]> git.saurik.com Git - redis.git/blob - src/redis.c
cfe060360660dcba092db7de1cbed03da5349477
[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 /* Keys dictionary */
598 if (dictIsRehashing(server.db[j].dict)) {
599 dictRehashMilliseconds(server.db[j].dict,1);
600 break; /* already used our millisecond for this loop... */
601 }
602 /* Expires */
603 if (dictIsRehashing(server.db[j].expires)) {
604 dictRehashMilliseconds(server.db[j].expires,1);
605 break; /* already used our millisecond for this loop... */
606 }
607 }
608 }
609
610 /* This function is called once a background process of some kind terminates,
611 * as we want to avoid resizing the hash tables when there is a child in order
612 * to play well with copy-on-write (otherwise when a resize happens lots of
613 * memory pages are copied). The goal of this function is to update the ability
614 * for dict.c to resize the hash tables accordingly to the fact we have o not
615 * running childs. */
616 void updateDictResizePolicy(void) {
617 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
618 dictEnableResize();
619 else
620 dictDisableResize();
621 }
622
623 /* ======================= Cron: called every 100 ms ======================== */
624
625 /* Try to expire a few timed out keys. The algorithm used is adaptive and
626 * will use few CPU cycles if there are few expiring keys, otherwise
627 * it will get more aggressive to avoid that too much memory is used by
628 * keys that can be removed from the keyspace. */
629 void activeExpireCycle(void) {
630 int j, iteration = 0;
631 long long start = ustime(), timelimit;
632
633 /* We can use at max REDIS_EXPIRELOOKUPS_TIME_PERC percentage of CPU time
634 * per iteration. Since this function gets called with a frequency of
635 * REDIS_HZ times per second, the following is the max amount of
636 * microseconds we can spend in this function. */
637 timelimit = 1000000*REDIS_EXPIRELOOKUPS_TIME_PERC/REDIS_HZ/100;
638 if (timelimit <= 0) timelimit = 1;
639
640 for (j = 0; j < server.dbnum; j++) {
641 int expired;
642 redisDb *db = server.db+j;
643
644 /* Continue to expire if at the end of the cycle more than 25%
645 * of the keys were expired. */
646 do {
647 unsigned long num = dictSize(db->expires);
648 unsigned long slots = dictSlots(db->expires);
649 long long now = mstime();
650
651 /* When there are less than 1% filled slots getting random
652 * keys is expensive, so stop here waiting for better times...
653 * The dictionary will be resized asap. */
654 if (num && slots > DICT_HT_INITIAL_SIZE &&
655 (num*100/slots < 1)) break;
656
657 /* The main collection cycle. Sample random keys among keys
658 * with an expire set, checking for expired ones. */
659 expired = 0;
660 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
661 num = REDIS_EXPIRELOOKUPS_PER_CRON;
662 while (num--) {
663 dictEntry *de;
664 long long t;
665
666 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
667 t = dictGetSignedIntegerVal(de);
668 if (now > t) {
669 sds key = dictGetKey(de);
670 robj *keyobj = createStringObject(key,sdslen(key));
671
672 propagateExpire(db,keyobj);
673 dbDelete(db,keyobj);
674 decrRefCount(keyobj);
675 expired++;
676 server.stat_expiredkeys++;
677 }
678 }
679 /* We can't block forever here even if there are many keys to
680 * expire. So after a given amount of milliseconds return to the
681 * caller waiting for the other active expire cycle. */
682 iteration++;
683 if ((iteration & 0xf) == 0 && /* check once every 16 cycles. */
684 (ustime()-start) > timelimit) return;
685 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
686 }
687 }
688
689 void updateLRUClock(void) {
690 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
691 REDIS_LRU_CLOCK_MAX;
692 }
693
694
695 /* Add a sample to the operations per second array of samples. */
696 void trackOperationsPerSecond(void) {
697 long long t = mstime() - server.ops_sec_last_sample_time;
698 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
699 long long ops_sec;
700
701 ops_sec = t > 0 ? (ops*1000/t) : 0;
702
703 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
704 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
705 server.ops_sec_last_sample_time = mstime();
706 server.ops_sec_last_sample_ops = server.stat_numcommands;
707 }
708
709 /* Return the mean of all the samples. */
710 long long getOperationsPerSecond(void) {
711 int j;
712 long long sum = 0;
713
714 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
715 sum += server.ops_sec_samples[j];
716 return sum / REDIS_OPS_SEC_SAMPLES;
717 }
718
719 /* Check for timeouts. Returns non-zero if the client was terminated */
720 int clientsCronHandleTimeout(redisClient *c) {
721 time_t now = server.unixtime;
722
723 if (server.maxidletime &&
724 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
725 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
726 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
727 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
728 listLength(c->pubsub_patterns) == 0 &&
729 (now - c->lastinteraction > server.maxidletime))
730 {
731 redisLog(REDIS_VERBOSE,"Closing idle client");
732 freeClient(c);
733 return 1;
734 } else if (c->flags & REDIS_BLOCKED) {
735 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
736 addReply(c,shared.nullmultibulk);
737 unblockClientWaitingData(c);
738 }
739 }
740 return 0;
741 }
742
743 /* The client query buffer is an sds.c string that can end with a lot of
744 * free space not used, this function reclaims space if needed.
745 *
746 * The funciton always returns 0 as it never terminates the client. */
747 int clientsCronResizeQueryBuffer(redisClient *c) {
748 size_t querybuf_size = sdsAllocSize(c->querybuf);
749 time_t idletime = server.unixtime - c->lastinteraction;
750
751 /* There are two conditions to resize the query buffer:
752 * 1) Query buffer is > BIG_ARG and too big for latest peak.
753 * 2) Client is inactive and the buffer is bigger than 1k. */
754 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
755 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
756 (querybuf_size > 1024 && idletime > 2))
757 {
758 /* Only resize the query buffer if it is actually wasting space. */
759 if (sdsavail(c->querybuf) > 1024) {
760 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
761 }
762 }
763 /* Reset the peak again to capture the peak memory usage in the next
764 * cycle. */
765 c->querybuf_peak = 0;
766 return 0;
767 }
768
769 void clientsCron(void) {
770 /* Make sure to process at least 1/(REDIS_HZ*10) of clients per call.
771 * Since this function is called REDIS_HZ times per second we are sure that
772 * in the worst case we process all the clients in 10 seconds.
773 * In normal conditions (a reasonable number of clients) we process
774 * all the clients in a shorter time. */
775 int numclients = listLength(server.clients);
776 int iterations = numclients/(REDIS_HZ*10);
777
778 if (iterations < 50)
779 iterations = (numclients < 50) ? numclients : 50;
780 while(listLength(server.clients) && iterations--) {
781 redisClient *c;
782 listNode *head;
783
784 /* Rotate the list, take the current head, process.
785 * This way if the client must be removed from the list it's the
786 * first element and we don't incur into O(N) computation. */
787 listRotate(server.clients);
788 head = listFirst(server.clients);
789 c = listNodeValue(head);
790 /* The following functions do different service checks on the client.
791 * The protocol is that they return non-zero if the client was
792 * terminated. */
793 if (clientsCronHandleTimeout(c)) continue;
794 if (clientsCronResizeQueryBuffer(c)) continue;
795 }
796 }
797
798 /* This is our timer interrupt, called REDIS_HZ times per second.
799 * Here is where we do a number of things that need to be done asynchronously.
800 * For instance:
801 *
802 * - Active expired keys collection (it is also performed in a lazy way on
803 * lookup).
804 * - Software watchdong.
805 * - Update some statistic.
806 * - Incremental rehashing of the DBs hash tables.
807 * - Triggering BGSAVE / AOF rewrite, and handling of terminated children.
808 * - Clients timeout of differnet kinds.
809 * - Replication reconnection.
810 * - Many more...
811 *
812 * Everything directly called here will be called REDIS_HZ times per second,
813 * so in order to throttle execution of things we want to do less frequently
814 * a macro is used: run_with_period(milliseconds) { .... }
815 */
816
817 /* Using the following macro you can run code inside serverCron() with the
818 * specified period, specified in milliseconds.
