]> git.saurik.com Git - redis.git/blob - src/redis.c
fd53e248eb2ab7312a467621bea89a4edc89f1cc
[redis.git] / src / redis.c
1 /*
2 * Copyright (c) 2009-2010, Salvatore Sanfilippo <antirez at gmail dot com>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are met:
7 *
8 * * Redistributions of source code must retain the above copyright notice,
9 * this list of conditions and the following disclaimer.
10 * * Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * * Neither the name of Redis nor the names of its contributors may be used
14 * to endorse or promote products derived from this software without
15 * specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27 * POSSIBILITY OF SUCH DAMAGE.
28 */
29
30 #include "redis.h"
31 #include "slowlog.h"
32 #include "bio.h"
33
34 #include <time.h>
35 #include <signal.h>
36 #include <sys/wait.h>
37 #include <errno.h>
38 #include <assert.h>
39 #include <ctype.h>
40 #include <stdarg.h>
41 #include <arpa/inet.h>
42 #include <sys/stat.h>
43 #include <fcntl.h>
44 #include <sys/time.h>
45 #include <sys/resource.h>
46 #include <sys/uio.h>
47 #include <limits.h>
48 #include <float.h>
49 #include <math.h>
50 #include <sys/resource.h>
51 #include <sys/utsname.h>
52
53 /* Our shared "common" objects */
54
55 struct sharedObjectsStruct shared;
56
57 /* Global vars that are actually used as constants. The following double
58 * values are used for double on-disk serialization, and are initialized
59 * at runtime to avoid strange compiler optimizations. */
60
61 double R_Zero, R_PosInf, R_NegInf, R_Nan;
62
63 /*================================= Globals ================================= */
64
65 /* Global vars */
66 struct redisServer server; /* server global state */
67 struct redisCommand *commandTable;
68
69 /* Our command table.
70 *
71 * Every entry is composed of the following fields:
72 *
73 * name: a string representing the command name.
74 * function: pointer to the C function implementing the command.
75 * arity: number of arguments, it is possible to use -N to say >= N
76 * sflags: command flags as string. See below for a table of flags.
77 * flags: flags as bitmask. Computed by Redis using the 'sflags' field.
78 * get_keys_proc: an optional function to get key arguments from a command.
79 * This is only used when the following three fields are not
80 * enough to specify what arguments are keys.
81 * first_key_index: first argument that is a key
82 * last_key_index: last argument that is a key
83 * key_step: step to get all the keys from first to last argument. For instance
84 * in MSET the step is two since arguments are key,val,key,val,...
85 * microseconds: microseconds of total execution time for this command.
86 * calls: total number of calls of this command.
87 *
88 * The flags, microseconds and calls fields are computed by Redis and should
89 * always be set to zero.
90 *
91 * Command flags are expressed using strings where every character represents
92 * a flag. Later the populateCommandTable() function will take care of
93 * populating the real 'flags' field using this characters.
94 *
95 * This is the meaning of the flags:
96 *
97 * w: write command (may modify the key space).
98 * r: read command (will never modify the key space).
99 * m: may increase memory usage once called. Don't allow if out of memory.
100 * a: admin command, like SAVE or SHUTDOWN.
101 * p: Pub/Sub related command.
102 * f: force replication of this command, regarless of server.dirty.
103 * s: command not allowed in scripts.
104 * R: random command. Command is not deterministic, that is, the same command
105 * with the same arguments, with the same key space, may have different
106 * results. For instance SPOP and RANDOMKEY are two random commands.
107 * S: Sort command output array if called from script, so that the output
108 * is deterministic.
109 */
110 struct redisCommand redisCommandTable[] = {
111 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
112 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
113 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
114 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
115 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
116 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
117 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
118 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
119 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
120 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
121 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
122 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
123 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
124 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
125 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
126 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
127 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
128 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
129 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
130 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
131 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
132 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
133 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
134 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
135 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
136 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
137 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
138 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
139 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
140 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
141 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
142 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
143 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
144 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
145 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
146 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
147 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
148 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
149 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
150 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
151 {"srandmember",srandmemberCommand,2,"rR",0,NULL,1,1,1,0,0},
152 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
153 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
154 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
155 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
156 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
157 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
158 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
159 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
160 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
161 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
162 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
163 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
164 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
165 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
166 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
167 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
168 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
169 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
170 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
171 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
172 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
173 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
174 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
175 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
176 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
177 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
178 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
179 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
180 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
182 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
183 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
184 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
185 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
186 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
187 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
188 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
189 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
190 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
191 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
192 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
193 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
194 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
195 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
196 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
197 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
198 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
199 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
200 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
201 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
202 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
203 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
204 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
205 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
206 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
207 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
208 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
209 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
210 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
211 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
212 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
213 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
214 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
215 {"exec",execCommand,1,"s",0,NULL,0,0,0,0,0},
216 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
217 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
218 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
219 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
220 {"sort",sortCommand,-2,"wmS",0,NULL,1,1,1,0,0},
221 {"info",infoCommand,-1,"r",0,NULL,0,0,0,0,0},
222 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
223 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
224 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
225 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
226 {"slaveof",slaveofCommand,3,"as",0,NULL,0,0,0,0,0},
227 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
228 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
229 {"subscribe",subscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
230 {"unsubscribe",unsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
231 {"psubscribe",psubscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
232 {"punsubscribe",punsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
233 {"publish",publishCommand,3,"pf",0,NULL,0,0,0,0,0},
234 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
235 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
236 {"cluster",clusterCommand,-2,"ar",0,NULL,0,0,0,0,0},
237 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
238 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
239 {"asking",askingCommand,1,"r",0,NULL,0,0,0,0,0},
240 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
241 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
242 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
243 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
244 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
245 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
246 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
247 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0}
248 };
249
250 /*============================ Utility functions ============================ */
251
252 /* Low level logging. To use only for very big messages, otherwise
253 * redisLog() is to prefer. */
254 void redisLogRaw(int level, const char *msg) {
255 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
256 const char *c = ".-*#";
257 FILE *fp;
258 char buf[64];
259 int rawmode = (level & REDIS_LOG_RAW);
260
261 level &= 0xff; /* clear flags */
262 if (level < server.verbosity) return;
263
264 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
265 if (!fp) return;
266
267 if (rawmode) {
268 fprintf(fp,"%s",msg);
269 } else {
270 int off;
271 struct timeval tv;
272
273 gettimeofday(&tv,NULL);
274 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
275 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
276 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
277 }
278 fflush(fp);
279
280 if (server.logfile) fclose(fp);
281
282 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
283 }
284
285 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
286 * is used across the code. The raw version is only used in order to dump
287 * the INFO output on crash. */
288 void redisLog(int level, const char *fmt, ...) {
289 va_list ap;
290 char msg[REDIS_MAX_LOGMSG_LEN];
291
292 if ((level&0xff) < server.verbosity) return;
293
294 va_start(ap, fmt);
295 vsnprintf(msg, sizeof(msg), fmt, ap);
296 va_end(ap);
297
298 redisLogRaw(level,msg);
299 }
300
301 /* Log a fixed message without printf-alike capabilities, in a way that is
302 * safe to call from a signal handler.
303 *
304 * We actually use this only for signals that are not fatal from the point
305 * of view of Redis. Signals that are going to kill the server anyway and
306 * where we need printf-alike features are served by redisLog(). */
307 void redisLogFromHandler(int level, const char *msg) {
308 int fd;
309 char buf[64];
310
311 if ((level&0xff) < server.verbosity ||
312 (server.logfile == NULL && server.daemonize)) return;
313 fd = server.logfile ?
314 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
315 STDOUT_FILENO;
316 if (fd == -1) return;
317 ll2string(buf,sizeof(buf),getpid());
318 if (write(fd,"[",1) == -1) goto err;
319 if (write(fd,buf,strlen(buf)) == -1) goto err;
320 if (write(fd," | signal handler] (",20) == -1) goto err;
321 ll2string(buf,sizeof(buf),time(NULL));
322 if (write(fd,buf,strlen(buf)) == -1) goto err;
323 if (write(fd,") ",2) == -1) goto err;
324 if (write(fd,msg,strlen(msg)) == -1) goto err;
325 if (write(fd,"\n",1) == -1) goto err;
326 err:
327 if (server.logfile) close(fd);
328 }
329
330 /* Redis generally does not try to recover from out of memory conditions
331 * when allocating objects or strings, it is not clear if it will be possible
332 * to report this condition to the client since the networking layer itself
333 * is based on heap allocation for send buffers, so we simply abort.
