0;
1212 server.unixtime = time(NULL);
1213 server.lastbgsave_status = REDIS_OK;
1214 server.stop_writes_on_bgsave_err = 1;
1215 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1216 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1217 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1218 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1219 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1220
1221 if (server.aof_state == REDIS_AOF_ON) {
1222 server.aof_fd = open(server.aof_filename,
1223 O_WRONLY|O_APPEND|O_CREAT,0644);
1224 if (server.aof_fd == -1) {
1225 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1226 strerror(errno));
1227 exit(1);
1228 }
1229 }
1230
1231 /* 32 bit instances are limited to 4GB of address space, so if there is
1232 * no explicit limit in the user provided configuration we set a limit
1233 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1234 * useless crashes of the Redis instance. */
1235 if (server.arch_bits == 32 && server.maxmemory == 0) {
1236 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1237 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1238 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1239 }
1240
1241 if (server.cluster_enabled) clusterInit();
1242 scriptingInit();
1243 slowlogInit();
1244 bioInit();
1245 }
1246
1247 /* Populates the Redis Command Table starting from the hard coded list
1248 * we have on top of redis.c file. */
1249 void populateCommandTable(void) {
1250 int j;
1251 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1252
1253 for (j = 0; j < numcommands; j++) {
1254 struct redisCommand *c = redisCommandTable+j;
1255 char *f = c->sflags;
1256 int retval;
1257
1258 while(*f != '\0') {
1259 switch(*f) {
1260 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1261 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1262 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1263 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1264 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1265 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1266 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1267 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1268 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1269 default: redisPanic("Unsupported command flag"); break;
1270 }
1271 f++;
1272 }
1273
1274 retval = dictAdd(server.commands, sdsnew(c->name), c);
1275 assert(retval == DICT_OK);
1276 }
1277 }
1278
1279 void resetCommandTableStats(void) {
1280 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1281 int j;
1282
1283 for (j = 0; j < numcommands; j++) {
1284 struct redisCommand *c = redisCommandTable+j;
1285
1286 c->microseconds = 0;
1287 c->calls = 0;
1288 }
1289 }
1290
1291 /* ========================== Redis OP Array API ============================ */
1292
1293 void redisOpArrayInit(redisOpArray *oa) {
1294 oa->ops = NULL;
1295 oa->numops = 0;
1296 }
1297
1298 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1299 robj **argv, int argc, int target)
1300 {
1301 redisOp *op;
1302
1303 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1304 op = oa->ops+oa->numops;
1305 op->cmd = cmd;
1306 op->dbid = dbid;
1307 op->argv = argv;
1308 op->argc = argc;
1309 op->target = target;
1310 oa->numops++;
1311 return oa->numops;
1312 }
1313
1314 void redisOpArrayFree(redisOpArray *oa) {
1315 while(oa->numops) {
1316 int j;
1317 redisOp *op;
1318
1319 oa->numops--;
1320 op = oa->ops+oa->numops;
1321 for (j = 0; j < op->argc; j++)
1322 decrRefCount(op->argv[j]);
1323 zfree(op->argv);
1324 }
1325 zfree(oa->ops);
1326 }
1327
1328 /* ====================== Commands lookup and execution ===================== */
1329
1330 struct redisCommand *lookupCommand(sds name) {
1331 return dictFetchValue(server.commands, name);
1332 }
1333
1334 struct redisCommand *lookupCommandByCString(char *s) {
1335 struct redisCommand *cmd;
1336 sds name = sdsnew(s);
1337
1338 cmd = dictFetchValue(server.commands, name);
1339 sdsfree(name);
1340 return cmd;
1341 }
1342
1343 /* Propagate the specified command (in the context of the specified database id)
1344 * to AOF, Slaves and Monitors.
1345 *
1346 * flags are an xor between:
1347 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1348 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1349 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1350 */
1351 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1352 int flags)
1353 {
1354 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1355 feedAppendOnlyFile(cmd,dbid,argv,argc);
1356 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1357 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1358 }
1359
1360 /* Used inside commands to schedule the propagation of additional commands
1361 * after the current command is propagated to AOF / Replication. */
1362 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1363 int target)
1364 {
1365 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1366 }
1367
1368 /* Call() is the core of Redis execution of a command */
1369 void call(redisClient *c, int flags) {
1370 long long dirty, start = ustime(), duration;
1371
1372 /* Sent the command to clients in MONITOR mode, only if the commands are
1373 * not geneated from reading an AOF. */
1374 if (listLength(server.monitors) && !server.loading)
1375 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1376
1377 /* Call the command. */
1378 redisOpArrayInit(&server.also_propagate);
1379 dirty = server.dirty;
1380 c->cmd->proc(c);
1381 dirty = server.dirty-dirty;
1382 duration = ustime()-start;
1383
1384 /* When EVAL is called loading the AOF we don't want commands called
1385 * from Lua to go into the slowlog or to populate statistics. */
1386 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1387 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1388
1389 /* Log the command into the Slow log if needed, and populate the
1390 * per-command statistics that we show in INFO commandstats. */
1391 if (flags & REDIS_CALL_SLOWLOG)
1392 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1393 if (flags & REDIS_CALL_STATS) {
1394 c->cmd->microseconds += duration;
1395 c->cmd->calls++;
1396 }
1397
1398 /* Propagate the command into the AOF and replication link */
1399 if (flags & REDIS_CALL_PROPAGATE) {
1400 int flags = REDIS_PROPAGATE_NONE;
1401
1402 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1403 flags |= REDIS_PROPAGATE_REPL;
1404 if (dirty)
1405 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1406 if (flags != REDIS_PROPAGATE_NONE)
1407 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1408 }
1409 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1410 * PUSH command. */
1411 if (server.also_propagate.numops) {
1412 int j;
1413 redisOp *rop;
1414
1415 for (j = 0; j < server.also_propagate.numops; j++) {
1416 rop = &server.also_propagate.ops[j];
1417 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1418 }
1419 redisOpArrayFree(&server.also_propagate);
1420 }
1421 server.stat_numcommands++;
1422 }
1423
1424 /* If this function gets called we already read a whole
1425 * command, argments are in the client argv/argc fields.