819 * The actual resolution depends on REDIS_HZ. */
820 #define run_with_period(_ms_) if (!(loops % ((_ms_)/(1000/REDIS_HZ))))
821
822 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
823 int j, loops = server.cronloops;
824 REDIS_NOTUSED(eventLoop);
825 REDIS_NOTUSED(id);
826 REDIS_NOTUSED(clientData);
827
828 /* Software watchdog: deliver the SIGALRM that will reach the signal
829 * handler if we don't return here fast enough. */
830 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
831
832 /* We take a cached value of the unix time in the global state because
833 * with virtual memory and aging there is to store the current time
834 * in objects at every object access, and accuracy is not needed.
835 * To access a global var is faster than calling time(NULL) */
836 server.unixtime = time(NULL);
837
838 run_with_period(100) trackOperationsPerSecond();
839
840 /* We have just 22 bits per object for LRU information.
841 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
842 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
843 *
844 * Note that even if this will wrap after 1.5 years it's not a problem,
845 * everything will still work but just some object will appear younger
846 * to Redis. But for this to happen a given object should never be touched
847 * for 1.5 years.
848 *
849 * Note that you can change the resolution altering the
850 * REDIS_LRU_CLOCK_RESOLUTION define.
851 */
852 updateLRUClock();
853
854 /* Record the max memory used since the server was started. */
855 if (zmalloc_used_memory() > server.stat_peak_memory)
856 server.stat_peak_memory = zmalloc_used_memory();
857
858 /* We received a SIGTERM, shutting down here in a safe way, as it is
859 * not ok doing so inside the signal handler. */
860 if (server.shutdown_asap) {
861 if (prepareForShutdown(0) == REDIS_OK) exit(0);
862 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
863 }
864
865 /* Show some info about non-empty databases */
866 run_with_period(5000) {
867 for (j = 0; j < server.dbnum; j++) {
868 long long size, used, vkeys;
869
870 size = dictSlots(server.db[j].dict);
871 used = dictSize(server.db[j].dict);
872 vkeys = dictSize(server.db[j].expires);
873 if (used || vkeys) {
874 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
875 /* dictPrintStats(server.dict); */
876 }
877 }
878 }
879
880 /* We don't want to resize the hash tables while a bacground saving
881 * is in progress: the saving child is created using fork() that is
882 * implemented with a copy-on-write semantic in most modern systems, so
883 * if we resize the HT while there is the saving child at work actually
884 * a lot of memory movements in the parent will cause a lot of pages
885 * copied. */
886 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
887 tryResizeHashTables();
888 if (server.activerehashing) incrementallyRehash();
889 }
890
891 /* Show information about connected clients */
892 run_with_period(5000) {
893 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
894 listLength(server.clients)-listLength(server.slaves),
895 listLength(server.slaves),
896 zmalloc_used_memory());
897 }
898
899 /* We need to do a few operations on clients asynchronously. */
900 clientsCron();
901
902 /* Start a scheduled AOF rewrite if this was requested by the user while
903 * a BGSAVE was in progress. */
904 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
905 server.aof_rewrite_scheduled)
906 {
907 rewriteAppendOnlyFileBackground();
908 }
909
910 /* Check if a background saving or AOF rewrite in progress terminated. */
911 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
912 int statloc;
913 pid_t pid;
914
915 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
916 int exitcode = WEXITSTATUS(statloc);
917 int bysignal = 0;
918
919 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
920
921 if (pid == server.rdb_child_pid) {
922 backgroundSaveDoneHandler(exitcode,bysignal);
923 } else {
924 backgroundRewriteDoneHandler(exitcode,bysignal);
925 }
926 updateDictResizePolicy();
927 }
928 } else {
929 /* If there is not a background saving/rewrite in progress check if
930 * we have to save/rewrite now */
931 for (j = 0; j < server.saveparamslen; j++) {
932 struct saveparam *sp = server.saveparams+j;
933
934 if (server.dirty >= sp->changes &&
935 server.unixtime-server.lastsave > sp->seconds) {
936 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
937 sp->changes, sp->seconds);
938 rdbSaveBackground(server.rdb_filename);
939 break;
940 }
941 }
942
943 /* Trigger an AOF rewrite if needed */
944 if (server.rdb_child_pid == -1 &&
945 server.aof_child_pid == -1 &&
946 server.aof_rewrite_perc &&
947 server.aof_current_size > server.aof_rewrite_min_size)
948 {
949 long long base = server.aof_rewrite_base_size ?
950 server.aof_rewrite_base_size : 1;
951 long long growth = (server.aof_current_size*100/base) - 100;
952 if (growth >= server.aof_rewrite_perc) {
953 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
954 rewriteAppendOnlyFileBackground();
955 }
956 }
957 }
958
959
960 /* If we postponed an AOF buffer flush, let's try to do it every time the
961 * cron function is called. */
962 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
963
964 /* Expire a few keys per cycle, only if this is a master.
965 * On slaves we wait for DEL operations synthesized by the master
966 * in order to guarantee a strict consistency. */
967 if (server.masterhost == NULL) activeExpireCycle();
968
969 /* Close clients that need to be closed asynchronous */
970 freeClientsInAsyncFreeQueue();
971
972 /* Replication cron function -- used to reconnect to master and
973 * to detect transfer failures. */
974 run_with_period(1000) replicationCron();
975
976 /* Run other sub-systems specific cron jobs */
977 run_with_period(1000) {
978 if (server.cluster_enabled) clusterCron();
979 }
980
981 server.cronloops++;
982 return 1000/REDIS_HZ;
983 }
984
985 /* This function gets called every time Redis is entering the
986 * main loop of the event driven library, that is, before to sleep
987 * for ready file descriptors. */
988 void beforeSleep(struct aeEventLoop *eventLoop) {
989 REDIS_NOTUSED(eventLoop);
990 listNode *ln;
991 redisClient *c;
992
993 /* Try to process pending commands for clients that were just unblocked. */
994 while (listLength(server.unblocked_clients)) {
995 ln = listFirst(server.unblocked_clients);
996 redisAssert(ln != NULL);
997 c = ln->value;
998 listDelNode(server.unblocked_clients,ln);
999 c->flags &= ~REDIS_UNBLOCKED;
1000
1001 /* Process remaining data in the input buffer. */
1002 if (c->querybuf && sdslen(c->querybuf) > 0) {
1003 server.current_client = c;
1004 processInputBuffer(c);
1005 server.current_client = NULL;
1006 }
1007 }
1008
1009 /* Write the AOF buffer on disk */
1010 flushAppendOnlyFile(0);
1011 }
1012
1013 /* =========================== Server initialization ======================== */
1014
1015 void createSharedObjects(void) {
1016 int j;
1017
1018 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
1019 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
1020 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
1021 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
1022 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
1023 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
1024 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
1025 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
1026 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
1027 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
1028 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
1029 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
1030 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
1031 "-ERR Operation against a key holding the wrong kind of value\r\n"));
1032 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
1033 "-ERR no such key\r\n"));
1034 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
1035 "-ERR syntax error\r\n"));
1036 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
1037 "-ERR source and destination objects are the same\r\n"));
1038 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
1039 "-ERR index out of range\r\n"));
1040 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
1041 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
1042 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
1043 "-LOADING Redis is loading the dataset in memory\r\n"));
1044 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
1045 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
1046 shared.masterdownerr = createObject(REDIS_STRING,sdsnew(
1047 "-MASTERDOWN Link with MASTER is down and slave-serve-stale-data is set to 'no'.\r\n"));
1048 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
1049 "-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"));
1050 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
1051 "-READONLY You can't write against a read only slave.\r\n"));
1052 shared.oomerr = createObject(REDIS_STRING,sdsnew(
1053 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
1054 shared.space = createObject(REDIS_STRING,sdsnew(" "));
1055 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1056 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1057
1058 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1059 shared.select[j] = createObject(REDIS_STRING,
1060 sdscatprintf(sdsempty(),"select %d\r\n", j));
1061 }