334 * At least the code will be simpler to read... */
335 void oom(const char *msg) {
336 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
337 sleep(1);
338 abort();
339 }
340
341 /* Return the UNIX time in microseconds */
342 long long ustime(void) {
343 struct timeval tv;
344 long long ust;
345
346 gettimeofday(&tv, NULL);
347 ust = ((long long)tv.tv_sec)*1000000;
348 ust += tv.tv_usec;
349 return ust;
350 }
351
352 /* Return the UNIX time in milliseconds */
353 long long mstime(void) {
354 return ustime()/1000;
355 }
356
357 /* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
358 * exit(), because the latter may interact with the same file objects used by
359 * the parent process. However if we are testing the coverage normal exit() is
360 * used in order to obtain the right coverage information. */
361 void exitFromChild(int retcode) {
362 #ifdef COVERAGE_TEST
363 exit(retcode);
364 #else
365 _exit(retcode);
366 #endif
367 }
368
369 /*====================== Hash table type implementation ==================== */
370
371 /* This is an hash table type that uses the SDS dynamic strings libary as
372 * keys and radis objects as values (objects can hold SDS strings,
373 * lists, sets). */
374
375 void dictVanillaFree(void *privdata, void *val)
376 {
377 DICT_NOTUSED(privdata);
378 zfree(val);
379 }
380
381 void dictListDestructor(void *privdata, void *val)
382 {
383 DICT_NOTUSED(privdata);
384 listRelease((list*)val);
385 }
386
387 int dictSdsKeyCompare(void *privdata, const void *key1,
388 const void *key2)
389 {
390 int l1,l2;
391 DICT_NOTUSED(privdata);
392
393 l1 = sdslen((sds)key1);
394 l2 = sdslen((sds)key2);
395 if (l1 != l2) return 0;
396 return memcmp(key1, key2, l1) == 0;
397 }
398
399 /* A case insensitive version used for the command lookup table. */
400 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
401 const void *key2)
402 {
403 DICT_NOTUSED(privdata);
404
405 return strcasecmp(key1, key2) == 0;
406 }
407
408 void dictRedisObjectDestructor(void *privdata, void *val)
409 {
410 DICT_NOTUSED(privdata);
411
412 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
413 decrRefCount(val);
414 }
415
416 void dictSdsDestructor(void *privdata, void *val)
417 {
418 DICT_NOTUSED(privdata);
419
420 sdsfree(val);
421 }
422
423 int dictObjKeyCompare(void *privdata, const void *key1,
424 const void *key2)
425 {
426 const robj *o1 = key1, *o2 = key2;
427 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
428 }
429
430 unsigned int dictObjHash(const void *key) {
431 const robj *o = key;
432 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
433 }
434
435 unsigned int dictSdsHash(const void *key) {
436 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
437 }
438
439 unsigned int dictSdsCaseHash(const void *key) {
440 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
441 }
442
443 int dictEncObjKeyCompare(void *privdata, const void *key1,
444 const void *key2)
445 {
446 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
447 int cmp;
448
449 if (o1->encoding == REDIS_ENCODING_INT &&
450 o2->encoding == REDIS_ENCODING_INT)
451 return o1->ptr == o2->ptr;
452
453 o1 = getDecodedObject(o1);
454 o2 = getDecodedObject(o2);
455 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
456 decrRefCount(o1);
457 decrRefCount(o2);
458 return cmp;
459 }
460
461 unsigned int dictEncObjHash(const void *key) {
462 robj *o = (robj*) key;
463
464 if (o->encoding == REDIS_ENCODING_RAW) {
465 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
466 } else {
467 if (o->encoding == REDIS_ENCODING_INT) {
468 char buf[32];
469 int len;
470
471 len = ll2string(buf,32,(long)o->ptr);
472 return dictGenHashFunction((unsigned char*)buf, len);
473 } else {
474 unsigned int hash;
475
476 o = getDecodedObject(o);
477 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
478 decrRefCount(o);
479 return hash;
480 }
481 }
482 }
483
484 /* Sets type hash table */
485 dictType setDictType = {
486 dictEncObjHash, /* hash function */
487 NULL, /* key dup */
488 NULL, /* val dup */
489 dictEncObjKeyCompare, /* key compare */
490 dictRedisObjectDestructor, /* key destructor */
491 NULL /* val destructor */
492 };
493
494 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
495 dictType zsetDictType = {
496 dictEncObjHash, /* hash function */
497 NULL, /* key dup */
498 NULL, /* val dup */
499 dictEncObjKeyCompare, /* key compare */
500 dictRedisObjectDestructor, /* key destructor */
501 NULL /* val destructor */
502 };
503
504 /* Db->dict, keys are sds strings, vals are Redis objects. */
505 dictType dbDictType = {
506 dictSdsHash, /* hash function */
507 NULL, /* key dup */
508 NULL, /* val dup */
509 dictSdsKeyCompare, /* key compare */
510 dictSdsDestructor, /* key destructor */
511 dictRedisObjectDestructor /* val destructor */
512 };
513
514 /* Db->expires */
515 dictType keyptrDictType = {
516 dictSdsHash, /* hash function */
517 NULL, /* key dup */
518 NULL, /* val dup */
519 dictSdsKeyCompare, /* key compare */
520 NULL, /* key destructor */
521 NULL /* val destructor */
522 };
523
524 /* Command table. sds string -> command struct pointer. */
525 dictType commandTableDictType = {
526 dictSdsCaseHash, /* hash function */
527 NULL, /* key dup */
528 NULL, /* val dup */
529 dictSdsKeyCaseCompare, /* key compare */
530 dictSdsDestructor, /* key destructor */
531 NULL /* val destructor */
532 };
533
534 /* Hash type hash table (note that small hashes are represented with zimpaps) */
535 dictType hashDictType = {
536 dictEncObjHash, /* hash function */
537 NULL, /* key dup */
538 NULL, /* val dup */
539 dictEncObjKeyCompare, /* key compare */
540 dictRedisObjectDestructor, /* key destructor */
541 dictRedisObjectDestructor /* val destructor */
542 };
543
544 /* Keylist hash table type has unencoded redis objects as keys and
545 * lists as values. It's used for blocking operations (BLPOP) and to
546 * map swapped keys to a list of clients waiting for this keys to be loaded. */
547 dictType keylistDictType = {
548 dictObjHash, /* hash function */
549 NULL, /* key dup */
550 NULL, /* val dup */
551 dictObjKeyCompare, /* key compare */
552 dictRedisObjectDestructor, /* key destructor */
553 dictListDestructor /* val destructor */
554 };
555
556 /* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
557 * clusterNode structures. */
558 dictType clusterNodesDictType = {
559 dictSdsHash, /* hash function */
560 NULL, /* key dup */
561 NULL, /* val dup */
562 dictSdsKeyCompare, /* key compare */
563 dictSdsDestructor, /* key destructor */
564 NULL /* val destructor */
565 };
566
567 int htNeedsResize(dict *dict) {
568 long long size, used;
569
570 size = dictSlots(dict);
571 used = dictSize(dict);
572 return (size && used && size > DICT_HT_INITIAL_SIZE &&
573 (used*100/size < REDIS_HT_MINFILL));
574 }
575
576 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
577 * we resize the hash table to save memory */
578 void tryResizeHashTables(void) {
579 int j;
580
581 for (j = 0; j < server.dbnum; j++) {
582 if (htNeedsResize(server.db[j].dict))
583 dictResize(server.db[j].dict);
584 if (htNeedsResize(server.db[j].expires))
585 dictResize(server.db[j].expires);
586 }
587 }
588
589 /* Our hash table implementation performs rehashing incrementally while
590 * we write/read from the hash table. Still if the server is idle, the hash
591 * table will use two tables for a long time. So we try to use 1 millisecond
592 * of CPU time at every serverCron() loop in order to rehash some key. */
593 void incrementallyRehash(void) {
594 int j;
595
596 for (j = 0; j < server.dbnum; j++) {
597 if (dictIsRehashing(server.db[j].dict)) {
598 dictRehashMilliseconds(server.db[j].dict,1);
599 break; /* already used our millisecond for this loop... */
600 }
601 }
602 }
603
604 /* This function is called once a background process of some kind terminates,
605 * as we want to avoid resizing the hash tables when there is a child in order
606 * to play well with copy-on-write (otherwise when a resize happens lots of
607 * memory pages are copied). The goal of this function is to update the ability
608 * for dict.c to resize the hash tables accordingly to the fact we have o not
609 * running childs. */
610 void updateDictResizePolicy(void) {
611 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
612 dictEnableResize();
613 else
614 dictDisableResize();
615 }
616
617 /* ======================= Cron: called every 100 ms ======================== */
618
619 /* Try to expire a few timed out keys. The algorithm used is adaptive and
620 * will use few CPU cycles if there are few expiring keys, otherwise
621 * it will get more aggressive to avoid that too much memory is used by
622 * keys that can be removed from the keyspace. */
623 void activeExpireCycle(void) {
624 int j;
625
626 for (j = 0; j < server.dbnum; j++) {
627 int expired;
628 redisDb *db = server.db+j;
629
630 /* Continue to expire if at the end of the cycle more than 25%
631 * of the keys were expired. */
632 do {
633 long num = dictSize(db->expires);
634 long long now = mstime();
635
636 expired = 0;
637 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
638 num = REDIS_EXPIRELOOKUPS_PER_CRON;
639 while (num--) {
640 dictEntry *de;
641 long long t;
642
643 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
644 t = dictGetSignedIntegerVal(de);
645 if (now > t) {
646 sds key = dictGetKey(de);
647 robj *keyobj = createStringObject(key,sdslen(key));
648
649 propagateExpire(db,keyobj);
650 dbDelete(db,keyobj);
651 decrRefCount(keyobj);
652 expired++;
653 server.stat_expiredkeys++;
654 }
655 }
656 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
657 }
658 }
659
660 void updateLRUClock(void) {
661 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
662 REDIS_LRU_CLOCK_MAX;
663 }
664
665
666 /* Add a sample to the operations per second array of samples. */
667 void trackOperationsPerSecond(void) {
668 long long t = mstime() - server.ops_sec_last_sample_time;
669 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
670 long long ops_sec;
671
672 ops_sec = t > 0 ? (ops*1000/t) : 0;
673
674 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
675 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
676 server.ops_sec_last_sample_time = mstime();
677 server.ops_sec_last_sample_ops = server.stat_numcommands;
678 }
679
680 /* Return the mean of all the samples. */
681 long long getOperationsPerSecond(void) {
682 int j;
683 long long sum = 0;
684
685 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
686 sum += server.ops_sec_samples[j];
687 return sum / REDIS_OPS_SEC_SAMPLES;
688 }
689
690 /* Check for timeouts. Returns non-zero if the client was terminated */
691 int clientsCronHandleTimeout(redisClient *c) {
692 time_t now = server.unixtime;
693
694 if (server.maxidletime &&
695 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
696 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
697 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
698 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
699 listLength(c->pubsub_patterns) == 0 &&
700 (now - c->lastinteraction > server.maxidletime))
701 {
702 redisLog(REDIS_VERBOSE,"Closing idle client");
703 freeClient(c);
704 return 1;
705 } else if (c->flags & REDIS_BLOCKED) {
706 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
707 addReply(c,shared.nullmultibulk);
708 unblockClientWaitingData(c);
709 }
710 }
711 return 0;
712 }
713
714 /* The client query buffer is an sds.c string that can end with a lot of
715 * free space not used, this function reclaims space if needed.
