]> git.saurik.com Git - redis.git/blob - src/redis.c
Fixed comment typo into time_independent_strcmp().
[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 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
237 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
238 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
239 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
240 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
241 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
242 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
243 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
244 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
245 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0},
246 {"bitop",bitopCommand,-4,"wm",0,NULL,2,-1,1,0,0},
247 {"bitcount",bitcountCommand,-2,"r",0,NULL,1,1,1,0,0}
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 int htNeedsResize(dict *dict) {
557 long long size, used;
558
559 size = dictSlots(dict);
560 used = dictSize(dict);
561 return (size && used && size > DICT_HT_INITIAL_SIZE &&
562 (used*100/size < REDIS_HT_MINFILL));
563 }
564
565 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
566 * we resize the hash table to save memory */
567 void tryResizeHashTables(void) {
568 int j;
569
570 for (j = 0; j < server.dbnum; j++) {
571 if (htNeedsResize(server.db[j].dict))
572 dictResize(server.db[j].dict);
573 if (htNeedsResize(server.db[j].expires))
574 dictResize(server.db[j].expires);
575 }
576 }
577
578 /* Our hash table implementation performs rehashing incrementally while
579 * we write/read from the hash table. Still if the server is idle, the hash
580 * table will use two tables for a long time. So we try to use 1 millisecond
581 * of CPU time at every serverCron() loop in order to rehash some key. */
582 void incrementallyRehash(void) {
583 int j;
584
585 for (j = 0; j < server.dbnum; j++) {
586 /* Keys dictionary */
587 if (dictIsRehashing(server.db[j].dict)) {
588 dictRehashMilliseconds(server.db[j].dict,1);
589 break; /* already used our millisecond for this loop... */
590 }
591 /* Expires */
592 if (dictIsRehashing(server.db[j].expires)) {
593 dictRehashMilliseconds(server.db[j].expires,1);
594 break; /* already used our millisecond for this loop... */
595 }
596 }
597 }
598
599 /* This function is called once a background process of some kind terminates,
600 * as we want to avoid resizing the hash tables when there is a child in order
601 * to play well with copy-on-write (otherwise when a resize happens lots of
602 * memory pages are copied). The goal of this function is to update the ability
603 * for dict.c to resize the hash tables accordingly to the fact we have o not
604 * running childs. */
605 void updateDictResizePolicy(void) {
606 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
607 dictEnableResize();
608 else
609 dictDisableResize();
610 }
611
612 /* ======================= Cron: called every 100 ms ======================== */
613
614 /* Try to expire a few timed out keys. The algorithm used is adaptive and
615 * will use few CPU cycles if there are few expiring keys, otherwise
616 * it will get more aggressive to avoid that too much memory is used by
617 * keys that can be removed from the keyspace. */
618 void activeExpireCycle(void) {
619 int j, iteration = 0;
620 long long start = ustime(), timelimit;
621
622 /* We can use at max REDIS_EXPIRELOOKUPS_TIME_PERC percentage of CPU time
623 * per iteration. Since this function gets called with a frequency of
624 * REDIS_HZ times per second, the following is the max amount of
625 * microseconds we can spend in this function. */
626 timelimit = 1000000*REDIS_EXPIRELOOKUPS_TIME_PERC/REDIS_HZ/100;
627 if (timelimit <= 0) timelimit = 1;
628
629 for (j = 0; j < server.dbnum; j++) {
630 int expired;
631 redisDb *db = server.db+j;
632
633 /* Continue to expire if at the end of the cycle more than 25%
634 * of the keys were expired. */
635 do {
636 unsigned long num = dictSize(db->expires);
637 unsigned long slots = dictSlots(db->expires);
638 long long now = mstime();
639
640 /* When there are less than 1% filled slots getting random
641 * keys is expensive, so stop here waiting for better times...