1062 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1063 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1064 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1065 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1066 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1067 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1068 shared.del = createStringObject("DEL",3);
1069 shared.rpop = createStringObject("RPOP",4);
1070 shared.lpop = createStringObject("LPOP",4);
1071 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1072 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1073 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1074 }
1075 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1076 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1077 sdscatprintf(sdsempty(),"*%d\r\n",j));
1078 shared.bulkhdr[j] = createObject(REDIS_STRING,
1079 sdscatprintf(sdsempty(),"$%d\r\n",j));
1080 }
1081 }
1082
1083 void initServerConfig() {
1084 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1085 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1086 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1087 server.port = REDIS_SERVERPORT;
1088 server.bindaddr = NULL;
1089 server.unixsocket = NULL;
1090 server.unixsocketperm = 0;
1091 server.ipfd = -1;
1092 server.sofd = -1;
1093 server.dbnum = REDIS_DEFAULT_DBNUM;
1094 server.verbosity = REDIS_NOTICE;
1095 server.maxidletime = REDIS_MAXIDLETIME;
1096 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1097 server.saveparams = NULL;
1098 server.loading = 0;
1099 server.logfile = NULL; /* NULL = log on standard output */
1100 server.syslog_enabled = 0;
1101 server.syslog_ident = zstrdup("redis");
1102 server.syslog_facility = LOG_LOCAL0;
1103 server.daemonize = 0;
1104 server.aof_state = REDIS_AOF_OFF;
1105 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1106 server.aof_no_fsync_on_rewrite = 0;
1107 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1108 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1109 server.aof_rewrite_base_size = 0;
1110 server.aof_rewrite_scheduled = 0;
1111 server.aof_last_fsync = time(NULL);
1112 server.aof_delayed_fsync = 0;
1113 server.aof_fd = -1;
1114 server.aof_selected_db = -1; /* Make sure the first time will not match */
1115 server.aof_flush_postponed_start = 0;
1116 server.pidfile = zstrdup("/var/run/redis.pid");
1117 server.rdb_filename = zstrdup("dump.rdb");
1118 server.aof_filename = zstrdup("appendonly.aof");
1119 server.requirepass = NULL;
1120 server.rdb_compression = 1;
1121 server.rdb_checksum = 1;
1122 server.activerehashing = 1;
1123 server.maxclients = REDIS_MAX_CLIENTS;
1124 server.bpop_blocked_clients = 0;
1125 server.maxmemory = 0;
1126 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1127 server.maxmemory_samples = 3;
1128 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1129 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1130 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1131 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1132 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1133 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1134 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1135 server.shutdown_asap = 0;
1136 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1137 server.repl_timeout = REDIS_REPL_TIMEOUT;
1138 server.cluster_enabled = 0;
1139 server.cluster.configfile = zstrdup("nodes.conf");
1140 server.lua_caller = NULL;
1141 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1142 server.lua_client = NULL;
1143 server.lua_timedout = 0;
1144
1145 updateLRUClock();
1146 resetServerSaveParams();
1147
1148 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1149 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1150 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1151 /* Replication related */
1152 server.masterauth = NULL;
1153 server.masterhost = NULL;
1154 server.masterport = 6379;
1155 server.master = NULL;
1156 server.repl_state = REDIS_REPL_NONE;
1157 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1158 server.repl_serve_stale_data = 1;
1159 server.repl_slave_ro = 1;
1160 server.repl_down_since = time(NULL);
1161
1162 /* Client output buffer limits */
1163 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1164 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1165 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1166 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1167 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1168 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1169 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1170 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1171 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1172
1173 /* Double constants initialization */
1174 R_Zero = 0.0;
1175 R_PosInf = 1.0/R_Zero;
1176 R_NegInf = -1.0/R_Zero;
1177 R_Nan = R_Zero/R_Zero;
1178
1179 /* Command table -- we intiialize it here as it is part of the
1180 * initial configuration, since command names may be changed via
1181 * redis.conf using the rename-command directive. */
1182 server.commands = dictCreate(&commandTableDictType,NULL);
1183 populateCommandTable();
1184 server.delCommand = lookupCommandByCString("del");
1185 server.multiCommand = lookupCommandByCString("multi");
1186 server.lpushCommand = lookupCommandByCString("lpush");
1187
1188 /* Slow log */
1189 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1190 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1191
1192 /* Debugging */
1193 server.assert_failed = "<no assertion failed>";
1194 server.assert_file = "<no file>";
1195 server.assert_line = 0;
1196 server.bug_report_start = 0;
1197 server.watchdog_period = 0;
1198 }
1199
1200 /* This function will try to raise the max number of open files accordingly to
1201 * the configured max number of clients. It will also account for 32 additional
1202 * file descriptors as we need a few more for persistence, listening
1203 * sockets, log files and so forth.
1204 *
1205 * If it will not be possible to set the limit accordingly to the configured
1206 * max number of clients, the function will do the reverse setting
1207 * server.maxclients to the value that we can actually handle. */
1208 void adjustOpenFilesLimit(void) {
1209 rlim_t maxfiles = server.maxclients+32;
1210 struct rlimit limit;
1211
1212 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1213 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1214 strerror(errno));
1215 server.maxclients = 1024-32;
1216 } else {
1217 rlim_t oldlimit = limit.rlim_cur;
1218
1219 /* Set the max number of files if the current limit is not enough
1220 * for our needs. */
1221 if (oldlimit < maxfiles) {
1222 rlim_t f;
1223
1224 f = maxfiles;
1225 while(f > oldlimit) {
1226 limit.rlim_cur = f;
1227 limit.rlim_max = f;
1228 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1229 f -= 128;
1230 }
1231 if (f < oldlimit) f = oldlimit;
1232 if (f != maxfiles) {
1233 server.maxclients = f-32;
1234 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1235 (int) maxfiles, strerror(errno), (int) server.maxclients);
1236 } else {
1237 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1238 (int) maxfiles);
1239 }
1240 }
1241 }
1242 }
1243
1244 void initServer() {
1245 int j;
1246
1247 signal(SIGHUP, SIG_IGN);
1248 signal(SIGPIPE, SIG_IGN);
1249 setupSignalHandlers();
1250
1251 if (server.syslog_enabled) {
1252 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1253 server.syslog_facility);
1254 }
1255
1256 server.current_client = NULL;
1257 server.clients = listCreate();
1258 server.clients_to_close = listCreate();
1259 server.slaves = listCreate();
1260 server.monitors = listCreate();
1261 server.unblocked_clients = listCreate();
1262
1263 createSharedObjects();
1264 adjustOpenFilesLimit();
1265 server.el = aeCreateEventLoop(server.maxclients+1024);
1266 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1267
1268 if (server.port != 0) {
1269 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1270 if (server.ipfd == ANET_ERR) {
1271 redisLog(REDIS_WARNING, "Opening port %d: %s",
1272 server.port, server.neterr);
1273 exit(1);
1274 }
1275 }
1276 if (server.unixsocket != NULL) {
1277 unlink(server.unixsocket); /* don't care if this fails */
1278 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1279 if (server.sofd == ANET_ERR) {
1280 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1281 exit(1);
1282 }
1283 }
1284 if (server.ipfd < 0 && server.sofd < 0) {
1285 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1286 exit(1);
1287 }
1288 for (j = 0; j < server.dbnum; j++) {
1289 server.db[j].dict = dictCreate(&dbDictType,NULL);
1290 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1291 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1292 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1293 server.db[j].id = j;
1294 }
1295 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1296 server.pubsub_patterns = listCreate();
1297 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1298 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1299 server.cronloops = 0;
1300 server.rdb_child_pid = -1;
1301 server.aof_child_pid = -1;
1302 aofRewriteBufferReset();
1303 server.aof_buf = sdsempty();
1304 server.lastsave = time(NULL);
1305 server.dirty = 0;
1306 server.stat_numcommands = 0;
1307 server.stat_numconnections = 0;
1308 server.stat_expiredkeys = 0;
1309 server.stat_evictedkeys = 0;
1310 server.stat_starttime = time(NULL);
1311 server.stat_keyspace_misses = 0;
1312 server.stat_keyspace_hits = 0;
1313 server.stat_peak_memory = 0;
1314 server.stat_fork_time = 0;
1315 server.stat_rejected_conn = 0;