1426 * processCommand() execute the command or prepare the
1427 * server for a bulk read from the client.
1428 *
1429 * If 1 is returned the client is still alive and valid and
1430 * and other operations can be performed by the caller. Otherwise
1431 * if 0 is returned the client was destroied (i.e. after QUIT). */
1432 int processCommand(redisClient *c) {
1433 /* The QUIT command is handled separately. Normal command procs will
1434 * go through checking for replication and QUIT will cause trouble
1435 * when FORCE_REPLICATION is enabled and would be implemented in
1436 * a regular command proc. */
1437 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1438 addReply(c,shared.ok);
1439 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1440 return REDIS_ERR;
1441 }
1442
1443 /* Now lookup the command and check ASAP about trivial error conditions
1444 * such as wrong arity, bad command name and so forth. */
1445 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1446 if (!c->cmd) {
1447 addReplyErrorFormat(c,"unknown command '%s'",
1448 (char*)c->argv[0]->ptr);
1449 return REDIS_OK;
1450 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1451 (c->argc < -c->cmd->arity)) {
1452 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1453 c->cmd->name);
1454 return REDIS_OK;
1455 }
1456
1457 /* Check if the user is authenticated */
1458 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1459 {
1460 addReplyError(c,"operation not permitted");
1461 return REDIS_OK;
1462 }
1463
1464 /* If cluster is enabled, redirect here */
1465 if (server.cluster_enabled &&
1466 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1467 int hashslot;
1468
1469 if (server.cluster.state != REDIS_CLUSTER_OK) {
1470 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1471 return REDIS_OK;
1472 } else {
1473 int ask;
1474 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1475 if (n == NULL) {
1476 addReplyError(c,"Multi keys request invalid in cluster");
1477 return REDIS_OK;
1478 } else if (n != server.cluster.myself) {
1479 addReplySds(c,sdscatprintf(sdsempty(),
1480 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1481 hashslot,n->ip,n->port));
1482 return REDIS_OK;
1483 }
1484 }
1485 }
1486
1487 /* Handle the maxmemory directive.
1488 *
1489 * First we try to free some memory if possible (if there are volatile
1490 * keys in the dataset). If there are not the only thing we can do
1491 * is returning an error. */
1492 if (server.maxmemory) {
1493 int retval = freeMemoryIfNeeded();
1494 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1495 addReply(c, shared.oomerr);
1496 return REDIS_OK;
1497 }
1498 }
1499
1500 /* Don't accept write commands if there are problems persisting on disk. */
1501 if (server.stop_writes_on_bgsave_err &&
1502 server.saveparamslen > 0
1503 && server.lastbgsave_status == REDIS_ERR &&
1504 c->cmd->flags & REDIS_CMD_WRITE)
1505 {
1506 addReply(c, shared.bgsaveerr);
1507 return REDIS_OK;
1508 }
1509
1510 /* Don't accept wirte commands if this is a read only slave. But
1511 * accept write commands if this is our master. */
1512 if (server.masterhost && server.repl_slave_ro &&
1513 !(c->flags & REDIS_MASTER) &&
1514 c->cmd->flags & REDIS_CMD_WRITE)
1515 {
1516 addReply(c, shared.roslaveerr);
1517 return REDIS_OK;
1518 }
1519
1520 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1521 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1522 &&
1523 c->cmd->proc != subscribeCommand &&
1524 c->cmd->proc != unsubscribeCommand &&
1525 c->cmd->proc != psubscribeCommand &&
1526 c->cmd->proc != punsubscribeCommand) {
1527 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1528 return REDIS_OK;
1529 }
1530
1531 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1532 * we are a slave with a broken link with master. */
1533 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1534 server.repl_serve_stale_data == 0 &&
1535 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1536 {
1537 addReplyError(c,
1538 "link with MASTER is down and slave-serve-stale-data is set to no");
1539 return REDIS_OK;
1540 }
1541
1542 /* Loading DB? Return an error if the command is not INFO */
1543 if (server.loading && c->cmd->proc != infoCommand) {
1544 addReply(c, shared.loadingerr);
1545 return REDIS_OK;
1546 }
1547
1548 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1549 if (server.lua_timedout &&
1550 !(c->cmd->proc != shutdownCommand &&
1551 c->argc == 2 &&
1552 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1553 !(c->cmd->proc == scriptCommand &&
1554 c->argc == 2 &&
1555 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1556 {
1557 addReply(c, shared.slowscripterr);
1558 return REDIS_OK;
1559 }
1560
1561 /* Exec the command */
1562 if (c->flags & REDIS_MULTI &&
1563 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1564 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1565 {
1566 queueMultiCommand(c);
1567 addReply(c,shared.queued);
1568 } else {
1569 call(c,REDIS_CALL_FULL);
1570 }
1571 return REDIS_OK;
1572 }
1573
1574 /*================================== Shutdown =============================== */
1575
1576 int prepareForShutdown(int flags) {
1577 int save = flags & REDIS_SHUTDOWN_SAVE;
1578 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1579
1580 redisLog(REDIS_WARNING,"User requested shutdown...");
1581 /* Kill the saving child if there is a background saving in progress.