716 *
717 * The funciton always returns 0 as it never terminates the client. */
718 int clientsCronResizeQueryBuffer(redisClient *c) {
719 size_t querybuf_size = sdsAllocSize(c->querybuf);
720 time_t idletime = server.unixtime - c->lastinteraction;
721
722 /* There are two conditions to resize the query buffer:
723 * 1) Query buffer is > BIG_ARG and too big for latest peak.
724 * 2) Client is inactive and the buffer is bigger than 1k. */
725 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
726 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
727 (querybuf_size > 1024 && idletime > 2))
728 {
729 /* Only resize the query buffer if it is actually wasting space. */
730 if (sdsavail(c->querybuf) > 1024) {
731 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
732 }
733 }
734 /* Reset the peak again to capture the peak memory usage in the next
735 * cycle. */
736 c->querybuf_peak = 0;
737 return 0;
738 }
739
740 void clientsCron(void) {
741 /* Make sure to process at least 1/100 of clients per call.
742 * Since this function is called 10 times per second we are sure that
743 * in the worst case we process all the clients in 10 seconds.
744 * In normal conditions (a reasonable number of clients) we process
745 * all the clients in a shorter time. */
746 int numclients = listLength(server.clients);
747 int iterations = numclients/100;
748
749 if (iterations < 50)
750 iterations = (numclients < 50) ? numclients : 50;
751 while(listLength(server.clients) && iterations--) {
752 redisClient *c;
753 listNode *head;
754
755 /* Rotate the list, take the current head, process.
756 * This way if the client must be removed from the list it's the
757 * first element and we don't incur into O(N) computation. */
758 listRotate(server.clients);
759 head = listFirst(server.clients);
760 c = listNodeValue(head);
761 /* The following functions do different service checks on the client.
762 * The protocol is that they return non-zero if the client was
763 * terminated. */
764 if (clientsCronHandleTimeout(c)) continue;
765 if (clientsCronResizeQueryBuffer(c)) continue;
766 }
767 }
768
769 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
770 int j, loops = server.cronloops;
771 REDIS_NOTUSED(eventLoop);
772 REDIS_NOTUSED(id);
773 REDIS_NOTUSED(clientData);
774
775 /* Software watchdog: deliver the SIGALRM that will reach the signal
776 * handler if we don't return here fast enough. */
777 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
778
779 /* We take a cached value of the unix time in the global state because
780 * with virtual memory and aging there is to store the current time
781 * in objects at every object access, and accuracy is not needed.
782 * To access a global var is faster than calling time(NULL) */
783 server.unixtime = time(NULL);
784
785 trackOperationsPerSecond();
786
787 /* We have just 22 bits per object for LRU information.
788 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
789 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
790 *
791 * Note that even if this will wrap after 1.5 years it's not a problem,
792 * everything will still work but just some object will appear younger
793 * to Redis. But for this to happen a given object should never be touched
794 * for 1.5 years.
795 *
796 * Note that you can change the resolution altering the
797 * REDIS_LRU_CLOCK_RESOLUTION define.
798 */
799 updateLRUClock();
800
801 /* Record the max memory used since the server was started. */
802 if (zmalloc_used_memory() > server.stat_peak_memory)
803 server.stat_peak_memory = zmalloc_used_memory();
804
805 /* We received a SIGTERM, shutting down here in a safe way, as it is
806 * not ok doing so inside the signal handler. */
807 if (server.shutdown_asap) {
808 if (prepareForShutdown(0) == REDIS_OK) exit(0);
809 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
810 }
811
812 /* Show some info about non-empty databases */
813 for (j = 0; j < server.dbnum; j++) {
814 long long size, used, vkeys;
815
816 size = dictSlots(server.db[j].dict);
817 used = dictSize(server.db[j].dict);
818 vkeys = dictSize(server.db[j].expires);
819 if (!(loops % 50) && (used || vkeys)) {
820 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
821 /* dictPrintStats(server.dict); */
822 }
823 }
824
825 /* We don't want to resize the hash tables while a bacground saving
826 * is in progress: the saving child is created using fork() that is
827 * implemented with a copy-on-write semantic in most modern systems, so
828 * if we resize the HT while there is the saving child at work actually
829 * a lot of memory movements in the parent will cause a lot of pages
830 * copied. */
831 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
832 if (!(loops % 10)) tryResizeHashTables();
833 if (server.activerehashing) incrementallyRehash();
834 }
835
836 /* Show information about connected clients */
837 if (!(loops % 50)) {
838 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
839 listLength(server.clients)-listLength(server.slaves),
840 listLength(server.slaves),
841 zmalloc_used_memory());
842 }
843
844 /* We need to do a few operations on clients asynchronously. */
845 clientsCron();
846
847 /* Start a scheduled AOF rewrite if this was requested by the user while
848 * a BGSAVE was in progress. */
849 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
850 server.aof_rewrite_scheduled)
851 {
852 rewriteAppendOnlyFileBackground();
853 }
854
855 /* Check if a background saving or AOF rewrite in progress terminated. */
856 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
857 int statloc;
858 pid_t pid;
859
860 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
861 int exitcode = WEXITSTATUS(statloc);
862 int bysignal = 0;
863
864 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
865
866 if (pid == server.rdb_child_pid) {
867 backgroundSaveDoneHandler(exitcode,bysignal);
868 } else {
869 backgroundRewriteDoneHandler(exitcode,bysignal);
870 }
871 updateDictResizePolicy();
872 }
873 } else {
874 /* If there is not a background saving/rewrite in progress check if
875 * we have to save/rewrite now */
876 for (j = 0; j < server.saveparamslen; j++) {
877 struct saveparam *sp = server.saveparams+j;
878
879 if (server.dirty >= sp->changes &&
880 server.unixtime-server.lastsave > sp->seconds) {
881 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
882 sp->changes, sp->seconds);
883 rdbSaveBackground(server.rdb_filename);
884 break;
885 }
886 }
887
888 /* Trigger an AOF rewrite if needed */
889 if (server.rdb_child_pid == -1 &&
890 server.aof_child_pid == -1 &&
891 server.aof_rewrite_perc &&
892 server.aof_current_size > server.aof_rewrite_min_size)
893 {
894 long long base = server.aof_rewrite_base_size ?
895 server.aof_rewrite_base_size : 1;
896 long long growth = (server.aof_current_size*100/base) - 100;
897 if (growth >= server.aof_rewrite_perc) {
898 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
899 rewriteAppendOnlyFileBackground();
900 }
901 }
902 }
903
904
905 /* If we postponed an AOF buffer flush, let's try to do it every time the
906 * cron function is called. */
907 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
908
909 /* Expire a few keys per cycle, only if this is a master.