1316 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1317 server.ops_sec_idx = 0;
1318 server.ops_sec_last_sample_time = mstime();
1319 server.ops_sec_last_sample_ops = 0;
1320 server.unixtime = time(NULL);
1321 server.lastbgsave_status = REDIS_OK;
1322 server.stop_writes_on_bgsave_err = 1;
1323 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1324 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1325 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1326 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1327 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1328
1329 if (server.aof_state == REDIS_AOF_ON) {
1330 server.aof_fd = open(server.aof_filename,
1331 O_WRONLY|O_APPEND|O_CREAT,0644);
1332 if (server.aof_fd == -1) {
1333 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1334 strerror(errno));
1335 exit(1);
1336 }
1337 }
1338
1339 /* 32 bit instances are limited to 4GB of address space, so if there is
1340 * no explicit limit in the user provided configuration we set a limit
1341 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1342 * useless crashes of the Redis instance. */
1343 if (server.arch_bits == 32 && server.maxmemory == 0) {
1344 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1345 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1346 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1347 }
1348
1349 if (server.cluster_enabled) clusterInit();
1350 scriptingInit();
1351 slowlogInit();
1352 bioInit();
1353 }
1354
1355 /* Populates the Redis Command Table starting from the hard coded list
1356 * we have on top of redis.c file. */
1357 void populateCommandTable(void) {
1358 int j;
1359 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1360
1361 for (j = 0; j < numcommands; j++) {
1362 struct redisCommand *c = redisCommandTable+j;
1363 char *f = c->sflags;
1364 int retval;
1365
1366 while(*f != '\0') {
1367 switch(*f) {
1368 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1369 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1370 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1371 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1372 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1373 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1374 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1375 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1376 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1377 default: redisPanic("Unsupported command flag"); break;
1378 }
1379 f++;
1380 }
1381
1382 retval = dictAdd(server.commands, sdsnew(c->name), c);
1383 assert(retval == DICT_OK);
1384 }
1385 }
1386
1387 void resetCommandTableStats(void) {
1388 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1389 int j;
1390
1391 for (j = 0; j < numcommands; j++) {
1392 struct redisCommand *c = redisCommandTable+j;
1393
1394 c->microseconds = 0;
1395 c->calls = 0;
1396 }
1397 }
1398
1399 /* ========================== Redis OP Array API ============================ */
1400
1401 void redisOpArrayInit(redisOpArray *oa) {
1402 oa->ops = NULL;
1403 oa->numops = 0;
1404 }
1405
1406 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1407 robj **argv, int argc, int target)
1408 {
1409 redisOp *op;
1410
1411 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1412 op = oa->ops+oa->numops;
1413 op->cmd = cmd;
1414 op->dbid = dbid;
1415 op->argv = argv;
1416 op->argc = argc;
1417 op->target = target;
1418 oa->numops++;
1419 return oa->numops;
1420 }
1421
1422 void redisOpArrayFree(redisOpArray *oa) {
1423 while(oa->numops) {
1424 int j;
1425 redisOp *op;
1426
1427 oa->numops--;
1428 op = oa->ops+oa->numops;
1429 for (j = 0; j < op->argc; j++)
1430 decrRefCount(op->argv[j]);
1431 zfree(op->argv);
1432 }
1433 zfree(oa->ops);
1434 }
1435
1436 /* ====================== Commands lookup and execution ===================== */
1437
1438 struct redisCommand *lookupCommand(sds name) {
1439 return dictFetchValue(server.commands, name);
1440 }
1441
1442 struct redisCommand *lookupCommandByCString(char *s) {
1443 struct redisCommand *cmd;
1444 sds name = sdsnew(s);
1445
1446 cmd = dictFetchValue(server.commands, name);
1447 sdsfree(name);
1448 return cmd;
1449 }
1450
1451 /* Propagate the specified command (in the context of the specified database id)
1452 * to AOF, Slaves and Monitors.
1453 *
1454 * flags are an xor between:
1455 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1456 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1457 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1458 */
1459 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1460 int flags)
1461 {
1462 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1463 feedAppendOnlyFile(cmd,dbid,argv,argc);
1464 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1465 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1466 }
1467
1468 /* Used inside commands to schedule the propagation of additional commands
1469 * after the current command is propagated to AOF / Replication. */
1470 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1471 int target)
1472 {
1473 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1474 }
1475
1476 /* Call() is the core of Redis execution of a command */
1477 void call(redisClient *c, int flags) {
1478 long long dirty, start = ustime(), duration;
1479
1480 /* Sent the command to clients in MONITOR mode, only if the commands are
1481 * not geneated from reading an AOF. */
1482 if (listLength(server.monitors) && !server.loading)
1483 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1484
1485 /* Call the command. */
1486 redisOpArrayInit(&server.also_propagate);
1487 dirty = server.dirty;
1488 c->cmd->proc(c);
1489 dirty = server.dirty-dirty;
1490 duration = ustime()-start;
1491
1492 /* When EVAL is called loading the AOF we don't want commands called
1493 * from Lua to go into the slowlog or to populate statistics. */
1494 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1495 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1496
1497 /* Log the command into the Slow log if needed, and populate the
1498 * per-command statistics that we show in INFO commandstats. */
1499 if (flags & REDIS_CALL_SLOWLOG)
1500 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1501 if (flags & REDIS_CALL_STATS) {
1502 c->cmd->microseconds += duration;
1503 c->cmd->calls++;
1504 }
1505
1506 /* Propagate the command into the AOF and replication link */
1507 if (flags & REDIS_CALL_PROPAGATE) {
1508 int flags = REDIS_PROPAGATE_NONE;
1509
1510 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1511 flags |= REDIS_PROPAGATE_REPL;
1512 if (dirty)
1513 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1514 if (flags != REDIS_PROPAGATE_NONE)
1515 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1516 }
1517 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1518 * PUSH command. */
1519 if (server.also_propagate.numops) {
1520 int j;
1521 redisOp *rop;
1522
1523 for (j = 0; j < server.also_propagate.numops; j++) {
1524 rop = &server.also_propagate.ops[j];
1525 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1526 }
1527 redisOpArrayFree(&server.also_propagate);
1528 }
1529 server.stat_numcommands++;
1530 }
1531
1532 /* If this function gets called we already read a whole
1533 * command, argments are in the client argv/argc fields.
1534 * processCommand() execute the command or prepare the
1535 * server for a bulk read from the client.
1536 *
1537 * If 1 is returned the client is still alive and valid and
1538 * and other operations can be performed by the caller. Otherwise
1539 * if 0 is returned the client was destroied (i.e. after QUIT). */
1540 int processCommand(redisClient *c) {
1541 /* The QUIT command is handled separately. Normal command procs will
1542 * go through checking for replication and QUIT will cause trouble
1543 * when FORCE_REPLICATION is enabled and would be implemented in
1544 * a regular command proc. */
1545 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1546 addReply(c,shared.ok);
1547 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1548 return REDIS_ERR;
1549 }
1550
1551 /* Now lookup the command and check ASAP about trivial error conditions
1552 * such as wrong arity, bad command name and so forth. */
1553 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1554 if (!c->cmd) {
1555 addReplyErrorFormat(c,"unknown command '%s'",
1556 (char*)c->argv[0]->ptr);
1557 return REDIS_OK;
1558 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1559 (c->argc < -c->cmd->arity)) {
1560 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1561 c->cmd->name);
1562 return REDIS_OK;
1563 }
1564
1565 /* Check if the user is authenticated */
1566 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1567 {
1568 addReplyError(c,"operation not permitted");
1569 return REDIS_OK;
1570 }
1571
1572 /* If cluster is enabled, redirect here */
1573 if (server.cluster_enabled &&
1574 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1575 int hashslot;
1576
1577 if (server.cluster.state != REDIS_CLUSTER_OK) {
1578 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1579 return REDIS_OK;
1580 } else {
1581 int ask;
1582 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1583 if (n == NULL) {
1584 addReplyError(c,"Multi keys request invalid in cluster");
1585 return REDIS_OK;
1586 } else if (n != server.cluster.myself) {
1587 addReplySds(c,sdscatprintf(sdsempty(),
1588 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1589 hashslot,n->ip,n->port));
1590 return REDIS_OK;
1591 }
1592 }
1593 }
1594
1595 /* Handle the maxmemory directive.