1582 We want to avoid race conditions, for instance our saving child may
1583 overwrite the synchronous saving did by SHUTDOWN. */
1584 if (server.rdb_child_pid != -1) {
1585 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1586 kill(server.rdb_child_pid,SIGKILL);
1587 rdbRemoveTempFile(server.rdb_child_pid);
1588 }
1589 if (server.aof_state != REDIS_AOF_OFF) {
1590 /* Kill the AOF saving child as the AOF we already have may be longer
1591 * but contains the full dataset anyway. */
1592 if (server.aof_child_pid != -1) {
1593 redisLog(REDIS_WARNING,
1594 "There is a child rewriting the AOF. Killing it!");
1595 kill(server.aof_child_pid,SIGKILL);
1596 }
1597 /* Append only file: fsync() the AOF and exit */
1598 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1599 aof_fsync(server.aof_fd);
1600 }
1601 if ((server.saveparamslen > 0 && !nosave) || save) {
1602 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1603 /* Snapshotting. Perform a SYNC SAVE and exit */
1604 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1605 /* Ooops.. error saving! The best we can do is to continue
1606 * operating. Note that if there was a background saving process,
1607 * in the next cron() Redis will be notified that the background
1608 * saving aborted, handling special stuff like slaves pending for
1609 * synchronization... */
1610 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1611 return REDIS_ERR;
1612 }
1613 }
1614 if (server.daemonize) {
1615 redisLog(REDIS_NOTICE,"Removing the pid file.");
1616 unlink(server.pidfile);
1617 }
1618 /* Close the listening sockets. Apparently this allows faster restarts. */
1619 if (server.ipfd != -1) close(server.ipfd);
1620 if (server.sofd != -1) close(server.sofd);
1621 if (server.unixsocket) {
1622 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1623 unlink(server.unixsocket); /* don't care if this fails */
1624 }
1625
1626 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1627 return REDIS_OK;
1628 }
1629
1630 /*================================== Commands =============================== */
1631
1632 void authCommand(redisClient *c) {
1633 if (!server.requirepass) {
1634 addReplyError(c,"Client sent AUTH, but no password is set");
1635 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1636 c->authenticated = 1;
1637 addReply(c,shared.ok);
1638 } else {
1639 c->authenticated = 0;
1640 addReplyError(c,"invalid password");
1641 }
1642 }
1643
1644 void pingCommand(redisClient *c) {
1645 addReply(c,shared.pong);
1646 }
1647
1648 void echoCommand(redisClient *c) {
1649 addReplyBulk(c,c->argv[1]);
1650 }
1651
1652 void timeCommand(redisClient *c) {
1653 struct timeval tv;
1654
1655 /* gettimeofday() can only fail if &tv is a bad addresss so we
1656 * don't check for errors. */
1657 gettimeofday(&tv,NULL);
1658 addReplyMultiBulkLen(c,2);
1659 addReplyBulkLongLong(c,tv.tv_sec);
1660 addReplyBulkLongLong(c,tv.tv_usec);
1661 }
1662
1663 /* Convert an amount of bytes into a human readable string in the form
1664 * of 100B, 2G, 100M, 4K, and so forth. */
1665 void bytesToHuman(char *s, unsigned long long n) {
1666 double d;
1667
1668 if (n < 1024) {
1669 /* Bytes */
1670 sprintf(s,"%lluB",n);
1671 return;
1672 } else if (n < (1024*1024)) {
1673 d = (double)n/(1024);
1674 sprintf(s,"%.2fK",d);
1675 } else if (n < (1024LL*1024*1024)) {
1676 d = (double)n/(1024*1024);
1677 sprintf(s,"%.2fM",d);