642 * The dictionary will be resized asap. */
643 if (num && slots > DICT_HT_INITIAL_SIZE &&
644 (num*100/slots < 1)) break;
645
646 /* The main collection cycle. Sample random keys among keys
647 * with an expire set, checking for expired ones. */
648 expired = 0;
649 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
650 num = REDIS_EXPIRELOOKUPS_PER_CRON;
651 while (num--) {
652 dictEntry *de;
653 long long t;
654
655 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
656 t = dictGetSignedIntegerVal(de);
657 if (now > t) {
658 sds key = dictGetKey(de);
659 robj *keyobj = createStringObject(key,sdslen(key));
660
661 propagateExpire(db,keyobj);
662 dbDelete(db,keyobj);
663 decrRefCount(keyobj);
664 expired++;
665 server.stat_expiredkeys++;
666 }
667 }
668 /* We can't block forever here even if there are many keys to
669 * expire. So after a given amount of milliseconds return to the
670 * caller waiting for the other active expire cycle. */
671 iteration++;
672 if ((iteration & 0xf) == 0 && /* check once every 16 cycles. */
673 (ustime()-start) > timelimit) return;
674 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
675 }
676 }
677
678 void updateLRUClock(void) {
679 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
680 REDIS_LRU_CLOCK_MAX;
681 }
682
683
684 /* Add a sample to the operations per second array of samples. */
685 void trackOperationsPerSecond(void) {
686 long long t = mstime() - server.ops_sec_last_sample_time;
687 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
688 long long ops_sec;
689
690 ops_sec = t > 0 ? (ops*1000/t) : 0;
691
692 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
693 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
694 server.ops_sec_last_sample_time = mstime();
695 server.ops_sec_last_sample_ops = server.stat_numcommands;
696 }
697
698 /* Return the mean of all the samples. */
699 long long getOperationsPerSecond(void) {
700 int j;
701 long long sum = 0;
702
703 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
704 sum += server.ops_sec_samples[j];
705 return sum / REDIS_OPS_SEC_SAMPLES;
706 }
707
708 /* Check for timeouts. Returns non-zero if the client was terminated */
709 int clientsCronHandleTimeout(redisClient *c) {
710 time_t now = server.unixtime;
711
712 if (server.maxidletime &&
713 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
714 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
715 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
716 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
717 listLength(c->pubsub_patterns) == 0 &&
718 (now - c->lastinteraction > server.maxidletime))
719 {
720 redisLog(REDIS_VERBOSE,"Closing idle client");
721 freeClient(c);
722 return 1;
723 } else if (c->flags & REDIS_BLOCKED) {
724 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
725 addReply(c,shared.nullmultibulk);
726 unblockClientWaitingData(c);
727 }
728 }
729 return 0;
730 }
731
732 /* The client query buffer is an sds.c string that can end with a lot of
733 * free space not used, this function reclaims space if needed.
734 *
735 * The funciton always returns 0 as it never terminates the client. */
736 int clientsCronResizeQueryBuffer(redisClient *c) {
737 size_t querybuf_size = sdsAllocSize(c->querybuf);
738 time_t idletime = server.unixtime - c->lastinteraction;
739
740 /* There are two conditions to resize the query buffer:
741 * 1) Query buffer is > BIG_ARG and too big for latest peak.
742 * 2) Client is inactive and the buffer is bigger than 1k. */
743 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
744 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
745 (querybuf_size > 1024 && idletime > 2))
746 {
747 /* Only resize the query buffer if it is actually wasting space. */
748 if (sdsavail(c->querybuf) > 1024) {
749 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
750 }
751 }
752 /* Reset the peak again to capture the peak memory usage in the next
753 * cycle. */
754 c->querybuf_peak = 0;
755 return 0;
756 }
757
758 void clientsCron(void) {
759 /* Make sure to process at least 1/(REDIS_HZ*10) of clients per call.
760 * Since this function is called REDIS_HZ times per second we are sure that
761 * in the worst case we process all the clients in 10 seconds.
762 * In normal conditions (a reasonable number of clients) we process
763 * all the clients in a shorter time. */
764 int numclients = listLength(server.clients);
765 int iterations = numclients/(REDIS_HZ*10);
766
767 if (iterations < 50)
768 iterations = (numclients < 50) ? numclients : 50;
769 while(listLength(server.clients) && iterations--) {
770 redisClient *c;
771 listNode *head;
772
773 /* Rotate the list, take the current head, process.
774 * This way if the client must be removed from the list it's the
775 * first element and we don't incur into O(N) computation. */
776 listRotate(server.clients);
777 head = listFirst(server.clients);
778 c = listNodeValue(head);
779 /* The following functions do different service checks on the client.