1596 *
1597 * First we try to free some memory if possible (if there are volatile
1598 * keys in the dataset). If there are not the only thing we can do
1599 * is returning an error. */
1600 if (server.maxmemory) {
1601 int retval = freeMemoryIfNeeded();
1602 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1603 addReply(c, shared.oomerr);
1604 return REDIS_OK;
1605 }
1606 }
1607
1608 /* Don't accept write commands if there are problems persisting on disk. */
1609 if (server.stop_writes_on_bgsave_err &&
1610 server.saveparamslen > 0
1611 && server.lastbgsave_status == REDIS_ERR &&
1612 c->cmd->flags & REDIS_CMD_WRITE)
1613 {
1614 addReply(c, shared.bgsaveerr);
1615 return REDIS_OK;
1616 }
1617
1618 /* Don't accept wirte commands if this is a read only slave. But
1619 * accept write commands if this is our master. */
1620 if (server.masterhost && server.repl_slave_ro &&
1621 !(c->flags & REDIS_MASTER) &&
1622 c->cmd->flags & REDIS_CMD_WRITE)
1623 {
1624 addReply(c, shared.roslaveerr);
1625 return REDIS_OK;
1626 }
1627
1628 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1629 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1630 &&
1631 c->cmd->proc != subscribeCommand &&
1632 c->cmd->proc != unsubscribeCommand &&
1633 c->cmd->proc != psubscribeCommand &&
1634 c->cmd->proc != punsubscribeCommand) {
1635 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1636 return REDIS_OK;
1637 }
1638
1639 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1640 * we are a slave with a broken link with master. */
1641 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1642 server.repl_serve_stale_data == 0 &&
1643 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1644 {
1645 addReply(c, shared.masterdownerr);
1646 return REDIS_OK;
1647 }
1648
1649 /* Loading DB? Return an error if the command is not INFO */
1650 if (server.loading && c->cmd->proc != infoCommand) {
1651 addReply(c, shared.loadingerr);
1652 return REDIS_OK;
1653 }
1654
1655 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1656 if (server.lua_timedout &&
1657 !(c->cmd->proc == shutdownCommand &&
1658 c->argc == 2 &&
1659 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1660 !(c->cmd->proc == scriptCommand &&
1661 c->argc == 2 &&
1662 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1663 {
1664 addReply(c, shared.slowscripterr);
1665 return REDIS_OK;
1666 }
1667
1668 /* Exec the command */
1669 if (c->flags & REDIS_MULTI &&
1670 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1671 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1672 {
1673 queueMultiCommand(c);
1674 addReply(c,shared.queued);
1675 } else {
1676 call(c,REDIS_CALL_FULL);
1677 }
1678 return REDIS_OK;
1679 }
1680
1681 /*================================== Shutdown =============================== */
1682
1683 int prepareForShutdown(int flags) {
1684 int save = flags & REDIS_SHUTDOWN_SAVE;
1685 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1686
1687 redisLog(REDIS_WARNING,"User requested shutdown...");
1688 /* Kill the saving child if there is a background saving in progress.
1689 We want to avoid race conditions, for instance our saving child may
1690 overwrite the synchronous saving did by SHUTDOWN. */
1691 if (server.rdb_child_pid != -1) {
1692 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1693 kill(server.rdb_child_pid,SIGKILL);
1694 rdbRemoveTempFile(server.rdb_child_pid);
1695 }
1696 if (server.aof_state != REDIS_AOF_OFF) {
1697 /* Kill the AOF saving child as the AOF we already have may be longer
1698 * but contains the full dataset anyway. */
1699 if (server.aof_child_pid != -1) {
1700 redisLog(REDIS_WARNING,
1701 "There is a child rewriting the AOF. Killing it!");
1702 kill(server.aof_child_pid,SIGKILL);
1703 }
1704 /* Append only file: fsync() the AOF and exit */
1705 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1706 aof_fsync(server.aof_fd);
1707 }
1708 if ((server.saveparamslen > 0 && !nosave) || save) {
1709 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1710 /* Snapshotting. Perform a SYNC SAVE and exit */
1711 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1712 /* Ooops.. error saving! The best we can do is to continue
1713 * operating. Note that if there was a background saving process,
1714 * in the next cron() Redis will be notified that the background
1715 * saving aborted, handling special stuff like slaves pending for
1716 * synchronization... */
1717 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1718 return REDIS_ERR;
1719 }
1720 }
1721 if (server.daemonize) {
1722 redisLog(REDIS_NOTICE,"Removing the pid file.");
1723 unlink(server.pidfile);
1724 }
1725 /* Close the listening sockets. Apparently this allows faster restarts. */
1726 if (server.ipfd != -1) close(server.ipfd);
1727 if (server.sofd != -1) close(server.sofd);
1728 if (server.unixsocket) {
1729 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1730 unlink(server.unixsocket); /* don't care if this fails */
1731 }
1732
1733 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1734 return REDIS_OK;
1735 }
1736
1737 /*================================== Commands =============================== */
1738
1739 void authCommand(redisClient *c) {
1740 if (!server.requirepass) {
1741 addReplyError(c,"Client sent AUTH, but no password is set");
1742 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1743 c->authenticated = 1;
1744 addReply(c,shared.ok);
1745 } else {
1746 c->authenticated = 0;
1747 addReplyError(c,"invalid password");
1748 }
1749 }
1750
1751 void pingCommand(redisClient *c) {
1752 addReply(c,shared.pong);
1753 }
1754
1755 void echoCommand(redisClient *c) {
1756 addReplyBulk(c,c->argv[1]);
1757 }
1758
1759 void timeCommand(redisClient *c) {
1760 struct timeval tv;
1761
1762 /* gettimeofday() can only fail if &tv is a bad addresss so we
1763 * don't check for errors. */
1764 gettimeofday(&tv,NULL);
1765 addReplyMultiBulkLen(c,2);
1766 addReplyBulkLongLong(c,tv.tv_sec);
1767 addReplyBulkLongLong(c,tv.tv_usec);
1768 }
1769
1770 /* Convert an amount of bytes into a human readable string in the form
1771 * of 100B, 2G, 100M, 4K, and so forth. */
1772 void bytesToHuman(char *s, unsigned long long n) {
1773 double d;
1774
1775 if (n < 1024) {
1776 /* Bytes */
1777 sprintf(s,"%lluB",n);
1778 return;
1779 } else if (n < (1024*1024)) {
1780 d = (double)n/(1024);
1781 sprintf(s,"%.2fK",d);
1782 } else if (n < (1024LL*1024*1024)) {
1783 d = (double)n/(1024*1024);
1784 sprintf(s,"%.2fM",d);
1785 } else if (n < (1024LL*1024*1024*1024)) {
1786 d = (double)n/(1024LL*1024*1024);
1787 sprintf(s,"%.2fG",d);
1788 }
1789 }
1790
1791 /* Create the string returned by the INFO command. This is decoupled
1792 * by the INFO command itself as we need to report the same information