1678 } else if (n < (1024LL*1024*1024*1024)) {
1679 d = (double)n/(1024LL*1024*1024);
1680 sprintf(s,"%.2fG",d);
1681 }
1682 }
1683
1684 /* Create the string returned by the INFO command. This is decoupled
1685 * by the INFO command itself as we need to report the same information
1686 * on memory corruption problems. */
1687 sds genRedisInfoString(char *section) {
1688 sds info = sdsempty();
1689 time_t uptime = server.unixtime-server.stat_starttime;
1690 int j, numcommands;
1691 struct rusage self_ru, c_ru;
1692 unsigned long lol, bib;
1693 int allsections = 0, defsections = 0;
1694 int sections = 0;
1695
1696 if (section) {
1697 allsections = strcasecmp(section,"all") == 0;
1698 defsections = strcasecmp(section,"default") == 0;
1699 }
1700
1701 getrusage(RUSAGE_SELF, &self_ru);
1702 getrusage(RUSAGE_CHILDREN, &c_ru);
1703 getClientsMaxBuffers(&lol,&bib);
1704
1705 /* Server */
1706 if (allsections || defsections || !strcasecmp(section,"server")) {
1707 if (sections++) info = sdscat(info,"\r\n");
1708 info = sdscatprintf(info,
1709 "# Server\r\n"
1710 "redis_version:%s\r\n"
1711 "redis_git_sha1:%s\r\n"
1712 "redis_git_dirty:%d\r\n"
1713 "arch_bits:%d\r\n"
1714 "multiplexing_api:%s\r\n"
1715 "gcc_version:%d.%d.%d\r\n"
1716 "process_id:%ld\r\n"
1717 "run_id:%s\r\n"
1718 "tcp_port:%d\r\n"
1719 "uptime_in_seconds:%ld\r\n"
1720 "uptime_in_days:%ld\r\n"
1721 "lru_clock:%ld\r\n",
1722 REDIS_VERSION,
1723 redisGitSHA1(),
1724 strtol(redisGitDirty(),NULL,10) > 0,
1725 server.arch_bits,
1726 aeGetApiName(),
1727 #ifdef __GNUC__
1728 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1729 #else
1730 0,0,0,
1731 #endif
1732 (long) getpid(),
1733 server.runid,
1734 server.port,
1735 uptime,
1736 uptime/(3600*24),
1737 (unsigned long) server.lruclock);
1738 }
1739
1740 /* Clients */
1741 if (allsections || defsections || !strcasecmp(section,"clients")) {
1742 if (sections++) info = sdscat(info,"\r\n");
1743 info = sdscatprintf(info,
1744 "# Clients\r\n"
1745 "connected_clients:%lu\r\n"
1746 "client_longest_output_list:%lu\r\n"
1747 "client_biggest_input_buf:%lu\r\n"
1748 "blocked_clients:%d\r\n",
1749 listLength(server.clients)-listLength(server.slaves),
1750 lol, bib,
1751 server.bpop_blocked_clients);
1752 }
1753
1754 /* Memory */
1755 if (allsections || defsections || !strcasecmp(section,"memory")) {
1756 char hmem[64];
1757 char peak_hmem[64];
1758
1759 bytesToHuman(hmem,zmalloc_used_memory());
1760 bytesToHuman(peak_hmem,server.stat_peak_memory);
1761 if (sections++) info = sdscat(info,"\r\n");
1762 info = sdscatprintf(info,
1763 "# Memory\r\n"
1764 "used_memory:%zu\r\n"
1765 "used_memory_human:%s\r\n"
1766 "used_memory_rss:%zu\r\n"
1767 "used_memory_peak:%zu\r\n"
1768 "used_memory_peak_human:%s\r\n"
1769 "used_memory_lua:%lld\r\n"
1770 "mem_fragmentation_ratio:%.2f\r\n"
1771 "mem_allocator:%s\r\n",
1772 zmalloc_used_memory(),
1773 hmem,
1774 zmalloc_get_rss(),
1775 server.stat_peak_memory,
1776 peak_hmem,
1777 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1778 zmalloc_get_fragmentation_ratio(),
1779 ZMALLOC_LIB
1780 );
1781 }
1782
1783 /* Persistence */
1784 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1785 if (sections++) info = sdscat(info,"\r\n");
1786 info = sdscatprintf(info,
1787 "# Persistence\r\n"
1788 "loading:%d\r\n"
1789 "aof_enabled:%d\r\n"
1790 "changes_since_last_save:%lld\r\n"
1791 "bgsave_in_progress:%d\r\n"