780 * The protocol is that they return non-zero if the client was
781 * terminated. */
782 if (clientsCronHandleTimeout(c)) continue;
783 if (clientsCronResizeQueryBuffer(c)) continue;
784 }
785 }
786
787 /* This is our timer interrupt, called REDIS_HZ times per second.
788 * Here is where we do a number of things that need to be done asynchronously.
789 * For instance:
790 *
791 * - Active expired keys collection (it is also performed in a lazy way on
792 * lookup).
793 * - Software watchdong.
794 * - Update some statistic.
795 * - Incremental rehashing of the DBs hash tables.
796 * - Triggering BGSAVE / AOF rewrite, and handling of terminated children.
797 * - Clients timeout of differnet kinds.
798 * - Replication reconnection.
799 * - Many more...
800 *
801 * Everything directly called here will be called REDIS_HZ times per second,
802 * so in order to throttle execution of things we want to do less frequently
803 * a macro is used: run_with_period(milliseconds) { .... }
804 */
805
806 /* Using the following macro you can run code inside serverCron() with the
807 * specified period, specified in milliseconds.
808 * The actual resolution depends on REDIS_HZ. */
809 #define run_with_period(_ms_) if (!(loops % ((_ms_)/(1000/REDIS_HZ))))
810
811 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
812 int j, loops = server.cronloops;
813 REDIS_NOTUSED(eventLoop);
814 REDIS_NOTUSED(id);
815 REDIS_NOTUSED(clientData);
816
817 /* Software watchdog: deliver the SIGALRM that will reach the signal
818 * handler if we don't return here fast enough. */
819 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
820
821 /* We take a cached value of the unix time in the global state because
822 * with virtual memory and aging there is to store the current time
823 * in objects at every object access, and accuracy is not needed.
824 * To access a global var is faster than calling time(NULL) */
825 server.unixtime = time(NULL);
826
827 run_with_period(100) trackOperationsPerSecond();
828
829 /* We have just 22 bits per object for LRU information.
830 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
831 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
832 *
833 * Note that even if this will wrap after 1.5 years it's not a problem,
834 * everything will still work but just some object will appear younger
835 * to Redis. But for this to happen a given object should never be touched
836 * for 1.5 years.
837 *
838 * Note that you can change the resolution altering the
839 * REDIS_LRU_CLOCK_RESOLUTION define.
840 */
841 updateLRUClock();
842
843 /* Record the max memory used since the server was started. */
844 if (zmalloc_used_memory() > server.stat_peak_memory)
845 server.stat_peak_memory = zmalloc_used_memory();
846
847 /* We received a SIGTERM, shutting down here in a safe way, as it is
848 * not ok doing so inside the signal handler. */
849 if (server.shutdown_asap) {
850 if (prepareForShutdown(0) == REDIS_OK) exit(0);
851 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
852 }
853
854 /* Show some info about non-empty databases */
855 run_with_period(5000) {
856 for (j = 0; j < server.dbnum; j++) {
857 long long size, used, vkeys;
858
859 size = dictSlots(server.db[j].dict);
860 used = dictSize(server.db[j].dict);
861 vkeys = dictSize(server.db[j].expires);
862 if (used || vkeys) {
863 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
864 /* dictPrintStats(server.dict); */
865 }
866 }
867 }
868
869 /* We don't want to resize the hash tables while a bacground saving
870 * is in progress: the saving child is created using fork() that is
871 * implemented with a copy-on-write semantic in most modern systems, so
872 * if we resize the HT while there is the saving child at work actually
873 * a lot of memory movements in the parent will cause a lot of pages
874 * copied. */
875 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
876 tryResizeHashTables();
877 if (server.activerehashing) incrementallyRehash();
878 }
879
880 /* Show information about connected clients */
881 run_with_period(5000) {
882 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
883 listLength(server.clients)-listLength(server.slaves),
884 listLength(server.slaves),
885 zmalloc_used_memory());
886 }
887
888 /* We need to do a few operations on clients asynchronously. */
889 clientsCron();
890
891 /* Start a scheduled AOF rewrite if this was requested by the user while
892 * a BGSAVE was in progress. */
893 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
894 server.aof_rewrite_scheduled)
895 {
896 rewriteAppendOnlyFileBackground();
897 }
898
899 /* Check if a background saving or AOF rewrite in progress terminated. */
900 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
901 int statloc;
902 pid_t pid;
903
904 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
905 int exitcode = WEXITSTATUS(statloc);
906 int bysignal = 0;
907
908 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
909
910 if (pid == server.rdb_child_pid) {
911 backgroundSaveDoneHandler(exitcode,bysignal);
912 } else {
913 backgroundRewriteDoneHandler(exitcode,bysignal);
914 }
915 updateDictResizePolicy();
916 }
917 } else {
918 /* If there is not a background saving/rewrite in progress check if
919 * we have to save/rewrite now */
920 for (j = 0; j < server.saveparamslen; j++) {
921 struct saveparam *sp = server.saveparams+j;
922
923 if (server.dirty >= sp->changes &&
924 server.unixtime-server.lastsave > sp->seconds) {