1793 * on memory corruption problems. */
1794 sds genRedisInfoString(char *section) {
1795 sds info = sdsempty();
1796 time_t uptime = server.unixtime-server.stat_starttime;
1797 int j, numcommands;
1798 struct rusage self_ru, c_ru;
1799 unsigned long lol, bib;
1800 int allsections = 0, defsections = 0;
1801 int sections = 0;
1802
1803 if (section) {
1804 allsections = strcasecmp(section,"all") == 0;
1805 defsections = strcasecmp(section,"default") == 0;
1806 }
1807
1808 getrusage(RUSAGE_SELF, &self_ru);
1809 getrusage(RUSAGE_CHILDREN, &c_ru);
1810 getClientsMaxBuffers(&lol,&bib);
1811
1812 /* Server */
1813 if (allsections || defsections || !strcasecmp(section,"server")) {
1814 struct utsname name;
1815
1816 if (sections++) info = sdscat(info,"\r\n");
1817 uname(&name);
1818 info = sdscatprintf(info,
1819 "# Server\r\n"
1820 "redis_version:%s\r\n"
1821 "redis_git_sha1:%s\r\n"
1822 "redis_git_dirty:%d\r\n"
1823 "os:%s %s %s\r\n"
1824 "arch_bits:%d\r\n"
1825 "multiplexing_api:%s\r\n"
1826 "gcc_version:%d.%d.%d\r\n"
1827 "process_id:%ld\r\n"
1828 "run_id:%s\r\n"
1829 "tcp_port:%d\r\n"
1830 "uptime_in_seconds:%ld\r\n"
1831 "uptime_in_days:%ld\r\n"
1832 "lru_clock:%ld\r\n",
1833 REDIS_VERSION,
1834 redisGitSHA1(),
1835 strtol(redisGitDirty(),NULL,10) > 0,
1836 name.sysname, name.release, name.machine,
1837 server.arch_bits,
1838 aeGetApiName(),
1839 #ifdef __GNUC__
1840 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1841 #else
1842 0,0,0,
1843 #endif
1844 (long) getpid(),
1845 server.runid,
1846 server.port,
1847 uptime,
1848 uptime/(3600*24),
1849 (unsigned long) server.lruclock);
1850 }
1851
1852 /* Clients */
1853 if (allsections || defsections || !strcasecmp(section,"clients")) {
1854 if (sections++) info = sdscat(info,"\r\n");
1855 info = sdscatprintf(info,
1856 "# Clients\r\n"
1857 "connected_clients:%lu\r\n"
1858 "client_longest_output_list:%lu\r\n"
1859 "client_biggest_input_buf:%lu\r\n"
1860 "blocked_clients:%d\r\n",
1861 listLength(server.clients)-listLength(server.slaves),
1862 lol, bib,
1863 server.bpop_blocked_clients);
1864 }
1865
1866 /* Memory */
1867 if (allsections || defsections || !strcasecmp(section,"memory")) {
1868 char hmem[64];
1869 char peak_hmem[64];
1870
1871 bytesToHuman(hmem,zmalloc_used_memory());
1872 bytesToHuman(peak_hmem,server.stat_peak_memory);
1873 if (sections++) info = sdscat(info,"\r\n");
1874 info = sdscatprintf(info,
1875 "# Memory\r\n"
1876 "used_memory:%zu\r\n"
1877 "used_memory_human:%s\r\n"
1878 "used_memory_rss:%zu\r\n"
1879 "used_memory_peak:%zu\r\n"
1880 "used_memory_peak_human:%s\r\n"
1881 "used_memory_lua:%lld\r\n"
1882 "mem_fragmentation_ratio:%.2f\r\n"
1883 "mem_allocator:%s\r\n",
1884 zmalloc_used_memory(),
1885 hmem,
1886 zmalloc_get_rss(),
1887 server.stat_peak_memory,
1888 peak_hmem,
1889 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1890 zmalloc_get_fragmentation_ratio(),
1891 ZMALLOC_LIB
1892 );
1893 }
1894
1895 /* Persistence */
1896 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1897 if (sections++) info = sdscat(info,"\r\n");
1898 info = sdscatprintf(info,
1899 "# Persistence\r\n"
1900 "loading:%d\r\n"
1901 "aof_enabled:%d\r\n"
1902 "changes_since_last_save:%lld\r\n"
1903 "bgsave_in_progress:%d\r\n"
1904 "last_save_time:%ld\r\n"
1905 "last_bgsave_status:%s\r\n"
1906 "bgrewriteaof_in_progress:%d\r\n"
1907 "bgrewriteaof_scheduled:%d\r\n",
1908 server.loading,
1909 server.aof_state != REDIS_AOF_OFF,
1910 server.dirty,
1911 server.rdb_child_pid != -1,
1912 server.lastsave,
1913 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1914 server.aof_child_pid != -1,
1915 server.aof_rewrite_scheduled);
1916
1917 if (server.aof_state != REDIS_AOF_OFF) {
1918 info = sdscatprintf(info,
1919 "aof_current_size:%lld\r\n"
1920 "aof_base_size:%lld\r\n"
1921 "aof_pending_rewrite:%d\r\n"
1922 "aof_buffer_length:%zu\r\n"
1923 "aof_rewrite_buffer_length:%zu\r\n"
1924 "aof_pending_bio_fsync:%llu\r\n"
1925 "aof_delayed_fsync:%lu\r\n",
1926 (long long) server.aof_current_size,
1927 (long long) server.aof_rewrite_base_size,
1928 server.aof_rewrite_scheduled,
1929 sdslen(server.aof_buf),
1930 aofRewriteBufferSize(),
1931 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1932 server.aof_delayed_fsync);
1933 }
1934
1935 if (server.loading) {
1936 double perc;
1937 time_t eta, elapsed;
1938 off_t remaining_bytes = server.loading_total_bytes-
1939 server.loading_loaded_bytes;
1940
1941 perc = ((double)server.loading_loaded_bytes /
1942 server.loading_total_bytes) * 100;
1943
1944 elapsed = server.unixtime-server.loading_start_time;
1945 if (elapsed == 0) {
1946 eta = 1; /* A fake 1 second figure if we don't have
1947 enough info */
1948 } else {
1949 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1950 }
1951
1952 info = sdscatprintf(info,
1953 "loading_start_time:%ld\r\n"
1954 "loading_total_bytes:%llu\r\n"
1955 "loading_loaded_bytes:%llu\r\n"
1956 "loading_loaded_perc:%.2f\r\n"
1957 "loading_eta_seconds:%ld\r\n"
1958 ,(unsigned long) server.loading_start_time,
1959 (unsigned long long) server.loading_total_bytes,
1960 (unsigned long long) server.loading_loaded_bytes,
1961 perc,
1962 eta
1963 );
1964 }
1965 }
1966
1967 /* Stats */
1968 if (allsections || defsections || !strcasecmp(section,"stats")) {
1969 if (sections++) info = sdscat(info,"\r\n");
1970 info = sdscatprintf(info,
1971 "# Stats\r\n"
1972 "total_connections_received:%lld\r\n"
1973 "total_commands_processed:%lld\r\n"
1974 "instantaneous_ops_per_sec:%lld\r\n"
1975 "rejected_connections:%lld\r\n"
1976 "expired_keys:%lld\r\n"
1977 "evicted_keys:%lld\r\n"
1978 "keyspace_hits:%lld\r\n"
1979 "keyspace_misses:%lld\r\n"
1980 "pubsub_channels:%ld\r\n"
1981 "pubsub_patterns:%lu\r\n"
1982 "latest_fork_usec:%lld\r\n",
1983 server.stat_numconnections,
1984 server.stat_numcommands,
1985 getOperationsPerSecond(),
1986 server.stat_rejected_conn,
1987 server.stat_expiredkeys,
1988 server.stat_evictedkeys,
1989 server.stat_keyspace_hits,
1990 server.stat_keyspace_misses,
1991 dictSize(server.pubsub_channels),
1992 listLength(server.pubsub_patterns),
1993 server.stat_fork_time);
1994 }
1995
1996 /* Replication */
1997 if (allsections || defsections || !strcasecmp(section,"replication")) {
1998 if (sections++) info = sdscat(info,"\r\n");
1999 info = sdscatprintf(info,
2000 "# Replication\r\n"
2001 "role:%s\r\n",
2002 server.masterhost == NULL ? "master" : "slave");
2003 if (server.masterhost) {
2004 info = sdscatprintf(info,
2005 "master_host:%s\r\n"
2006 "master_port:%d\r\n"
2007 "master_link_status:%s\r\n"
2008 "master_last_io_seconds_ago:%d\r\n"
2009 "master_sync_in_progress:%d\r\n"
2010 ,server.masterhost,
2011 server.masterport,
2012 (server.repl_state == REDIS_REPL_CONNECTED) ?
2013 "up" : "down",
2014 server.master ?