910 * On slaves we wait for DEL operations synthesized by the master
911 * in order to guarantee a strict consistency. */
912 if (server.masterhost == NULL) activeExpireCycle();
913
914 /* Close clients that need to be closed asynchronous */
915 freeClientsInAsyncFreeQueue();
916
917 /* Replication cron function -- used to reconnect to master and
918 * to detect transfer failures. */
919 if (!(loops % 10)) replicationCron();
920
921 /* Run other sub-systems specific cron jobs */
922 if (server.cluster_enabled && !(loops % 10)) clusterCron();
923
924 server.cronloops++;
925 return 100;
926 }
927
928 /* This function gets called every time Redis is entering the
929 * main loop of the event driven library, that is, before to sleep
930 * for ready file descriptors. */
931 void beforeSleep(struct aeEventLoop *eventLoop) {
932 REDIS_NOTUSED(eventLoop);
933 listNode *ln;
934 redisClient *c;
935
936 /* Try to process pending commands for clients that were just unblocked. */
937 while (listLength(server.unblocked_clients)) {
938 ln = listFirst(server.unblocked_clients);
939 redisAssert(ln != NULL);
940 c = ln->value;
941 listDelNode(server.unblocked_clients,ln);
942 c->flags &= ~REDIS_UNBLOCKED;
943
944 /* Process remaining data in the input buffer. */
945 if (c->querybuf && sdslen(c->querybuf) > 0) {
946 server.current_client = c;
947 processInputBuffer(c);
948 server.current_client = NULL;
949 }
950 }
951
952 /* Write the AOF buffer on disk */
953 flushAppendOnlyFile(0);
954 }
955
956 /* =========================== Server initialization ======================== */
957
958 void createSharedObjects(void) {
959 int j;
960
961 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
962 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
963 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
964 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
965 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
966 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
967 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
968 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
969 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
970 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
971 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
972 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
973 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
974 "-ERR Operation against a key holding the wrong kind of value\r\n"));
975 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
976 "-ERR no such key\r\n"));
977 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
978 "-ERR syntax error\r\n"));
979 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
980 "-ERR source and destination objects are the same\r\n"));
981 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
982 "-ERR index out of range\r\n"));
983 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
984 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
985 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
986 "-LOADING Redis is loading the dataset in memory\r\n"));
987 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
988 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
989 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
990 "-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"));
991 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
992 "-READONLY You can't write against a read only slave.\r\n"));
993 shared.oomerr = createObject(REDIS_STRING,sdsnew(
994 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
995 shared.space = createObject(REDIS_STRING,sdsnew(" "));
996 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
997 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
998
999 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1000 shared.select[j] = createObject(REDIS_STRING,
1001 sdscatprintf(sdsempty(),"select %d\r\n", j));
1002 }
1003 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1004 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1005 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1006 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1007 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1008 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1009 shared.del = createStringObject("DEL",3);
1010 shared.rpop = createStringObject("RPOP",4);
1011 shared.lpop = createStringObject("LPOP",4);
1012 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1013 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1014 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1015 }
1016 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1017 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1018 sdscatprintf(sdsempty(),"*%d\r\n",j));
1019 shared.bulkhdr[j] = createObject(REDIS_STRING,
1020 sdscatprintf(sdsempty(),"$%d\r\n",j));
1021 }
1022 }
1023
1024 void initServerConfig() {
1025 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1026 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1027 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1028 server.port = REDIS_SERVERPORT;
1029 server.bindaddr = NULL;
1030 server.unixsocket = NULL;
1031 server.unixsocketperm = 0;
1032 server.ipfd = -1;
1033 server.sofd = -1;
1034 server.dbnum = REDIS_DEFAULT_DBNUM;
1035 server.verbosity = REDIS_NOTICE;
1036 server.maxidletime = REDIS_MAXIDLETIME;
1037 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1038 server.saveparams = NULL;
1039 server.loading = 0;
1040 server.logfile = NULL; /* NULL = log on standard output */
1041 server.syslog_enabled = 0;
1042 server.syslog_ident = zstrdup("redis");
1043 server.syslog_facility = LOG_LOCAL0;
1044 server.daemonize = 0;
1045 server.aof_state = REDIS_AOF_OFF;
1046 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1047 server.aof_no_fsync_on_rewrite = 0;
1048 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1049 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1050 server.aof_rewrite_base_size = 0;
1051 server.aof_rewrite_scheduled = 0;
1052 server.aof_last_fsync = time(NULL);
1053 server.aof_delayed_fsync = 0;
1054 server.aof_fd = -1;
1055 server.aof_selected_db = -1; /* Make sure the first time will not match */
1056 server.aof_flush_postponed_start = 0;
1057 server.pidfile = zstrdup("/var/run/redis.pid");
1058 server.rdb_filename = zstrdup("dump.rdb");
1059 server.aof_filename = zstrdup("appendonly.aof");
1060 server.requirepass = NULL;
1061 server.rdb_compression = 1;
1062 server.rdb_checksum = 1;
1063 server.activerehashing = 1;
1064 server.maxclients = REDIS_MAX_CLIENTS;
1065 server.bpop_blocked_clients = 0;
1066 server.maxmemory = 0;
1067 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1068 server.maxmemory_samples = 3;
1069 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1070 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1071 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1072 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1073 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1074 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1075 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1076 server.shutdown_asap = 0;
1077 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1078 server.repl_timeout = REDIS_REPL_TIMEOUT;
1079 server.cluster_enabled = 0;
1080 server.cluster.configfile = zstrdup("nodes.conf");
1081 server.lua_caller = NULL;
1082 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1083 server.lua_client = NULL;
1084 server.lua_timedout = 0;
1085
1086 updateLRUClock();
1087 resetServerSaveParams();
1088
1089 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1090 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1091 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1092 /* Replication related */
1093 server.masterauth = NULL;
1094 server.masterhost = NULL;
1095 server.masterport = 6379;
1096 server.master = NULL;
1097 server.repl_state = REDIS_REPL_NONE;
1098 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1099 server.repl_serve_stale_data = 1;
1100 server.repl_slave_ro = 1;
1101 server.repl_down_since = time(NULL);
1102
1103 /* Client output buffer limits */
1104 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1105 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1106 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1107 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1108 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1109 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1110 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1111 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1112 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1113
1114 /* Double constants initialization */
1115 R_Zero = 0.0;
1116 R_PosInf = 1.0/R_Zero;
1117 R_NegInf = -1.0/R_Zero;
1118 R_Nan = R_Zero/R_Zero;
1119
1120 /* Command table -- we intiialize it here as it is part of the
1121 * initial configuration, since command names may be changed via
1122 * redis.conf using the rename-command directive. */
1123 server.commands = dictCreate(&commandTableDictType,NULL);
1124 populateCommandTable();
1125 server.delCommand = lookupCommandByCString("del");
1126 server.multiCommand = lookupCommandByCString("multi");
1127 server.lpushCommand = lookupCommandByCString("lpush");
1128
1129 /* Slow log */
1130 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1131 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1132
1133 /* Debugging */
1134 server.assert_failed = "<no assertion failed>";
1135 server.assert_file = "<no file>";
1136 server.assert_line = 0;
1137 server.bug_report_start = 0;
1138 server.watchdog_period = 0;
1139 }
1140
1141 /* This function will try to raise the max number of open files accordingly to
1142 * the configured max number of clients. It will also account for 32 additional
1143 * file descriptors as we need a few more for persistence, listening
1144 * sockets, log files and so forth.
1145 *
1146 * If it will not be possible to set the limit accordingly to the configured
1147 * max number of clients, the function will do the reverse setting
1148 * server.maxclients to the value that we can actually handle. */
1149 void adjustOpenFilesLimit(void) {
1150 rlim_t maxfiles = server.maxclients+32;
1151 struct rlimit limit;
1152
1153 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1154 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1155 strerror(errno));
1156 server.maxclients = 1024-32;
1157 } else {
1158 rlim_t oldlimit = limit.rlim_cur;
1159
1160 /* Set the max number of files if the current limit is not enough
1161 * for our needs. */
1162 if (oldlimit < maxfiles) {
1163 rlim_t f;
1164
1165 f = maxfiles;
1166 while(f > oldlimit) {
1167 limit.rlim_cur = f;
1168 limit.rlim_max = f;
1169 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1170 f -= 128;
1171 }
1172 if (f < oldlimit) f = oldlimit;
1173 if (f != maxfiles) {
1174 server.maxclients = f-32;
1175 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1176 (int) maxfiles, strerror(errno), (int) server.maxclients);
1177 } else {
1178 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1179 (int) maxfiles);
1180 }
1181 }
1182 }
1183 }
1184
1185 void initServer() {
1186 int j;
1187
1188 signal(SIGHUP, SIG_IGN);
1189 signal(SIGPIPE, SIG_IGN);
1190 setupSignalHandlers();
1191
1192 if (server.syslog_enabled) {
1193 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1194 server.syslog_facility);
1195 }
1196
1197 server.current_client = NULL;
1198 server.clients = listCreate();
1199 server.clients_to_close = listCreate();
1200 server.slaves = listCreate();
1201 server.monitors = listCreate();
1202 server.unblocked_clients = listCreate();
1203
1204 createSharedObjects();
1205 adjustOpenFilesLimit();
1206 server.el = aeCreateEventLoop(server.maxclients+1024);
1207 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1208
1209 if (server.port != 0) {
1210 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1211 if (server.ipfd == ANET_ERR) {
1212 redisLog(REDIS_WARNING, "Opening port %d: %s",
1213 server.port, server.neterr);
1214 exit(1);
1215 }
1216 }
1217 if (server.unixsocket != NULL) {
1218 unlink(server.unixsocket); /* don't care if this fails */
1219 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1220 if (server.sofd == ANET_ERR) {
1221 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1222 exit(1);
1223 }
1224 }
1225 if (server.ipfd < 0 && server.sofd < 0) {
1226 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1227 exit(1);
1228 }
1229 for (j = 0; j < server.dbnum; j++) {
1230 server.db[j].dict = dictCreate(&dbDictType,NULL);
1231 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1232 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1233 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1234 server.db[j].id = j;
1235 }
1236 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1237 server.pubsub_patterns = listCreate();
1238 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1239 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1240 server.cronloops = 0;
1241 server.rdb_child_pid = -1;
1242 server.aof_child_pid = -1;
1243 server.aof_rewrite_buf = sdsempty();
1244 server.aof_buf = sdsempty();
1245 server.lastsave = time(NULL);
1246 server.dirty = 0;
1247 server.stat_numcommands = 0;
1248 server.stat_numconnections = 0;
1249 server.stat_expiredkeys = 0;
1250 server.stat_evictedkeys = 0;
1251 server.stat_starttime = time(NULL);
1252 server.stat_keyspace_misses = 0;
1253 server.stat_keyspace_hits = 0;
1254 server.stat_peak_memory = 0;
1255 server.stat_fork_time = 0;
1256 server.stat_rejected_conn = 0;
1257 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1258 server.ops_sec_idx = 0;
1259 server.ops_sec_last_sample_time = mstime();
1260 server.ops_sec_last_sample_ops = 0;
1261 server.unixtime = time(NULL);
1262 server.lastbgsave_status = REDIS_OK;
1263 server.stop_writes_on_bgsave_err = 1;
1264 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1265 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1266 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1267 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1268 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1269
1270 if (server.aof_state == REDIS_AOF_ON) {
1271 server.aof_fd = open(server.aof_filename,
1272 O_WRONLY|O_APPEND|O_CREAT,0644);
1273 if (server.aof_fd == -1) {
1274 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1275 strerror(errno));
1276 exit(1);
1277 }
1278 }
1279
1280 /* 32 bit instances are limited to 4GB of address space, so if there is
1281 * no explicit limit in the user provided configuration we set a limit
1282 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1283 * useless crashes of the Redis instance. */
1284 if (server.arch_bits == 32 && server.maxmemory == 0) {
1285 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1286 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1287 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1288 }
1289
1290 if (server.cluster_enabled) clusterInit();
1291 scriptingInit();
1292 slowlogInit();
1293 bioInit();
1294 }
1295
1296 /* Populates the Redis Command Table starting from the hard coded list
1297 * we have on top of redis.c file. */
1298 void populateCommandTable(void) {
1299 int j;
1300 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1301
1302 for (j = 0; j < numcommands; j++) {
1303 struct redisCommand *c = redisCommandTable+j;
1304 char *f = c->sflags;
1305 int retval;
1306
1307 while(*f != '\0') {
1308 switch(*f) {
1309 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1310 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1311 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1312 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1313 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1314 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1315 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1316 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1317 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1318 default: redisPanic("Unsupported command flag"); break;
1319 }
1320 f++;
1321 }
1322
1323 retval = dictAdd(server.commands, sdsnew(c->name), c);
1324 assert(retval == DICT_OK);
1325 }
1326 }
1327
1328 void resetCommandTableStats(void) {
1329 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1330 int j;
1331
1332 for (j = 0; j < numcommands; j++) {
1333 struct redisCommand *c = redisCommandTable+j;
1334
1335 c->microseconds = 0;
1336 c->calls = 0;
1337 }
1338 }
1339
1340 /* ========================== Redis OP Array API ============================ */
1341
1342 void redisOpArrayInit(redisOpArray *oa) {
1343 oa->ops = NULL;
1344 oa->numops = 0;
1345 }
1346
1347 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1348 robj **argv, int argc, int target)
1349 {
1350 redisOp *op;
1351
1352 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1353 op = oa->ops+oa->numops;
1354 op->cmd = cmd;
1355 op->dbid = dbid;
1356 op->argv = argv;
1357 op->argc = argc;
1358 op->target = target;
1359 oa->numops++;
1360 return oa->numops;
1361 }
1362
1363 void redisOpArrayFree(redisOpArray *oa) {
1364 while(oa->numops) {
1365 int j;
1366 redisOp *op;
1367
1368 oa->numops--;
1369 op = oa->ops+oa->numops;
1370 for (j = 0; j < op->argc; j++)
1371 decrRefCount(op->argv[j]);
1372 zfree(op->argv);
1373 }
1374 zfree(oa->ops);
1375 }
1376
1377 /* ====================== Commands lookup and execution ===================== */
1378
1379 struct redisCommand *lookupCommand(sds name) {
1380 return dictFetchValue(server.commands, name);
1381 }
1382
1383 struct redisCommand *lookupCommandByCString(char *s) {
1384 struct redisCommand *cmd;
1385 sds name = sdsnew(s);
1386
1387 cmd = dictFetchValue(server.commands, name);
1388 sdsfree(name);
1389 return cmd;
1390 }
1391
1392 /* Propagate the specified command (in the context of the specified database id)
1393 * to AOF, Slaves and Monitors.