925 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
926 sp->changes, sp->seconds);
927 rdbSaveBackground(server.rdb_filename);
928 break;
929 }
930 }
931
932 /* Trigger an AOF rewrite if needed */
933 if (server.rdb_child_pid == -1 &&
934 server.aof_child_pid == -1 &&
935 server.aof_rewrite_perc &&
936 server.aof_current_size > server.aof_rewrite_min_size)
937 {
938 long long base = server.aof_rewrite_base_size ?
939 server.aof_rewrite_base_size : 1;
940 long long growth = (server.aof_current_size*100/base) - 100;
941 if (growth >= server.aof_rewrite_perc) {
942 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
943 rewriteAppendOnlyFileBackground();
944 }
945 }
946 }
947
948
949 /* If we postponed an AOF buffer flush, let's try to do it every time the
950 * cron function is called. */
951 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
952
953 /* Expire a few keys per cycle, only if this is a master.
954 * On slaves we wait for DEL operations synthesized by the master
955 * in order to guarantee a strict consistency. */
956 if (server.masterhost == NULL) activeExpireCycle();
957
958 /* Close clients that need to be closed asynchronous */
959 freeClientsInAsyncFreeQueue();
960
961 /* Replication cron function -- used to reconnect to master and
962 * to detect transfer failures. */
963 run_with_period(1000) replicationCron();
964
965 server.cronloops++;
966 return 1000/REDIS_HZ;
967 }
968
969 /* This function gets called every time Redis is entering the
970 * main loop of the event driven library, that is, before to sleep
971 * for ready file descriptors. */
972 void beforeSleep(struct aeEventLoop *eventLoop) {
973 REDIS_NOTUSED(eventLoop);
974 listNode *ln;
975 redisClient *c;
976
977 /* Try to process pending commands for clients that were just unblocked. */
978 while (listLength(server.unblocked_clients)) {
979 ln = listFirst(server.unblocked_clients);
980 redisAssert(ln != NULL);
981 c = ln->value;
982 listDelNode(server.unblocked_clients,ln);
983 c->flags &= ~REDIS_UNBLOCKED;
984
985 /* Process remaining data in the input buffer. */
986 if (c->querybuf && sdslen(c->querybuf) > 0) {
987 server.current_client = c;
988 processInputBuffer(c);
989 server.current_client = NULL;
990 }
991 }
992
993 /* Write the AOF buffer on disk */
994 flushAppendOnlyFile(0);
995 }
996
997 /* =========================== Server initialization ======================== */
998
999 void createSharedObjects(void) {
1000 int j;
1001
1002 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
1003 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
1004 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
1005 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
1006 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
1007 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
1008 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
1009 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
1010 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
1011 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
1012 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
1013 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
1014 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
1015 "-ERR Operation against a key holding the wrong kind of value\r\n"));
1016 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
1017 "-ERR no such key\r\n"));
1018 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
1019 "-ERR syntax error\r\n"));
1020 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
1021 "-ERR source and destination objects are the same\r\n"));
1022 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
1023 "-ERR index out of range\r\n"));
1024 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
1025 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
1026 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
1027 "-LOADING Redis is loading the dataset in memory\r\n"));
1028 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
1029 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
1030 shared.masterdownerr = createObject(REDIS_STRING,sdsnew(
1031 "-MASTERDOWN Link with MASTER is down and slave-serve-stale-data is set to 'no'.\r\n"));
1032 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
1033 "-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"));
1034 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
1035 "-READONLY You can't write against a read only slave.\r\n"));
1036 shared.oomerr = createObject(REDIS_STRING,sdsnew(
1037 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
1038 shared.space = createObject(REDIS_STRING,sdsnew(" "));
1039 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1040 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1041
1042 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1043 shared.select[j] = createObject(REDIS_STRING,
1044 sdscatprintf(sdsempty(),"select %d\r\n", j));
1045 }
1046 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1047 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1048 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1049 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1050 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1051 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1052 shared.del = createStringObject("DEL",3);
1053 shared.rpop = createStringObject("RPOP",4);
1054 shared.lpop = createStringObject("LPOP",4);