2015 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
2016 server.repl_state == REDIS_REPL_TRANSFER
2017 );
2018
2019 if (server.repl_state == REDIS_REPL_TRANSFER) {
2020 info = sdscatprintf(info,
2021 "master_sync_left_bytes:%ld\r\n"
2022 "master_sync_last_io_seconds_ago:%d\r\n"
2023 ,(long)server.repl_transfer_left,
2024 (int)(server.unixtime-server.repl_transfer_lastio)
2025 );
2026 }
2027
2028 if (server.repl_state != REDIS_REPL_CONNECTED) {
2029 info = sdscatprintf(info,
2030 "master_link_down_since_seconds:%ld\r\n",
2031 (long)server.unixtime-server.repl_down_since);
2032 }
2033 }
2034 info = sdscatprintf(info,
2035 "connected_slaves:%lu\r\n",
2036 listLength(server.slaves));
2037 if (listLength(server.slaves)) {
2038 int slaveid = 0;
2039 listNode *ln;
2040 listIter li;
2041
2042 listRewind(server.slaves,&li);
2043 while((ln = listNext(&li))) {
2044 redisClient *slave = listNodeValue(ln);
2045 char *state = NULL;
2046 char ip[32];
2047 int port;
2048
2049 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
2050 switch(slave->replstate) {
2051 case REDIS_REPL_WAIT_BGSAVE_START:
2052 case REDIS_REPL_WAIT_BGSAVE_END:
2053 state = "wait_bgsave";
2054 break;
2055 case REDIS_REPL_SEND_BULK:
2056 state = "send_bulk";
2057 break;
2058 case REDIS_REPL_ONLINE:
2059 state = "online";
2060 break;
2061 }
2062 if (state == NULL) continue;
2063 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2064 slaveid,ip,port,state);
2065 slaveid++;
2066 }
2067 }
2068 }
2069
2070 /* CPU */
2071 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2072 if (sections++) info = sdscat(info,"\r\n");
2073 info = sdscatprintf(info,
2074 "# CPU\r\n"
2075 "used_cpu_sys:%.2f\r\n"
2076 "used_cpu_user:%.2f\r\n"
2077 "used_cpu_sys_children:%.2f\r\n"
2078 "used_cpu_user_children:%.2f\r\n",
2079 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2080 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2081 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2082 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2083 }
2084
2085 /* cmdtime */
2086 if (allsections || !strcasecmp(section,"commandstats")) {
2087 if (sections++) info = sdscat(info,"\r\n");
2088 info = sdscatprintf(info, "# Commandstats\r\n");
2089 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2090 for (j = 0; j < numcommands; j++) {
2091 struct redisCommand *c = redisCommandTable+j;
2092
2093 if (!c->calls) continue;
2094 info = sdscatprintf(info,
2095 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2096 c->name, c->calls, c->microseconds,
2097 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2098 }
2099 }
2100
2101 /* Cluster */
2102 if (allsections || defsections || !strcasecmp(section,"cluster")) {
2103 if (sections++) info = sdscat(info,"\r\n");
2104 info = sdscatprintf(info,
2105 "# Cluster\r\n"
2106 "cluster_enabled:%d\r\n",
2107 server.cluster_enabled);
2108 }
2109
2110 /* Key space */
2111 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2112 if (sections++) info = sdscat(info,"\r\n");
2113 info = sdscatprintf(info, "# Keyspace\r\n");
2114 for (j = 0; j < server.dbnum; j++) {
2115 long long keys, vkeys;
2116
2117 keys = dictSize(server.db[j].dict);
2118 vkeys = dictSize(server.db[j].expires);
2119 if (keys || vkeys) {
2120 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2121 j, keys, vkeys);
2122 }
2123 }
2124 }
2125 return info;
2126 }
2127
2128 void infoCommand(redisClient *c) {
2129 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2130
2131 if (c->argc > 2) {
2132 addReply(c,shared.syntaxerr);
2133 return;
2134 }
2135 sds info = genRedisInfoString(section);
2136 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2137 (unsigned long)sdslen(info)));
2138 addReplySds(c,info);
2139 addReply(c,shared.crlf);
2140 }
2141
2142 void monitorCommand(redisClient *c) {
2143 /* ignore MONITOR if aleady slave or in monitor mode */
2144 if (c->flags & REDIS_SLAVE) return;
2145
2146 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2147 c->slaveseldb = 0;
2148 listAddNodeTail(server.monitors,c);
2149 addReply(c,shared.ok);
2150 }
2151
2152 /* ============================ Maxmemory directive ======================== */
2153
2154 /* This function gets called when 'maxmemory' is set on the config file to limit
2155 * the max memory used by the server, before processing a command.
2156 *
2157 * The goal of the function is to free enough memory to keep Redis under the
2158 * configured memory limit.
2159 *
2160 * The function starts calculating how many bytes should be freed to keep
2161 * Redis under the limit, and enters a loop selecting the best keys to
2162 * evict accordingly to the configured policy.
2163 *
2164 * If all the bytes needed to return back under the limit were freed the
2165 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2166 * should block the execution of commands that will result in more memory
2167 * used by the server.
2168 */
2169 int freeMemoryIfNeeded(void) {
2170 size_t mem_used, mem_tofree, mem_freed;
2171 int slaves = listLength(server.slaves);
2172
2173 /* Remove the size of slaves output buffers and AOF buffer from the
2174 * count of used memory. */
2175 mem_used = zmalloc_used_memory();
2176 if (slaves) {
2177 listIter li;
2178 listNode *ln;
2179
2180 listRewind(server.slaves,&li);
2181 while((ln = listNext(&li))) {
2182 redisClient *slave = listNodeValue(ln);
2183 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2184 if (obuf_bytes > mem_used)
2185 mem_used = 0;
2186 else
2187 mem_used -= obuf_bytes;
2188 }
2189 }
2190 if (server.aof_state != REDIS_AOF_OFF) {
2191 mem_used -= sdslen(server.aof_buf);
2192 mem_used -= aofRewriteBufferSize();
2193 }
2194
2195 /* Check if we are over the memory limit. */
2196 if (mem_used <= server.maxmemory) return REDIS_OK;
2197
2198 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2199 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2200
2201 /* Compute how much memory we need to free. */
2202 mem_tofree = mem_used - server.maxmemory;
2203 mem_freed = 0;
2204 while (mem_freed < mem_tofree) {
2205 int j, k, keys_freed = 0;
2206
2207 for (j = 0; j < server.dbnum; j++) {
2208 long bestval = 0; /* just to prevent warning */
2209 sds bestkey = NULL;
2210 struct dictEntry *de;
2211 redisDb *db = server.db+j;
2212 dict *dict;
2213
2214 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2215 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2216 {
2217 dict = server.db[j].dict;
2218 } else {
2219 dict = server.db[j].expires;
2220 }
2221 if (dictSize(dict) == 0) continue;
2222
2223 /* volatile-random and allkeys-random policy */
2224 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2225 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2226 {
2227 de = dictGetRandomKey(dict);
2228 bestkey = dictGetKey(de);
2229 }
2230
2231 /* volatile-lru and allkeys-lru policy */
2232 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2233 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2234 {
2235 for (k = 0; k < server.maxmemory_samples; k++) {
2236 sds thiskey;
2237 long thisval;
2238 robj *o;
2239
2240 de = dictGetRandomKey(dict);
2241 thiskey = dictGetKey(de);
2242 /* When policy is volatile-lru we need an additonal lookup
2243 * to locate the real key, as dict is set to db->expires. */
2244 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2245 de = dictFind(db->dict, thiskey);
2246 o = dictGetVal(de);
2247 thisval = estimateObjectIdleTime(o);
2248
2249 /* Higher idle time is better candidate for deletion */
2250 if (bestkey == NULL || thisval > bestval) {
2251 bestkey = thiskey;
2252 bestval = thisval;
2253 }
2254 }
2255 }
2256
2257 /* volatile-ttl */
2258 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2259 for (k = 0; k < server.maxmemory_samples; k++) {
2260 sds thiskey;
2261 long thisval;
2262
2263 de = dictGetRandomKey(dict);
2264 thiskey = dictGetKey(de);
2265 thisval = (long) dictGetVal(de);
2266
2267 /* Expire sooner (minor expire unix timestamp) is better
2268 * candidate for deletion */
2269 if (bestkey == NULL || thisval < bestval) {
2270 bestkey = thiskey;
2271 bestval = thisval;
2272 }
2273 }
2274 }
2275
2276 /* Finally remove the selected key. */
2277 if (bestkey) {
2278 long long delta;
2279
2280 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2281 propagateExpire(db,keyobj);
2282 /* We compute the amount of memory freed by dbDelete() alone.