1394 *
1395 * flags are an xor between:
1396 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1397 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1398 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1399 */
1400 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1401 int flags)
1402 {
1403 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1404 feedAppendOnlyFile(cmd,dbid,argv,argc);
1405 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1406 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1407 }
1408
1409 /* Used inside commands to schedule the propagation of additional commands
1410 * after the current command is propagated to AOF / Replication. */
1411 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1412 int target)
1413 {
1414 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1415 }
1416
1417 /* Call() is the core of Redis execution of a command */
1418 void call(redisClient *c, int flags) {
1419 long long dirty, start = ustime(), duration;
1420
1421 /* Sent the command to clients in MONITOR mode, only if the commands are
1422 * not geneated from reading an AOF. */
1423 if (listLength(server.monitors) && !server.loading)
1424 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1425
1426 /* Call the command. */
1427 redisOpArrayInit(&server.also_propagate);
1428 dirty = server.dirty;
1429 c->cmd->proc(c);
1430 dirty = server.dirty-dirty;
1431 duration = ustime()-start;
1432
1433 /* When EVAL is called loading the AOF we don't want commands called
1434 * from Lua to go into the slowlog or to populate statistics. */
1435 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1436 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1437
1438 /* Log the command into the Slow log if needed, and populate the
1439 * per-command statistics that we show in INFO commandstats. */
1440 if (flags & REDIS_CALL_SLOWLOG)
1441 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1442 if (flags & REDIS_CALL_STATS) {
1443 c->cmd->microseconds += duration;
1444 c->cmd->calls++;
1445 }
1446
1447 /* Propagate the command into the AOF and replication link */
1448 if (flags & REDIS_CALL_PROPAGATE) {
1449 int flags = REDIS_PROPAGATE_NONE;
1450
1451 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1452 flags |= REDIS_PROPAGATE_REPL;
1453 if (dirty)
1454 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1455 if (flags != REDIS_PROPAGATE_NONE)
1456 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1457 }
1458 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1459 * PUSH command. */
1460 if (server.also_propagate.numops) {
1461 int j;
1462 redisOp *rop;
1463
1464 for (j = 0; j < server.also_propagate.numops; j++) {
1465 rop = &server.also_propagate.ops[j];
1466 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1467 }
1468 redisOpArrayFree(&server.also_propagate);
1469 }
1470 server.stat_numcommands++;
1471 }
1472
1473 /* If this function gets called we already read a whole
1474 * command, argments are in the client argv/argc fields.
1475 * processCommand() execute the command or prepare the
1476 * server for a bulk read from the client.
1477 *
1478 * If 1 is returned the client is still alive and valid and
1479 * and other operations can be performed by the caller. Otherwise
1480 * if 0 is returned the client was destroied (i.e. after QUIT). */
1481 int processCommand(redisClient *c) {
1482 /* The QUIT command is handled separately. Normal command procs will
1483 * go through checking for replication and QUIT will cause trouble
1484 * when FORCE_REPLICATION is enabled and would be implemented in
1485 * a regular command proc. */
1486 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1487 addReply(c,shared.ok);
1488 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1489 return REDIS_ERR;
1490 }
1491
1492 /* Now lookup the command and check ASAP about trivial error conditions
1493 * such as wrong arity, bad command name and so forth. */
1494 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1495 if (!c->cmd) {
1496 addReplyErrorFormat(c,"unknown command '%s'",
1497 (char*)c->argv[0]->ptr);
1498 return REDIS_OK;
1499 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1500 (c->argc < -c->cmd->arity)) {
1501 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1502 c->cmd->name);
1503 return REDIS_OK;
1504 }
1505
1506 /* Check if the user is authenticated */
1507 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1508 {
1509 addReplyError(c,"operation not permitted");
1510 return REDIS_OK;
1511 }
1512
1513 /* If cluster is enabled, redirect here */
1514 if (server.cluster_enabled &&
1515 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1516 int hashslot;
1517
1518 if (server.cluster.state != REDIS_CLUSTER_OK) {
1519 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1520 return REDIS_OK;
1521 } else {
1522 int ask;
1523 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1524 if (n == NULL) {
1525 addReplyError(c,"Multi keys request invalid in cluster");
1526 return REDIS_OK;
1527 } else if (n != server.cluster.myself) {
1528 addReplySds(c,sdscatprintf(sdsempty(),
1529 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1530 hashslot,n->ip,n->port));
1531 return REDIS_OK;
1532 }
1533 }
1534 }
1535
1536 /* Handle the maxmemory directive.
1537 *
1538 * First we try to free some memory if possible (if there are volatile
1539 * keys in the dataset). If there are not the only thing we can do
1540 * is returning an error. */
1541 if (server.maxmemory) {
1542 int retval = freeMemoryIfNeeded();
1543 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1544 addReply(c, shared.oomerr);
1545 return REDIS_OK;
1546 }
1547 }
1548
1549 /* Don't accept write commands if there are problems persisting on disk. */
1550 if (server.stop_writes_on_bgsave_err &&
1551 server.saveparamslen > 0
1552 && server.lastbgsave_status == REDIS_ERR &&
1553 c->cmd->flags & REDIS_CMD_WRITE)
1554 {
1555 addReply(c, shared.bgsaveerr);
1556 return REDIS_OK;
1557 }
1558
1559 /* Don't accept wirte commands if this is a read only slave. But
1560 * accept write commands if this is our master. */
1561 if (server.masterhost && server.repl_slave_ro &&
1562 !(c->flags & REDIS_MASTER) &&
1563 c->cmd->flags & REDIS_CMD_WRITE)
1564 {
1565 addReply(c, shared.roslaveerr);
1566 return REDIS_OK;
1567 }
1568
1569 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1570 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1571 &&
1572 c->cmd->proc != subscribeCommand &&
1573 c->cmd->proc != unsubscribeCommand &&
1574 c->cmd->proc != psubscribeCommand &&
1575 c->cmd->proc != punsubscribeCommand) {
1576 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1577 return REDIS_OK;
1578 }
1579
1580 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1581 * we are a slave with a broken link with master. */
1582 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1583 server.repl_serve_stale_data == 0 &&
1584 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1585 {
1586 addReplyError(c,
1587 "link with MASTER is down and slave-serve-stale-data is set to no");
1588 return REDIS_OK;
1589 }
1590
1591 /* Loading DB? Return an error if the command is not INFO */
1592 if (server.loading && c->cmd->proc != infoCommand) {
1593 addReply(c, shared.loadingerr);
1594 return REDIS_OK;
1595 }
1596
1597 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1598 if (server.lua_timedout &&
1599 !(c->cmd->proc == shutdownCommand &&
1600 c->argc == 2 &&
1601 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1602 !(c->cmd->proc == scriptCommand &&
1603 c->argc == 2 &&
1604 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1605 {
1606 addReply(c, shared.slowscripterr);
1607 return REDIS_OK;
1608 }
1609
1610 /* Exec the command */
1611 if (c->flags & REDIS_MULTI &&
1612 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1613 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1614 {
1615 queueMultiCommand(c);
1616 addReply(c,shared.queued);
1617 } else {
1618 call(c,REDIS_CALL_FULL);
1619 }
1620 return REDIS_OK;
1621 }
1622
1623 /*================================== Shutdown =============================== */
1624
1625 int prepareForShutdown(int flags) {
1626 int save = flags & REDIS_SHUTDOWN_SAVE;
1627 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1628
1629 redisLog(REDIS_WARNING,"User requested shutdown...");
1630 /* Kill the saving child if there is a background saving in progress.