1055 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1056 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1057 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1058 }
1059 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1060 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1061 sdscatprintf(sdsempty(),"*%d\r\n",j));
1062 shared.bulkhdr[j] = createObject(REDIS_STRING,
1063 sdscatprintf(sdsempty(),"$%d\r\n",j));
1064 }
1065 }
1066
1067 void initServerConfig() {
1068 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1069 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1070 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1071 server.port = REDIS_SERVERPORT;
1072 server.bindaddr = NULL;
1073 server.unixsocket = NULL;
1074 server.unixsocketperm = 0;
1075 server.ipfd = -1;
1076 server.sofd = -1;
1077 server.dbnum = REDIS_DEFAULT_DBNUM;
1078 server.verbosity = REDIS_NOTICE;
1079 server.maxidletime = REDIS_MAXIDLETIME;
1080 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1081 server.saveparams = NULL;
1082 server.loading = 0;
1083 server.logfile = NULL; /* NULL = log on standard output */
1084 server.syslog_enabled = 0;
1085 server.syslog_ident = zstrdup("redis");
1086 server.syslog_facility = LOG_LOCAL0;
1087 server.daemonize = 0;
1088 server.aof_state = REDIS_AOF_OFF;
1089 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1090 server.aof_no_fsync_on_rewrite = 0;
1091 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1092 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1093 server.aof_rewrite_base_size = 0;
1094 server.aof_rewrite_scheduled = 0;
1095 server.aof_last_fsync = time(NULL);
1096 server.aof_rewrite_time_last = -1;
1097 server.aof_rewrite_time_start = -1;
1098 server.aof_delayed_fsync = 0;
1099 server.aof_fd = -1;
1100 server.aof_selected_db = -1; /* Make sure the first time will not match */
1101 server.aof_flush_postponed_start = 0;
1102 server.pidfile = zstrdup("/var/run/redis.pid");
1103 server.rdb_filename = zstrdup("dump.rdb");
1104 server.aof_filename = zstrdup("appendonly.aof");
1105 server.requirepass = NULL;
1106 server.rdb_compression = 1;
1107 server.rdb_checksum = 1;
1108 server.activerehashing = 1;
1109 server.maxclients = REDIS_MAX_CLIENTS;
1110 server.bpop_blocked_clients = 0;
1111 server.maxmemory = 0;
1112 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1113 server.maxmemory_samples = 3;
1114 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1115 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1116 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1117 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1118 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1119 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1120 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1121 server.shutdown_asap = 0;
1122 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1123 server.repl_timeout = REDIS_REPL_TIMEOUT;
1124 server.lua_caller = NULL;
1125 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1126 server.lua_client = NULL;
1127 server.lua_timedout = 0;
1128
1129 updateLRUClock();
1130 resetServerSaveParams();
1131
1132 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1133 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1134 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1135 /* Replication related */
1136 server.masterauth = NULL;
1137 server.masterhost = NULL;
1138 server.masterport = 6379;
1139 server.master = NULL;
1140 server.repl_state = REDIS_REPL_NONE;
1141 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1142 server.repl_serve_stale_data = 1;
1143 server.repl_slave_ro = 1;
1144 server.repl_down_since = time(NULL);
1145
1146 /* Client output buffer limits */
1147 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1148 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1149 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1150 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1151 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1152 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1153 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1154 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1155 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1156
1157 /* Double constants initialization */
1158 R_Zero = 0.0;
1159 R_PosInf = 1.0/R_Zero;
1160 R_NegInf = -1.0/R_Zero;
1161 R_Nan = R_Zero/R_Zero;
1162
1163 /* Command table -- we intiialize it here as it is part of the
1164 * initial configuration, since command names may be changed via
1165 * redis.conf using the rename-command directive. */
1166 server.commands = dictCreate(&commandTableDictType,NULL);
1167 populateCommandTable();
1168 server.delCommand = lookupCommandByCString("del");
1169 server.multiCommand = lookupCommandByCString("multi");
1170 server.lpushCommand = lookupCommandByCString("lpush");
1171
1172 /* Slow log */
1173 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1174 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1175
1176 /* Debugging */
1177 server.assert_failed = "<no assertion failed>";
1178 server.assert_file = "<no file>";
1179 server.assert_line = 0;
1180 server.bug_report_start = 0;
1181 server.watchdog_period = 0;
1182 }
1183
1184 /* This function will try to raise the max number of open files accordingly to
1185 * the configured max number of clients. It will also account for 32 additional
1186 * file descriptors as we need a few more for persistence, listening
1187 * sockets, log files and so forth.