2283 * It is possible that actually the memory needed to propagate
2284 * the DEL in AOF and replication link is greater than the one
2285 * we are freeing removing the key, but we can't account for
2286 * that otherwise we would never exit the loop.
2287 *
2288 * AOF and Output buffer memory will be freed eventually so
2289 * we only care about memory used by the key space. */
2290 delta = (long long) zmalloc_used_memory();
2291 dbDelete(db,keyobj);
2292 delta -= (long long) zmalloc_used_memory();
2293 mem_freed += delta;
2294 server.stat_evictedkeys++;
2295 decrRefCount(keyobj);
2296 keys_freed++;
2297
2298 /* When the memory to free starts to be big enough, we may
2299 * start spending so much time here that is impossible to
2300 * deliver data to the slaves fast enough, so we force the
2301 * transmission here inside the loop. */
2302 if (slaves) flushSlavesOutputBuffers();
2303 }
2304 }
2305 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2306 }
2307 return REDIS_OK;
2308 }
2309
2310 /* =================================== Main! ================================ */
2311
2312 #ifdef __linux__
2313 int linuxOvercommitMemoryValue(void) {
2314 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2315 char buf[64];
2316
2317 if (!fp) return -1;
2318 if (fgets(buf,64,fp) == NULL) {
2319 fclose(fp);
2320 return -1;
2321 }
2322 fclose(fp);
2323
2324 return atoi(buf);
2325 }
2326
2327 void linuxOvercommitMemoryWarning(void) {
2328 if (linuxOvercommitMemoryValue() == 0) {
2329 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.");
2330 }
2331 }
2332 #endif /* __linux__ */
2333
2334 void createPidFile(void) {
2335 /* Try to write the pid file in a best-effort way. */
2336 FILE *fp = fopen(server.pidfile,"w");
2337 if (fp) {
2338 fprintf(fp,"%d\n",(int)getpid());
2339 fclose(fp);
2340 }
2341 }
2342
2343 void daemonize(void) {
2344 int fd;
2345
2346 if (fork() != 0) exit(0); /* parent exits */
2347 setsid(); /* create a new session */
2348
2349 /* Every output goes to /dev/null. If Redis is daemonized but
2350 * the 'logfile' is set to 'stdout' in the configuration file
2351 * it will not log at all. */
2352 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2353 dup2(fd, STDIN_FILENO);
2354 dup2(fd, STDOUT_FILENO);
2355 dup2(fd, STDERR_FILENO);
2356 if (fd > STDERR_FILENO) close(fd);
2357 }
2358 }
2359
2360 void version() {
2361 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2362 REDIS_VERSION,
2363 redisGitSHA1(),
2364 atoi(redisGitDirty()) > 0,
2365 ZMALLOC_LIB,
2366 sizeof(long) == 4 ? 32 : 64);
2367 exit(0);
2368 }
2369
2370 void usage() {
2371 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2372 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2373 fprintf(stderr," ./redis-server -v or --version\n");
2374 fprintf(stderr," ./redis-server -h or --help\n");
2375 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2376 fprintf(stderr,"Examples:\n");
2377 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2378 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2379 fprintf(stderr," ./redis-server --port 7777\n");
2380 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2381 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2382 exit(1);
2383 }
2384
2385 void redisAsciiArt(void) {
2386 #include "asciilogo.h"
2387 char *buf = zmalloc(1024*16);
2388
2389 snprintf(buf,1024*16,ascii_logo,
2390 REDIS_VERSION,
2391 redisGitSHA1(),
2392 strtol(redisGitDirty(),NULL,10) > 0,
2393 (sizeof(long) == 8) ? "64" : "32",
2394 server.cluster_enabled ? "cluster" : "stand alone",
2395 server.port,
2396 (long) getpid()
2397 );
2398 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2399 zfree(buf);
2400 }
2401
2402 static void sigtermHandler(int sig) {
2403 REDIS_NOTUSED(sig);
2404
2405 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2406 server.shutdown_asap = 1;
2407 }
2408
2409 void setupSignalHandlers(void) {
2410 struct sigaction act;
2411
2412 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2413 * Otherwise, sa_handler is used. */
2414 sigemptyset(&act.sa_mask);
2415 act.sa_flags = 0;
2416 act.sa_handler = sigtermHandler;
2417 sigaction(SIGTERM, &act, NULL);
2418
2419 #ifdef HAVE_BACKTRACE
2420 sigemptyset(&act.sa_mask);
2421 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2422 act.sa_sigaction = sigsegvHandler;
2423 sigaction(SIGSEGV, &act, NULL);
2424 sigaction(SIGBUS, &act, NULL);
2425 sigaction(SIGFPE, &act, NULL);
2426 sigaction(SIGILL, &act, NULL);
2427 #endif
2428 return;
2429 }
2430
2431 void memtest(size_t megabytes, int passes);
2432
2433 int main(int argc, char **argv) {
2434 long long start;
2435 struct timeval tv;
2436
2437 /* We need to initialize our libraries, and the server configuration. */
2438 zmalloc_enable_thread_safeness();
2439 srand(time(NULL)^getpid());
2440 gettimeofday(&tv,NULL);
2441 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2442 initServerConfig();
2443
2444 if (argc >= 2) {
2445 int j = 1; /* First option to parse in argv[] */
2446 sds options = sdsempty();
2447 char *configfile = NULL;
2448
2449 /* Handle special options --help and --version */
2450 if (strcmp(argv[1], "-v") == 0 ||
2451 strcmp(argv[1], "--version") == 0) version();
2452 if (strcmp(argv[1], "--help") == 0 ||
2453 strcmp(argv[1], "-h") == 0) usage();
2454 if (strcmp(argv[1], "--test-memory") == 0) {
2455 if (argc == 3) {
2456 memtest(atoi(argv[2]),50);
2457 exit(0);
2458 } else {
2459 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2460 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2461 exit(1);
2462 }
2463 }
2464
2465 /* First argument is the config file name? */
2466 if (argv[j][0] != '-' || argv[j][1] != '-')
2467 configfile = argv[j++];
2468 /* All the other options are parsed and conceptually appended to the
2469 * configuration file. For instance --port 6380 will generate the
2470 * string "port 6380\n" to be parsed after the actual file name
2471 * is parsed, if any. */
2472 while(j != argc) {
2473 if (argv[j][0] == '-' && argv[j][1] == '-') {
2474 /* Option name */
2475 if (sdslen(options)) options = sdscat(options,"\n");
2476 options = sdscat(options,argv[j]+2);
2477 options = sdscat(options," ");
2478 } else {
2479 /* Option argument */
2480 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2481 options = sdscat(options," ");
2482 }
2483 j++;
2484 }
2485 resetServerSaveParams();
2486 loadServerConfig(configfile,options);
2487 sdsfree(options);
2488 } else {
2489 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'");
2490 }
2491 if (server.daemonize) daemonize();
2492 initServer();
2493 if (server.daemonize) createPidFile();
2494 redisAsciiArt();
2495 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2496 #ifdef __linux__
2497 linuxOvercommitMemoryWarning();
2498 #endif
2499 start = ustime();
2500 if (server.aof_state == REDIS_AOF_ON) {
2501 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2502 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2503 } else {
2504 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2505 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2506 (float)(ustime()-start)/1000000);
2507 } else if (errno != ENOENT) {
2508 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2509 exit(1);
2510 }
2511 }
2512 if (server.ipfd > 0)
2513 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2514 if (server.sofd > 0)
2515 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2516 aeSetBeforeSleepProc(server.el,beforeSleep);
2517 aeMain(server.el);
2518 aeDeleteEventLoop(server.el);
2519 return 0;
2520 }
2521
2522 /* The End */