1631 We want to avoid race conditions, for instance our saving child may
1632 overwrite the synchronous saving did by SHUTDOWN. */
1633 if (server.rdb_child_pid != -1) {
1634 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1635 kill(server.rdb_child_pid,SIGKILL);
1636 rdbRemoveTempFile(server.rdb_child_pid);
1637 }
1638 if (server.aof_state != REDIS_AOF_OFF) {
1639 /* Kill the AOF saving child as the AOF we already have may be longer
1640 * but contains the full dataset anyway. */
1641 if (server.aof_child_pid != -1) {
1642 redisLog(REDIS_WARNING,
1643 "There is a child rewriting the AOF. Killing it!");
1644 kill(server.aof_child_pid,SIGKILL);
1645 }
1646 /* Append only file: fsync() the AOF and exit */
1647 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1648 aof_fsync(server.aof_fd);
1649 }
1650 if ((server.saveparamslen > 0 && !nosave) || save) {
1651 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1652 /* Snapshotting. Perform a SYNC SAVE and exit */
1653 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1654 /* Ooops.. error saving! The best we can do is to continue
1655 * operating. Note that if there was a background saving process,
1656 * in the next cron() Redis will be notified that the background
1657 * saving aborted, handling special stuff like slaves pending for
1658 * synchronization... */
1659 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1660 return REDIS_ERR;
1661 }
1662 }
1663 if (server.daemonize) {
1664 redisLog(REDIS_NOTICE,"Removing the pid file.");
1665 unlink(server.pidfile);
1666 }
1667 /* Close the listening sockets. Apparently this allows faster restarts. */
1668 if (server.ipfd != -1) close(server.ipfd);
1669 if (server.sofd != -1) close(server.sofd);
1670 if (server.unixsocket) {
1671 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1672 unlink(server.unixsocket); /* don't care if this fails */
1673 }
1674
1675 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1676 return REDIS_OK;
1677 }
1678
1679 /*================================== Commands =============================== */
1680
1681 void authCommand(redisClient *c) {
1682 if (!server.requirepass) {
1683 addReplyError(c,"Client sent AUTH, but no password is set");
1684 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1685 c->authenticated = 1;
1686 addReply(c,shared.ok);
1687 } else {
1688 c->authenticated = 0;
1689 addReplyError(c,"invalid password");
1690 }
1691 }
1692
1693 void pingCommand(redisClient *c) {
1694 addReply(c,shared.pong);
1695 }
1696
1697 void echoCommand(redisClient *c) {
1698 addReplyBulk(c,c->argv[1]);
1699 }
1700
1701 void timeCommand(redisClient *c) {
1702 struct timeval tv;
1703
1704 /* gettimeofday() can only fail if &tv is a bad addresss so we
1705 * don't check for errors. */
1706 gettimeofday(&tv,NULL);
1707 addReplyMultiBulkLen(c,2);
1708 addReplyBulkLongLong(c,tv.tv_sec);
1709 addReplyBulkLongLong(c,tv.tv_usec);
1710 }
1711
1712 /* Convert an amount of bytes into a human readable string in the form
1713 * of 100B, 2G, 100M, 4K, and so forth. */
1714 void bytesToHuman(char *s, unsigned long long n) {
1715 double d;
1716
1717 if (n < 1024) {
1718 /* Bytes */
1719 sprintf(s,"%lluB",n);
1720 return;
1721 } else if (n < (1024*1024)) {
1722 d = (double)n/(1024);
1723 sprintf(s,"%.2fK",d);
1724 } else if (n < (1024LL*1024*1024)) {
1725 d = (double)n/(1024*1024);
1726 sprintf(s,"%.2fM",d);
1727 } else if (n < (1024LL*1024*1024*1024)) {
1728 d = (double)n/(1024LL*1024*1024);
1729 sprintf(s,"%.2fG",d);
1730 }
1731 }
1732
1733 /* Create the string returned by the INFO command. This is decoupled
1734 * by the INFO command itself as we need to report the same information
1735 * on memory corruption problems. */
1736 sds genRedisInfoString(char *section) {
1737 sds info = sdsempty();
1738 time_t uptime = server.unixtime-server.stat_starttime;
1739 int j, numcommands;
1740 struct rusage self_ru, c_ru;
1741 unsigned long lol, bib;
1742 int allsections = 0, defsections = 0;
1743 int sections = 0;
1744
1745 if (section) {
1746 allsections = strcasecmp(section,"all") == 0;
1747 defsections = strcasecmp(section,"default") == 0;
1748 }
1749
1750 getrusage(RUSAGE_SELF, &self_ru);
1751 getrusage(RUSAGE_CHILDREN, &c_ru);
1752 getClientsMaxBuffers(&lol,&bib);
1753
1754 /* Server */
1755 if (allsections || defsections || !strcasecmp(section,"server")) {
1756 struct utsname name;
1757
1758 if (sections++) info = sdscat(info,"\r\n");
1759 uname(&name);
1760 info = sdscatprintf(info,
1761 "# Server\r\n"
1762 "redis_version:%s\r\n"
1763 "redis_git_sha1:%s\r\n"
1764 "redis_git_dirty:%d\r\n"
1765 "os:%s %s %s\r\n"
1766 "arch_bits:%d\r\n"
1767 "multiplexing_api:%s\r\n"
1768 "gcc_version:%d.%d.%d\r\n"
1769 "process_id:%ld\r\n"
1770 "run_id:%s\r\n"
1771 "tcp_port:%d\r\n"
1772 "uptime_in_seconds:%ld\r\n"
1773 "uptime_in_days:%ld\r\n"
1774 "lru_clock:%ld\r\n",
1775 REDIS_VERSION,
1776 redisGitSHA1(),
1777 strtol(redisGitDirty(),NULL,10) > 0,
1778 name.sysname, name.release, name.machine,
1779 server.arch_bits,
1780 aeGetApiName(),
1781 #ifdef __GNUC__
1782 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1783 #else
1784 0,0,0,
1785 #endif
1786 (long) getpid(),
1787 server.runid,
1788 server.port,
1789 uptime,
1790 uptime/(3600*24),
1791 (unsigned long) server.lruclock);
1792 }
1793
1794 /* Clients */
1795 if (allsections || defsections || !strcasecmp(section,"clients")) {
1796 if (sections++) info = sdscat(info,"\r\n");
1797 info = sdscatprintf(info,
1798 "# Clients\r\n"
1799 "connected_clients:%lu\r\n"
1800 "client_longest_output_list:%lu\r\n"
1801 "client_biggest_input_buf:%lu\r\n"
1802 "blocked_clients:%d\r\n",
1803 listLength(server.clients)-listLength(server.slaves),
1804 lol, bib,
1805 server.bpop_blocked_clients);
1806 }
1807
1808 /* Memory */
1809 if (allsections || defsections || !strcasecmp(section,"memory")) {
1810 char hmem[64];
1811 char peak_hmem[64];
1812
1813 bytesToHuman(hmem,zmalloc_used_memory());
1814 bytesToHuman(peak_hmem,server.stat_peak_memory);
1815 if (sections++) info = sdscat(info,"\r\n");
1816 info = sdscatprintf(info,
1817 "# Memory\r\n"
1818 "used_memory:%zu\r\n"
1819 "used_memory_human:%s\r\n"
1820 "used_memory_rss:%zu\r\n"
1821 "used_memory_peak:%zu\r\n"
1822 "used_memory_peak_human:%s\r\n"
1823 "used_memory_lua:%lld\r\n"
1824 "mem_fragmentation_ratio:%.2f\r\n"
1825 "mem_allocator:%s\r\n",
1826 zmalloc_used_memory(),
1827 hmem,
1828 zmalloc_get_rss(),
1829 server.stat_peak_memory,
1830 peak_hmem,
1831 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1832 zmalloc_get_fragmentation_ratio(),
1833 ZMALLOC_LIB
1834 );
1835 }
1836
1837 /* Persistence */
1838 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1839 if (sections++) info = sdscat(info,"\r\n");
1840 info = sdscatprintf(info,
1841 "# Persistence\r\n"
1842 "loading:%d\r\n"
1843 "aof_enabled:%d\r\n"
1844 "changes_since_last_save:%lld\r\n"
1845 "bgsave_in_progress:%d\r\n"
1846 "last_save_time:%ld\r\n"
1847 "last_bgsave_status:%s\r\n"
1848 "bgrewriteaof_in_progress:%d\r\n"
1849 "bgrewriteaof_scheduled:%d\r\n",
1850 server.loading,
1851 server.aof_state != REDIS_AOF_OFF,
1852 server.dirty,
1853 server.rdb_child_pid != -1,
1854 server.lastsave,
1855 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1856 server.aof_child_pid != -1,
1857 server.aof_rewrite_scheduled);
1858
1859 if (server.aof_state != REDIS_AOF_OFF) {
1860 info = sdscatprintf(info,
1861 "aof_current_size:%lld\r\n"
1862 "aof_base_size:%lld\r\n"
1863 "aof_pending_rewrite:%d\r\n"
1864 "aof_buffer_length:%zu\r\n"
1865 "aof_pending_bio_fsync:%llu\r\n"
1866 "aof_delayed_fsync:%lu\r\n",
1867 (long long) server.aof_current_size,