1188 *
1189 * If it will not be possible to set the limit accordingly to the configured
1190 * max number of clients, the function will do the reverse setting
1191 * server.maxclients to the value that we can actually handle. */
1192 void adjustOpenFilesLimit(void) {
1193 rlim_t maxfiles = server.maxclients+32;
1194 struct rlimit limit;
1195
1196 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1197 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1198 strerror(errno));
1199 server.maxclients = 1024-32;
1200 } else {
1201 rlim_t oldlimit = limit.rlim_cur;
1202
1203 /* Set the max number of files if the current limit is not enough
1204 * for our needs. */
1205 if (oldlimit < maxfiles) {
1206 rlim_t f;
1207
1208 f = maxfiles;
1209 while(f > oldlimit) {
1210 limit.rlim_cur = f;
1211 limit.rlim_max = f;
1212 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1213 f -= 128;
1214 }
1215 if (f < oldlimit) f = oldlimit;
1216 if (f != maxfiles) {
1217 server.maxclients = f-32;
1218 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1219 (int) maxfiles, strerror(errno), (int) server.maxclients);
1220 } else {
1221 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1222 (int) maxfiles);
1223 }
1224 }
1225 }
1226 }
1227
1228 void initServer() {
1229 int j;
1230
1231 signal(SIGHUP, SIG_IGN);
1232 signal(SIGPIPE, SIG_IGN);
1233 setupSignalHandlers();
1234
1235 if (server.syslog_enabled) {
1236 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1237 server.syslog_facility);
1238 }
1239
1240 server.current_client = NULL;
1241 server.clients = listCreate();
1242 server.clients_to_close = listCreate();
1243 server.slaves = listCreate();
1244 server.monitors = listCreate();
1245 server.unblocked_clients = listCreate();
1246
1247 createSharedObjects();
1248 adjustOpenFilesLimit();
1249 server.el = aeCreateEventLoop(server.maxclients+1024);
1250 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1251
1252 if (server.port != 0) {
1253 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1254 if (server.ipfd == ANET_ERR) {
1255 redisLog(REDIS_WARNING, "Opening port %d: %s",
1256 server.port, server.neterr);
1257 exit(1);
1258 }
1259 }
1260 if (server.unixsocket != NULL) {
1261 unlink(server.unixsocket); /* don't care if this fails */
1262 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1263 if (server.sofd == ANET_ERR) {
1264 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1265 exit(1);
1266 }
1267 }
1268 if (server.ipfd < 0 && server.sofd < 0) {
1269 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1270 exit(1);
1271 }
1272 for (j = 0; j < server.dbnum; j++) {
1273 server.db[j].dict = dictCreate(&dbDictType,NULL);
1274 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1275 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1276 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1277 server.db[j].id = j;
1278 }
1279 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1280 server.pubsub_patterns = listCreate();
1281 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1282 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1283 server.cronloops = 0;
1284 server.rdb_child_pid = -1;
1285 server.aof_child_pid = -1;
1286 aofRewriteBufferReset();
1287 server.aof_buf = sdsempty();
1288 server.lastsave = time(NULL);
1289 server.rdb_save_time_last = -1;
1290 server.rdb_save_time_start = -1;
1291 server.dirty = 0;
1292 server.stat_numcommands = 0;
1293 server.stat_numconnections = 0;
1294 server.stat_expiredkeys = 0;
1295 server.stat_evictedkeys = 0;
1296 server.stat_starttime = time(NULL);
1297 server.stat_keyspace_misses = 0;
1298 server.stat_keyspace_hits = 0;
1299 server.stat_peak_memory = 0;
1300 server.stat_fork_time = 0;
1301 server.stat_rejected_conn = 0;
1302 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1303 server.ops_sec_idx = 0;
1304 server.ops_sec_last_sample_time = mstime();
1305 server.ops_sec_last_sample_ops = 0;
1306 server.unixtime = time(NULL);
1307 server.lastbgsave_status = REDIS_OK;