1792 "last_save_time:%ld\r\n"
1793 "last_bgsave_status:%s\r\n"
1794 "bgrewriteaof_in_progress:%d\r\n",
1795 server.loading,
1796 server.aof_state != REDIS_AOF_OFF,
1797 server.dirty,
1798 server.rdb_child_pid != -1,
1799 server.lastsave,
1800 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1801 server.aof_child_pid != -1);
1802
1803 if (server.aof_state != REDIS_AOF_OFF) {
1804 info = sdscatprintf(info,
1805 "aof_current_size:%lld\r\n"
1806 "aof_base_size:%lld\r\n"
1807 "aof_pending_rewrite:%d\r\n"
1808 "aof_buffer_length:%zu\r\n"
1809 "aof_pending_bio_fsync:%llu\r\n"
1810 "aof_delayed_fsync:%lu\r\n",
1811 (long long) server.aof_current_size,
1812 (long long) server.aof_rewrite_base_size,
1813 server.aof_rewrite_scheduled,
1814 sdslen(server.aof_buf),
1815 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1816 server.aof_delayed_fsync);
1817 }
1818
1819 if (server.loading) {
1820 double perc;
1821 time_t eta, elapsed;
1822 off_t remaining_bytes = server.loading_total_bytes-
1823 server.loading_loaded_bytes;
1824
1825 perc = ((double)server.loading_loaded_bytes /
1826 server.loading_total_bytes) * 100;
1827
1828 elapsed = server.unixtime-server.loading_start_time;
1829 if (elapsed == 0) {
1830 eta = 1; /* A fake 1 second figure if we don't have
1831 enough info */
1832 } else {
1833 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1834 }
1835
1836 info = sdscatprintf(info,
1837 "loading_start_time:%ld\r\n"
1838 "loading_total_bytes:%llu\r\n"
1839 "loading_loaded_bytes:%llu\r\n"
1840 "loading_loaded_perc:%.2f\r\n"
1841 "loading_eta_seconds:%ld\r\n"
1842 ,(unsigned long) server.loading_start_time,
1843 (unsigned long long) server.loading_total_bytes,
1844 (unsigned long long) server.loading_loaded_bytes,
1845 perc,
1846 eta
1847 );
1848 }
1849 }
1850
1851 /* Stats */
1852 if (allsections || defsections || !strcasecmp(section,"stats")) {
1853 if (sections++) info = sdscat(info,"\r\n");
1854 info = sdscatprintf(info,
1855 "# Stats\r\n"
1856 "total_connections_received:%lld\r\n"
1857 "total_commands_processed:%lld\r\n"
1858 "instantaneous_ops_per_sec:%lld\r\n"
1859 "rejected_connections:%lld\r\n"
1860 "expired_keys:%lld\r\n"
1861 "evicted_keys:%lld\r\n"
1862 "keyspace_hits:%lld\r\n"
1863 "keyspace_misses:%lld\r\n"
1864 "pubsub_channels:%ld\r\n"
1865 "pubsub_patterns:%lu\r\n"
1866 "latest_fork_usec:%lld\r\n",
1867 server.stat_numconnections,
1868 server.stat_numcommands,
1869 getOperationsPerSecond(),
1870 server.stat_rejected_conn,
1871 server.stat_expiredkeys,
1872 server.stat_evictedkeys,
1873 server.stat_keyspace_hits,
1874 server.stat_keyspace_misses,
1875 dictSize(server.pubsub_channels),
1876 listLength(server.pubsub_patterns),
1877 server.stat_fork_time);
1878 }
1879
1880 /* Replication */
1881 if (allsections || defsections || !strcasecmp(section,"replication")) {
1882 if (sections++) info = sdscat(info,"\r\n");
1883 info = sdscatprintf(info,
1884 "# Replication\r\n"
1885 "role:%s\r\n",
1886 server.masterhost == NULL ? "master" : "slave");
1887 if (server.masterhost) {
1888 info = sdscatprintf(info,
1889 "master_host:%s\r\n"
1890 "master_port:%d\r\n"
1891 "master_link_status:%s\r\n"
1892 "master_last_io_seconds_ago:%d\r\n"
1893 "master_sync_in_progress:%d\r\n"
1894 ,server.masterhost,
1895 server.masterport,
1896 (server.repl_state == REDIS_REPL_CONNECTED) ?
1897 "up" : "down",
1898 server.master ?