1868 (long long) server.aof_rewrite_base_size,
1869 server.aof_rewrite_scheduled,
1870 sdslen(server.aof_buf),
1871 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1872 server.aof_delayed_fsync);
1873 }
1874
1875 if (server.loading) {
1876 double perc;
1877 time_t eta, elapsed;
1878 off_t remaining_bytes = server.loading_total_bytes-
1879 server.loading_loaded_bytes;
1880
1881 perc = ((double)server.loading_loaded_bytes /
1882 server.loading_total_bytes) * 100;
1883
1884 elapsed = server.unixtime-server.loading_start_time;
1885 if (elapsed == 0) {
1886 eta = 1; /* A fake 1 second figure if we don't have
1887 enough info */
1888 } else {
1889 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1890 }
1891
1892 info = sdscatprintf(info,
1893 "loading_start_time:%ld\r\n"
1894 "loading_total_bytes:%llu\r\n"
1895 "loading_loaded_bytes:%llu\r\n"
1896 "loading_loaded_perc:%.2f\r\n"
1897 "loading_eta_seconds:%ld\r\n"
1898 ,(unsigned long) server.loading_start_time,
1899 (unsigned long long) server.loading_total_bytes,
1900 (unsigned long long) server.loading_loaded_bytes,
1901 perc,
1902 eta
1903 );
1904 }
1905 }
1906
1907 /* Stats */
1908 if (allsections || defsections || !strcasecmp(section,"stats")) {
1909 if (sections++) info = sdscat(info,"\r\n");
1910 info = sdscatprintf(info,
1911 "# Stats\r\n"
1912 "total_connections_received:%lld\r\n"
1913 "total_commands_processed:%lld\r\n"
1914 "instantaneous_ops_per_sec:%lld\r\n"
1915 "rejected_connections:%lld\r\n"
1916 "expired_keys:%lld\r\n"
1917 "evicted_keys:%lld\r\n"
1918 "keyspace_hits:%lld\r\n"
1919 "keyspace_misses:%lld\r\n"
1920 "pubsub_channels:%ld\r\n"
1921 "pubsub_patterns:%lu\r\n"
1922 "latest_fork_usec:%lld\r\n",
1923 server.stat_numconnections,
1924 server.stat_numcommands,
1925 getOperationsPerSecond(),
1926 server.stat_rejected_conn,
1927 server.stat_expiredkeys,
1928 server.stat_evictedkeys,
1929 server.stat_keyspace_hits,
1930 server.stat_keyspace_misses,
1931 dictSize(server.pubsub_channels),
1932 listLength(server.pubsub_patterns),
1933 server.stat_fork_time);
1934 }
1935
1936 /* Replication */
1937 if (allsections || defsections || !strcasecmp(section,"replication")) {
1938 if (sections++) info = sdscat(info,"\r\n");
1939 info = sdscatprintf(info,
1940 "# Replication\r\n"
1941 "role:%s\r\n",
1942 server.masterhost == NULL ? "master" : "slave");
1943 if (server.masterhost) {
1944 info = sdscatprintf(info,
1945 "master_host:%s\r\n"
1946 "master_port:%d\r\n"
1947 "master_link_status:%s\r\n"
1948 "master_last_io_seconds_ago:%d\r\n"
1949 "master_sync_in_progress:%d\r\n"
1950 ,server.masterhost,
1951 server.masterport,
1952 (server.repl_state == REDIS_REPL_CONNECTED) ?
1953 "up" : "down",
1954 server.master ?
1955 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
1956 server.repl_state == REDIS_REPL_TRANSFER
1957 );
1958
1959 if (server.repl_state == REDIS_REPL_TRANSFER) {
1960 info = sdscatprintf(info,
1961 "master_sync_left_bytes:%ld\r\n"
1962 "master_sync_last_io_seconds_ago:%d\r\n"
1963 ,(long)server.repl_transfer_left,
1964 (int)(server.unixtime-server.repl_transfer_lastio)
1965 );
1966 }
1967
1968 if (server.repl_state != REDIS_REPL_CONNECTED) {
1969 info = sdscatprintf(info,
1970 "master_link_down_since_seconds:%ld\r\n",
1971 (long)server.unixtime-server.repl_down_since);
1972 }
1973 }
1974 info = sdscatprintf(info,
1975 "connected_slaves:%lu\r\n",
1976 listLength(server.slaves));
1977 if (listLength(server.slaves)) {
1978 int slaveid = 0;
1979 listNode *ln;
1980 listIter li;
1981
1982 listRewind(server.slaves,&li);
1983 while((ln = listNext(&li))) {
1984 redisClient *slave = listNodeValue(ln);
1985 char *state = NULL;
1986 char ip[32];
1987 int port;
1988
1989 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
1990 switch(slave->replstate) {
1991 case REDIS_REPL_WAIT_BGSAVE_START:
1992 case REDIS_REPL_WAIT_BGSAVE_END:
1993 state = "wait_bgsave";
1994 break;
1995 case REDIS_REPL_SEND_BULK:
1996 state = "send_bulk";
1997 break;
1998 case REDIS_REPL_ONLINE:
1999 state = "online";
2000 break;
2001 }
2002 if (state == NULL) continue;
2003 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2004 slaveid,ip,port,state);
2005 slaveid++;
2006 }
2007 }
2008 }
2009
2010 /* CPU */
2011 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2012 if (sections++) info = sdscat(info,"\r\n");
2013 info = sdscatprintf(info,
2014 "# CPU\r\n"
2015 "used_cpu_sys:%.2f\r\n"
2016 "used_cpu_user:%.2f\r\n"
2017 "used_cpu_sys_children:%.2f\r\n"
2018 "used_cpu_user_children:%.2f\r\n",
2019 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2020 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2021 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2022 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2023 }
2024
2025 /* cmdtime */
2026 if (allsections || !strcasecmp(section,"commandstats")) {
2027 if (sections++) info = sdscat(info,"\r\n");
2028 info = sdscatprintf(info, "# Commandstats\r\n");
2029 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2030 for (j = 0; j < numcommands; j++) {
2031 struct redisCommand *c = redisCommandTable+j;
2032
2033 if (!c->calls) continue;
2034 info = sdscatprintf(info,
2035 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2036 c->name, c->calls, c->microseconds,
2037 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2038 }
2039 }
2040
2041 /* Cluster */
2042 if (allsections || defsections || !strcasecmp(section,"cluster")) {
2043 if (sections++) info = sdscat(info,"\r\n");
2044 info = sdscatprintf(info,
2045 "# Cluster\r\n"
2046 "cluster_enabled:%d\r\n",
2047 server.cluster_enabled);
2048 }
2049
2050 /* Key space */
2051 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2052 if (sections++) info = sdscat(info,"\r\n");
2053 info = sdscatprintf(info, "# Keyspace\r\n");
2054 for (j = 0; j < server.dbnum; j++) {
2055 long long keys, vkeys;
2056
2057 keys = dictSize(server.db[j].dict);
2058 vkeys = dictSize(server.db[j].expires);
2059 if (keys || vkeys) {
2060 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2061 j, keys, vkeys);
2062 }
2063 }
2064 }
2065 return info;
2066 }
2067
2068 void infoCommand(redisClient *c) {
2069 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2070
2071 if (c->argc > 2) {
2072 addReply(c,shared.syntaxerr);
2073 return;
2074 }
2075 sds info = genRedisInfoString(section);
2076 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2077 (unsigned long)sdslen(info)));
2078 addReplySds(c,info);
2079 addReply(c,shared.crlf);
2080 }
2081
2082 void monitorCommand(redisClient *c) {
2083 /* ignore MONITOR if aleady slave or in monitor mode */
2084 if (c->flags & REDIS_SLAVE) return;
2085
2086 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2087 c->slaveseldb = 0;
2088 listAddNodeTail(server.monitors,c);
2089 addReply(c,shared.ok);
2090 }
2091
2092 /* ============================ Maxmemory directive ======================== */
2093
2094 /* This function gets called when 'maxmemory' is set on the config file to limit
2095 * the max memory used by the server, before processing a command.
2096 *
2097 * The goal of the function is to free enough memory to keep Redis under the
2098 * configured memory limit.
2099 *
2100 * The function starts calculating how many bytes should be freed to keep
2101 * Redis under the limit, and enters a loop selecting the best keys to
2102 * evict accordingly to the configured policy.
2103 *
2104 * If all the bytes needed to return back under the limit were freed the
2105 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2106 * should block the execution of commands that will result in more memory
2107 * used by the server.