1308 server.stop_writes_on_bgsave_err = 1;
1309 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1310 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1311 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1312 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1313 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1314
1315 if (server.aof_state == REDIS_AOF_ON) {
1316 server.aof_fd = open(server.aof_filename,
1317 O_WRONLY|O_APPEND|O_CREAT,0644);
1318 if (server.aof_fd == -1) {
1319 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1320 strerror(errno));
1321 exit(1);
1322 }
1323 }
1324
1325 /* 32 bit instances are limited to 4GB of address space, so if there is
1326 * no explicit limit in the user provided configuration we set a limit
1327 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1328 * useless crashes of the Redis instance. */
1329 if (server.arch_bits == 32 && server.maxmemory == 0) {
1330 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1331 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1332 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1333 }
1334
1335 scriptingInit();
1336 slowlogInit();
1337 bioInit();
1338 }
1339
1340 /* Populates the Redis Command Table starting from the hard coded list
1341 * we have on top of redis.c file. */
1342 void populateCommandTable(void) {
1343 int j;
1344 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1345
1346 for (j = 0; j < numcommands; j++) {
1347 struct redisCommand *c = redisCommandTable+j;
1348 char *f = c->sflags;
1349 int retval;
1350
1351 while(*f != '\0') {
1352 switch(*f) {
1353 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1354 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1355 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1356 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1357 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1358 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1359 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1360 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1361 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1362 default: redisPanic("Unsupported command flag"); break;
1363 }
1364 f++;
1365 }
1366
1367 retval = dictAdd(server.commands, sdsnew(c->name), c);
1368 assert(retval == DICT_OK);
1369 }
1370 }
1371
1372 void resetCommandTableStats(void) {
1373 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1374 int j;
1375
1376 for (j = 0; j < numcommands; j++) {
1377 struct redisCommand *c = redisCommandTable+j;
1378
1379 c->microseconds = 0;
1380 c->calls = 0;
1381 }
1382 }
1383
1384 /* ========================== Redis OP Array API ============================ */
1385
1386 void redisOpArrayInit(redisOpArray *oa) {
1387 oa->ops = NULL;
1388 oa->numops = 0;
1389 }
1390
1391 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1392 robj **argv, int argc, int target)
1393 {
1394 redisOp *op;
1395
1396 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1397 op = oa->ops+oa->numops;
1398 op->cmd = cmd;
1399 op->dbid = dbid;
1400 op->argv = argv;
1401 op->argc = argc;
1402 op->target = target;
1403 oa->numops++;
1404 return oa->numops;
1405 }
1406
1407 void redisOpArrayFree(redisOpArray *oa) {
1408 while(oa->numops) {
1409 int j;
1410 redisOp *op;
1411
1412 oa->numops--;
1413 op = oa->ops+oa->numops;
1414 for (j = 0; j < op->argc; j++)
1415 decrRefCount(op->argv[j]);
1416 zfree(op->argv);
1417 }
1418 zfree(oa->ops);
1419 }
1420
1421 /* ====================== Commands lookup and execution ===================== */
1422
1423 struct redisCommand *lookupCommand(sds name) {
1424 return dictFetchValue(server.commands, name);
1425 }
1426
1427 struct redisCommand *lookupCommandByCString(char *s) {
1428 struct redisCommand *cmd;
1429 sds name = sdsnew(s);
1430
1431 cmd = dictFetchValue(server.commands, name);
1432 sdsfree(name);
1433 return cmd;
1434 }
1435
1436 /* Propagate the specified command (in the context of the specified database id)
1437 * to AOF, Slaves and Monitors.