1899 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
1900 server.repl_state == REDIS_REPL_TRANSFER
1901 );
1902
1903 if (server.repl_state == REDIS_REPL_TRANSFER) {
1904 info = sdscatprintf(info,
1905 "master_sync_left_bytes:%ld\r\n"
1906 "master_sync_last_io_seconds_ago:%d\r\n"
1907 ,(long)server.repl_transfer_left,
1908 (int)(server.unixtime-server.repl_transfer_lastio)
1909 );
1910 }
1911
1912 if (server.repl_state != REDIS_REPL_CONNECTED) {
1913 info = sdscatprintf(info,
1914 "master_link_down_since_seconds:%ld\r\n",
1915 (long)server.unixtime-server.repl_down_since);
1916 }
1917 }
1918 info = sdscatprintf(info,
1919 "connected_slaves:%lu\r\n",
1920 listLength(server.slaves));
1921 if (listLength(server.slaves)) {
1922 int slaveid = 0;
1923 listNode *ln;
1924 listIter li;
1925
1926 listRewind(server.slaves,&li);
1927 while((ln = listNext(&li))) {
1928 redisClient *slave = listNodeValue(ln);
1929 char *state = NULL;
1930 char ip[32];
1931 int port;
1932
1933 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
1934 switch(slave->replstate) {
1935 case REDIS_REPL_WAIT_BGSAVE_START:
1936 case REDIS_REPL_WAIT_BGSAVE_END:
1937 state = "wait_bgsave";
1938 break;
1939 case REDIS_REPL_SEND_BULK:
1940 state = "send_bulk";
1941 break;
1942 case REDIS_REPL_ONLINE:
1943 state = "online";
1944 break;
1945 }
1946 if (state == NULL) continue;
1947 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
1948 slaveid,ip,port,state);
1949 slaveid++;
1950 }
1951 }
1952 }
1953
1954 /* CPU */
1955 if (allsections || defsections || !strcasecmp(section,"cpu")) {
1956 if (sections++) info = sdscat(info,"\r\n");
1957 info = sdscatprintf(info,
1958 "# CPU\r\n"
1959 "used_cpu_sys:%.2f\r\n"
1960 "used_cpu_user:%.2f\r\n"
1961 "used_cpu_sys_children:%.2f\r\n"
1962 "used_cpu_user_children:%.2f\r\n",
1963 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
1964 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
1965 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
1966 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
1967 }
1968
1969 /* cmdtime */
1970 if (allsections || !strcasecmp(section,"commandstats")) {
1971 if (sections++) info = sdscat(info,"\r\n");
1972 info = sdscatprintf(info, "# Commandstats\r\n");
1973 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1974 for (j = 0; j < numcommands; j++) {
1975 struct redisCommand *c = redisCommandTable+j;
1976
1977 if (!c->calls) continue;
1978 info = sdscatprintf(info,
1979 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
1980 c->name, c->calls, c->microseconds,
1981 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
1982 }
1983 }
1984
1985 /* Clusetr */
1986 if (allsections || defsections || !strcasecmp(section,"cluster")) {
1987 if (sections++) info = sdscat(info,"\r\n");
1988 info = sdscatprintf(info,
1989 "# Cluster\r\n"
1990 "cluster_enabled:%d\r\n",
1991 server.cluster_enabled);
1992 }
1993
1994 /* Key space */
1995 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
1996 if (sections++) info = sdscat(info,"\r\n");
1997 info = sdscatprintf(info, "# Keyspace\r\n");
1998 for (j = 0; j < server.dbnum; j++) {
1999 long long keys, vkeys;
2000
2001 keys = dictSize(server.db[j].dict);
2002 vkeys = dictSize(server.db[j].expires);
2003 if (keys || vkeys) {
2004 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2005 j, keys, vkeys);
2006 }
2007 }
2008 }
2009 return info;
2010 }
2011
2012 void infoCommand(redisClient *c) {
2013 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2014
2015 if (c->argc > 2) {
2016 addReply(c,shared.syntaxerr);
2017 return;
2018 }
2019 sds info = genRedisInfoString(section);
2020 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2021 (unsigned long)sdslen(info)));
2022 addReplySds(c,info);
2023 addReply(c,shared.crlf);
2024 }
2025
2026 void monitorCommand(redisClient *c) {
2027 /* ignore MONITOR if aleady slave or in monitor mode */
2028 if (c->flags & REDIS_SLAVE) return;
2029
2030 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2031 c->slaveseldb = 0;
2032 listAddNodeTail(server.monitors,c);
2033 addReply(c,shared.ok);
2034 }
2035
2036 /* ============================ Maxmemory directive ======================== */
2037
2038 /* This function gets called when 'maxmemory' is set on the config file to limit
2039 * the max memory used by the server, before processing a command.
2040 *
2041 * The goal of the function is to free enough memory to keep Redis under the
2042 * configured memory limit.
2043 *
2044 * The function starts calculating how many bytes should be freed to keep
2045 * Redis under the limit, and enters a loop selecting the best keys to
2046 * evict accordingly to the configured policy.
2047 *
2048 * If all the bytes needed to return back under the limit were freed the
2049 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2050 * should block the execution of commands that will result in more memory
2051 * used by the server.