2108 */
2109 int freeMemoryIfNeeded(void) {
2110 size_t mem_used, mem_tofree, mem_freed;
2111 int slaves = listLength(server.slaves);
2112
2113 /* Remove the size of slaves output buffers and AOF buffer from the
2114 * count of used memory. */
2115 mem_used = zmalloc_used_memory();
2116 if (slaves) {
2117 listIter li;
2118 listNode *ln;
2119
2120 listRewind(server.slaves,&li);
2121 while((ln = listNext(&li))) {
2122 redisClient *slave = listNodeValue(ln);
2123 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2124 if (obuf_bytes > mem_used)
2125 mem_used = 0;
2126 else
2127 mem_used -= obuf_bytes;
2128 }
2129 }
2130 if (server.aof_state != REDIS_AOF_OFF) {
2131 mem_used -= sdslen(server.aof_buf);
2132 mem_used -= sdslen(server.aof_rewrite_buf);
2133 }
2134
2135 /* Check if we are over the memory limit. */
2136 if (mem_used <= server.maxmemory) return REDIS_OK;
2137
2138 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2139 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2140
2141 /* Compute how much memory we need to free. */
2142 mem_tofree = mem_used - server.maxmemory;
2143 mem_freed = 0;
2144 while (mem_freed < mem_tofree) {
2145 int j, k, keys_freed = 0;
2146
2147 for (j = 0; j < server.dbnum; j++) {
2148 long bestval = 0; /* just to prevent warning */
2149 sds bestkey = NULL;
2150 struct dictEntry *de;
2151 redisDb *db = server.db+j;
2152 dict *dict;
2153
2154 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2155 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2156 {
2157 dict = server.db[j].dict;
2158 } else {
2159 dict = server.db[j].expires;
2160 }
2161 if (dictSize(dict) == 0) continue;
2162
2163 /* volatile-random and allkeys-random policy */
2164 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2165 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2166 {
2167 de = dictGetRandomKey(dict);
2168 bestkey = dictGetKey(de);
2169 }
2170
2171 /* volatile-lru and allkeys-lru policy */
2172 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2173 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2174 {
2175 for (k = 0; k < server.maxmemory_samples; k++) {
2176 sds thiskey;
2177 long thisval;
2178 robj *o;
2179
2180 de = dictGetRandomKey(dict);
2181 thiskey = dictGetKey(de);
2182 /* When policy is volatile-lru we need an additonal lookup
2183 * to locate the real key, as dict is set to db->expires. */
2184 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2185 de = dictFind(db->dict, thiskey);
2186 o = dictGetVal(de);
2187 thisval = estimateObjectIdleTime(o);
2188
2189 /* Higher idle time is better candidate for deletion */
2190 if (bestkey == NULL || thisval > bestval) {
2191 bestkey = thiskey;
2192 bestval = thisval;
2193 }
2194 }
2195 }
2196
2197 /* volatile-ttl */
2198 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2199 for (k = 0; k < server.maxmemory_samples; k++) {
2200 sds thiskey;
2201 long thisval;
2202
2203 de = dictGetRandomKey(dict);
2204 thiskey = dictGetKey(de);
2205 thisval = (long) dictGetVal(de);
2206
2207 /* Expire sooner (minor expire unix timestamp) is better
2208 * candidate for deletion */
2209 if (bestkey == NULL || thisval < bestval) {
2210 bestkey = thiskey;
2211 bestval = thisval;
2212 }
2213 }
2214 }
2215
2216 /* Finally remove the selected key. */
2217 if (bestkey) {
2218 long long delta;
2219
2220 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2221 propagateExpire(db,keyobj);
2222 /* We compute the amount of memory freed by dbDelete() alone.
2223 * It is possible that actually the memory needed to propagate
2224 * the DEL in AOF and replication link is greater than the one
2225 * we are freeing removing the key, but we can't account for
2226 * that otherwise we would never exit the loop.
2227 *
2228 * AOF and Output buffer memory will be freed eventually so
2229 * we only care about memory used by the key space. */
2230 delta = (long long) zmalloc_used_memory();
2231 dbDelete(db,keyobj);
2232 delta -= (long long) zmalloc_used_memory();
2233 mem_freed += delta;
2234 server.stat_evictedkeys++;
2235 decrRefCount(keyobj);
2236 keys_freed++;
2237
2238 /* When the memory to free starts to be big enough, we may
2239 * start spending so much time here that is impossible to
2240 * deliver data to the slaves fast enough, so we force the
2241 * transmission here inside the loop. */
2242 if (slaves) flushSlavesOutputBuffers();
2243 }
2244 }
2245 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2246 }
2247 return REDIS_OK;
2248 }
2249
2250 /* =================================== Main! ================================ */
2251
2252 #ifdef __linux__
2253 int linuxOvercommitMemoryValue(void) {
2254 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2255 char buf[64];
2256
2257 if (!fp) return -1;
2258 if (fgets(buf,64,fp) == NULL) {
2259 fclose(fp);
2260 return -1;
2261 }
2262 fclose(fp);
2263
2264 return atoi(buf);
2265 }
2266
2267 void linuxOvercommitMemoryWarning(void) {
2268 if (linuxOvercommitMemoryValue() == 0) {
2269 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.");
2270 }
2271 }
2272 #endif /* __linux__ */
2273
2274 void createPidFile(void) {
2275 /* Try to write the pid file in a best-effort way. */
2276 FILE *fp = fopen(server.pidfile,"w");
2277 if (fp) {
2278 fprintf(fp,"%d\n",(int)getpid());
2279 fclose(fp);
2280 }
2281 }
2282
2283 void daemonize(void) {
2284 int fd;
2285
2286 if (fork() != 0) exit(0); /* parent exits */
2287 setsid(); /* create a new session */
2288
2289 /* Every output goes to /dev/null. If Redis is daemonized but
2290 * the 'logfile' is set to 'stdout' in the configuration file
2291 * it will not log at all. */
2292 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2293 dup2(fd, STDIN_FILENO);
2294 dup2(fd, STDOUT_FILENO);
2295 dup2(fd, STDERR_FILENO);
2296 if (fd > STDERR_FILENO) close(fd);
2297 }
2298 }
2299
2300 void version() {
2301 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2302 REDIS_VERSION,
2303 redisGitSHA1(),
2304 atoi(redisGitDirty()) > 0,
2305 ZMALLOC_LIB,
2306 sizeof(long) == 4 ? 32 : 64);
2307 exit(0);
2308 }
2309
2310 void usage() {
2311 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2312 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2313 fprintf(stderr," ./redis-server -v or --version\n");
2314 fprintf(stderr," ./redis-server -h or --help\n");
2315 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2316 fprintf(stderr,"Examples:\n");
2317 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2318 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2319 fprintf(stderr," ./redis-server --port 7777\n");
2320 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2321 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2322 exit(1);
2323 }
2324
2325 void redisAsciiArt(void) {
2326 #include "asciilogo.h"
2327 char *buf = zmalloc(1024*16);
2328
2329 snprintf(buf,1024*16,ascii_logo,
2330 REDIS_VERSION,
2331 redisGitSHA1(),
2332 strtol(redisGitDirty(),NULL,10) > 0,
2333 (sizeof(long) == 8) ? "64" : "32",
2334 server.cluster_enabled ? "cluster" : "stand alone",
2335 server.port,
2336 (long) getpid()
2337 );
2338 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2339 zfree(buf);
2340 }
2341
2342 static void sigtermHandler(int sig) {
2343 REDIS_NOTUSED(sig);
2344
2345 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2346 server.shutdown_asap = 1;
2347 }
2348
2349 void setupSignalHandlers(void) {
2350 struct sigaction act;
2351
2352 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2353 * Otherwise, sa_handler is used. */
2354 sigemptyset(&act.sa_mask);
2355 act.sa_flags = 0;
2356 act.sa_handler = sigtermHandler;
2357 sigaction(SIGTERM, &act, NULL);
2358
2359 #ifdef HAVE_BACKTRACE
2360 sigemptyset(&act.sa_mask);
2361 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2362 act.sa_sigaction = sigsegvHandler;
2363 sigaction(SIGSEGV, &act, NULL);
2364 sigaction(SIGBUS, &act, NULL);
2365 sigaction(SIGFPE, &act, NULL);
2366 sigaction(SIGILL, &act, NULL);
2367 #endif
2368 return;
2369 }
2370
2371 void memtest(size_t megabytes, int passes);
2372
2373 int main(int argc, char **argv) {
2374 long long start;
2375 struct timeval tv;
2376
2377 /* We need to initialize our libraries, and the server configuration. */
2378 zmalloc_enable_thread_safeness();
2379 srand(time(NULL)^getpid());
2380 gettimeofday(&tv,NULL);
2381 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2382 initServerConfig();
2383
2384 if (argc >= 2) {
2385 int j = 1; /* First option to parse in argv[] */
2386 sds options = sdsempty();
2387 char *configfile = NULL;
2388
2389 /* Handle special options --help and --version */
2390 if (strcmp(argv[1], "-v") == 0 ||
2391 strcmp(argv[1], "--version") == 0) version();
2392 if (strcmp(argv[1], "--help") == 0 ||
2393 strcmp(argv[1], "-h") == 0) usage();
2394 if (strcmp(argv[1], "--test-memory") == 0) {
2395 if (argc == 3) {
2396 memtest(atoi(argv[2]),50);
2397 exit(0);
2398 } else {
2399 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2400 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2401 exit(1);
2402 }
2403 }
2404
2405 /* First argument is the config file name? */
2406 if (argv[j][0] != '-' || argv[j][1] != '-')
2407 configfile = argv[j++];
2408 /* All the other options are parsed and conceptually appended to the
2409 * configuration file. For instance --port 6380 will generate the
2410 * string "port 6380\n" to be parsed after the actual file name
2411 * is parsed, if any. */
2412 while(j != argc) {
2413 if (argv[j][0] == '-' && argv[j][1] == '-') {
2414 /* Option name */
2415 if (sdslen(options)) options = sdscat(options,"\n");
2416 options = sdscat(options,argv[j]+2);
2417 options = sdscat(options," ");
2418 } else {
2419 /* Option argument */
2420 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2421 options = sdscat(options," ");
2422 }
2423 j++;
2424 }
2425 resetServerSaveParams();
2426 loadServerConfig(configfile,options);
2427 sdsfree(options);
2428 } else {
2429 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'");
2430 }
2431 if (server.daemonize) daemonize();
2432 initServer();
2433 if (server.daemonize) createPidFile();
2434 redisAsciiArt();
2435 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2436 #ifdef __linux__
2437 linuxOvercommitMemoryWarning();
2438 #endif
2439 start = ustime();
2440 if (server.aof_state == REDIS_AOF_ON) {
2441 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2442 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2443 } else {
2444 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2445 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2446 (float)(ustime()-start)/1000000);
2447 } else if (errno != ENOENT) {
2448 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2449 exit(1);
2450 }
2451 }
2452 if (server.ipfd > 0)
2453 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2454 if (server.sofd > 0)
2455 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2456 aeSetBeforeSleepProc(server.el,beforeSleep);
2457 aeMain(server.el);
2458 aeDeleteEventLoop(server.el);
2459 return 0;
2460 }
2461
2462 /* The End */