2052 */
2053 int freeMemoryIfNeeded(void) {
2054 size_t mem_used, mem_tofree, mem_freed;
2055 int slaves = listLength(server.slaves);
2056
2057 /* Remove the size of slaves output buffers and AOF buffer from the
2058 * count of used memory. */
2059 mem_used = zmalloc_used_memory();
2060 if (slaves) {
2061 listIter li;
2062 listNode *ln;
2063
2064 listRewind(server.slaves,&li);
2065 while((ln = listNext(&li))) {
2066 redisClient *slave = listNodeValue(ln);
2067 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2068 if (obuf_bytes > mem_used)
2069 mem_used = 0;
2070 else
2071 mem_used -= obuf_bytes;
2072 }
2073 }
2074 if (server.aof_state != REDIS_AOF_OFF) {
2075 mem_used -= sdslen(server.aof_buf);
2076 mem_used -= sdslen(server.aof_rewrite_buf);
2077 }
2078
2079 /* Check if we are over the memory limit. */
2080 if (mem_used <= server.maxmemory) return REDIS_OK;
2081
2082 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2083 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2084
2085 /* Compute how much memory we need to free. */
2086 mem_tofree = mem_used - server.maxmemory;
2087 mem_freed = 0;
2088 while (mem_freed < mem_tofree) {
2089 int j, k, keys_freed = 0;
2090
2091 for (j = 0; j < server.dbnum; j++) {
2092 long bestval = 0; /* just to prevent warning */
2093 sds bestkey = NULL;
2094 struct dictEntry *de;
2095 redisDb *db = server.db+j;
2096 dict *dict;
2097
2098 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2099 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2100 {
2101 dict = server.db[j].dict;
2102 } else {
2103 dict = server.db[j].expires;
2104 }
2105 if (dictSize(dict) == 0) continue;
2106
2107 /* volatile-random and allkeys-random policy */
2108 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2109 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2110 {
2111 de = dictGetRandomKey(dict);
2112 bestkey = dictGetKey(de);
2113 }
2114
2115 /* volatile-lru and allkeys-lru policy */
2116 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2117 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2118 {
2119 for (k = 0; k < server.maxmemory_samples; k++) {
2120 sds thiskey;
2121 long thisval;
2122 robj *o;
2123
2124 de = dictGetRandomKey(dict);
2125 thiskey = dictGetKey(de);
2126 /* When policy is volatile-lru we need an additonal lookup
2127 * to locate the real key, as dict is set to db->expires. */
2128 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2129 de = dictFind(db->dict, thiskey);
2130 o = dictGetVal(de);
2131 thisval = estimateObjectIdleTime(o);
2132
2133 /* Higher idle time is better candidate for deletion */
2134 if (bestkey == NULL || thisval > bestval) {
2135 bestkey = thiskey;
2136 bestval = thisval;
2137 }
2138 }
2139 }
2140
2141 /* volatile-ttl */
2142 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2143 for (k = 0; k < server.maxmemory_samples; k++) {
2144 sds thiskey;
2145 long thisval;
2146
2147 de = dictGetRandomKey(dict);
2148 thiskey = dictGetKey(de);
2149 thisval = (long) dictGetVal(de);
2150
2151 /* Expire sooner (minor expire unix timestamp) is better
2152 * candidate for deletion */
2153 if (bestkey == NULL || thisval < bestval) {
2154 bestkey = thiskey;
2155 bestval = thisval;
2156 }
2157 }
2158 }
2159
2160 /* Finally remove the selected key. */
2161 if (bestkey) {
2162 long long delta;
2163
2164 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2165 propagateExpire(db,keyobj);
2166 /* We compute the amount of memory freed by dbDelete() alone.
2167 * It is possible that actually the memory needed to propagate
2168 * the DEL in AOF and replication link is greater than the one
2169 * we are freeing removing the key, but we can't account for
2170 * that otherwise we would never exit the loop.
2171 *
2172 * AOF and Output buffer memory will be freed eventually so
2173 * we only care about memory used by the key space. */
2174 delta = (long long) zmalloc_used_memory();
2175 dbDelete(db,keyobj);
2176 delta -= (long long) zmalloc_used_memory();
2177 mem_freed += delta;
2178 server.stat_evictedkeys++;
2179 decrRefCount(keyobj);
2180 keys_freed++;
2181
2182 /* When the memory to free starts to be big enough, we may
2183 * start spending so much time here that is impossible to
2184 * deliver data to the slaves fast enough, so we force the
2185 * transmission here inside the loop. */
2186 if (slaves) flushSlavesOutputBuffers();
2187 }
2188 }
2189 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2190 }
2191 return REDIS_OK;
2192 }
2193
2194 /* =================================== Main! ================================ */
2195
2196 #ifdef __linux__
2197 int linuxOvercommitMemoryValue(void) {
2198 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2199 char buf[64];
2200
2201 if (!fp) return -1;
2202 if (fgets(buf,64,fp) == NULL) {
2203 fclose(fp);
2204 return -1;
2205 }
2206 fclose(fp);
2207
2208 return atoi(buf);
2209 }
2210
2211 void linuxOvercommitMemoryWarning(void) {
2212 if (linuxOvercommitMemoryValue() == 0) {
2213 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.");
2214 }
2215 }
2216 #endif /* __linux__ */
2217
2218 void createPidFile(void) {
2219 /* Try to write the pid file in a best-effort way. */
2220 FILE *fp = fopen(server.pidfile,"w");
2221 if (fp) {
2222 fprintf(fp,"%d\n",(int)getpid());
2223 fclose(fp);
2224 }
2225 }
2226
2227 void daemonize(void) {
2228 int fd;
2229
2230 if (fork() != 0) exit(0); /* parent exits */
2231 setsid(); /* create a new session */
2232
2233 /* Every output goes to /dev/null. If Redis is daemonized but
2234 * the 'logfile' is set to 'stdout' in the configuration file
2235 * it will not log at all. */
2236 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2237 dup2(fd, STDIN_FILENO);
2238 dup2(fd, STDOUT_FILENO);
2239 dup2(fd, STDERR_FILENO);
2240 if (fd > STDERR_FILENO) close(fd);
2241 }
2242 }
2243
2244 void version() {
2245 printf("Redis server v=%s sha=%s:%d malloc=%s\n", REDIS_VERSION,
2246 redisGitSHA1(), atoi(redisGitDirty()) > 0, ZMALLOC_LIB);
2247 exit(0);
2248 }
2249
2250 void usage() {
2251 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2252 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2253 fprintf(stderr," ./redis-server -v or --version\n");
2254 fprintf(stderr," ./redis-server -h or --help\n");
2255 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2256 fprintf(stderr,"Examples:\n");
2257 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2258 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2259 fprintf(stderr," ./redis-server --port 7777\n");
2260 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2261 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2262 exit(1);
2263 }
2264
2265 void redisAsciiArt(void) {
2266 #include "asciilogo.h"
2267 char *buf = zmalloc(1024*16);
2268
2269 snprintf(buf,1024*16,ascii_logo,
2270 REDIS_VERSION,
2271 redisGitSHA1(),
2272 strtol(redisGitDirty(),NULL,10) > 0,
2273 (sizeof(long) == 8) ? "64" : "32",
2274 server.cluster_enabled ? "cluster" : "stand alone",
2275 server.port,
2276 (long) getpid()
2277 );
2278 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2279 zfree(buf);
2280 }
2281
2282 static void sigtermHandler(int sig) {
2283 REDIS_NOTUSED(sig);
2284
2285 redisLog(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2286 server.shutdown_asap = 1;
2287 }
2288
2289 void setupSignalHandlers(void) {
2290 struct sigaction act;
2291
2292 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2293 * Otherwise, sa_handler is used. */
2294 sigemptyset(&act.sa_mask);
2295 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
2296 act.sa_handler = sigtermHandler;
2297 sigaction(SIGTERM, &act, NULL);
2298
2299 #ifdef HAVE_BACKTRACE
2300 sigemptyset(&act.sa_mask);
2301 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
2302 act.sa_sigaction = sigsegvHandler;
2303 sigaction(SIGSEGV, &act, NULL);
2304 sigaction(SIGBUS, &act, NULL);
2305 sigaction(SIGFPE, &act, NULL);
2306 sigaction(SIGILL, &act, NULL);
2307 #endif
2308 return;
2309 }
2310
2311 void memtest(size_t megabytes, int passes);
2312
2313 int main(int argc, char **argv) {
2314 long long start;
2315 struct timeval tv;
2316
2317 /* We need to initialize our libraries, and the server configuration. */
2318 zmalloc_enable_thread_safeness();
2319 srand(time(NULL)^getpid());
2320 gettimeofday(&tv,NULL);
2321 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2322 initServerConfig();
2323
2324 if (argc >= 2) {
2325 int j = 1; /* First option to parse in argv[] */
2326 sds options = sdsempty();
2327 char *configfile = NULL;
2328
2329 /* Handle special options --help and --version */
2330 if (strcmp(argv[1], "-v") == 0 ||
2331 strcmp(argv[1], "--version") == 0) version();
2332 if (strcmp(argv[1], "--help") == 0 ||
2333 strcmp(argv[1], "-h") == 0) usage();
2334 if (strcmp(argv[1], "--test-memory") == 0) {
2335 if (argc == 3) {
2336 memtest(atoi(argv[2]),50);
2337 exit(0);
2338 } else {
2339 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2340 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2341 exit(1);
2342 }
2343 }
2344
2345 /* First argument is the config file name? */
2346 if (argv[j][0] != '-' || argv[j][1] != '-')
2347 configfile = argv[j++];
2348 /* All the other options are parsed and conceptually appended to the
2349 * configuration file. For instance --port 6380 will generate the
2350 * string "port 6380\n" to be parsed after the actual file name
2351 * is parsed, if any. */
2352 while(j != argc) {
2353 if (argv[j][0] == '-' && argv[j][1] == '-') {
2354 /* Option name */
2355 if (sdslen(options)) options = sdscat(options,"\n");
2356 options = sdscat(options,argv[j]+2);
2357 options = sdscat(options," ");
2358 } else {
2359 /* Option argument */
2360 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2361 options = sdscat(options," ");
2362 }
2363 j++;
2364 }
2365 resetServerSaveParams();
2366 loadServerConfig(configfile,options);
2367 sdsfree(options);
2368 } else {
2369 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'");
2370 }
2371 if (server.daemonize) daemonize();
2372 initServer();
2373 if (server.daemonize) createPidFile();
2374 redisAsciiArt();
2375 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2376 #ifdef __linux__
2377 linuxOvercommitMemoryWarning();
2378 #endif
2379 start = ustime();
2380 if (server.aof_state == REDIS_AOF_ON) {
2381 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2382 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2383 } else {
2384 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2385 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2386 (float)(ustime()-start)/1000000);
2387 } else if (errno != ENOENT) {
2388 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2389 exit(1);
2390 }
2391 }
2392 if (server.ipfd > 0)
2393 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2394 if (server.sofd > 0)
2395 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2396 aeSetBeforeSleepProc(server.el,beforeSleep);
2397 aeMain(server.el);
2398 aeDeleteEventLoop(server.el);
2399 return 0;
2400 }
2401
2402 /* The End */