1438 *
1439 * flags are an xor between:
1440 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1441 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1442 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1443 */
1444 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1445 int flags)
1446 {
1447 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1448 feedAppendOnlyFile(cmd,dbid,argv,argc);
1449 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1450 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1451 }
1452
1453 /* Used inside commands to schedule the propagation of additional commands
1454 * after the current command is propagated to AOF / Replication. */
1455 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1456 int target)
1457 {
1458 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1459 }
1460
1461 /* Call() is the core of Redis execution of a command */
1462 void call(redisClient *c, int flags) {
1463 long long dirty, start = ustime(), duration;
1464
1465 /* Sent the command to clients in MONITOR mode, only if the commands are
1466 * not geneated from reading an AOF. */
1467 if (listLength(server.monitors) && !server.loading)
1468 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1469
1470 /* Call the command. */
1471 redisOpArrayInit(&server.also_propagate);
1472 dirty = server.dirty;
1473 c->cmd->proc(c);
1474 dirty = server.dirty-dirty;
1475 duration = ustime()-start;
1476
1477 /* When EVAL is called loading the AOF we don't want commands called
1478 * from Lua to go into the slowlog or to populate statistics. */
1479 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1480 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1481
1482 /* Log the command into the Slow log if needed, and populate the
1483 * per-command statistics that we show in INFO commandstats. */
1484 if (flags & REDIS_CALL_SLOWLOG)
1485 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1486 if (flags & REDIS_CALL_STATS) {
1487 c->cmd->microseconds += duration;
1488 c->cmd->calls++;
1489 }
1490
1491 /* Propagate the command into the AOF and replication link */
1492 if (flags & REDIS_CALL_PROPAGATE) {
1493 int flags = REDIS_PROPAGATE_NONE;
1494
1495 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1496 flags |= REDIS_PROPAGATE_REPL;
1497 if (dirty)
1498 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1499 if (flags != REDIS_PROPAGATE_NONE)
1500 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1501 }
1502 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1503 * PUSH command. */
1504 if (server.also_propagate.numops) {
1505 int j;
1506 redisOp *rop;
1507
1508 for (j = 0; j < server.also_propagate.numops; j++) {
1509 rop = &server.also_propagate.ops[j];
1510 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1511 }
1512 redisOpArrayFree(&server.also_propagate);
1513 }
1514 server.stat_numcommands++;
1515 }
1516
1517 /* If this function gets called we already read a whole
1518 * command, argments are in the client argv/argc fields.
1519 * processCommand() execute the command or prepare the
1520 * server for a bulk read from the client.
1521 *
1522 * If 1 is returned the client is still alive and valid and
1523 * and other operations can be performed by the caller. Otherwise
1524 * if 0 is returned the client was destroied (i.e. after QUIT). */
1525 int processCommand(redisClient *c) {
1526 /* The QUIT command is handled separately. Normal command procs will
1527 * go through checking for replication and QUIT will cause trouble
1528 * when FORCE_REPLICATION is enabled and would be implemented in
1529 * a regular command proc. */
1530 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1531 addReply(c,shared.ok);
1532 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1533 return REDIS_ERR;
1534 }
1535
1536 /* Now lookup the command and check ASAP about trivial error conditions
1537 * such as wrong arity, bad command name and so forth. */
1538 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1539 if (!c->cmd) {
1540 addReplyErrorFormat(c,"unknown command '%s'",
1541 (char*)c->argv[0]->ptr);
1542 return REDIS_OK;
1543 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1544 (c->argc < -c->cmd->arity)) {
1545 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1546 c->cmd->name);
1547 return REDIS_OK;
1548 }
1549
1550 /* Check if the user is authenticated */
1551 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1552 {
1553 addReplyError(c,"operation not permitted");
1554 return REDIS_OK;
1555 }
1556
1557 /* Handle the maxmemory directive.
1558 *
1559 * First we try to free some memory if possible (if there are volatile
1560 * keys in the dataset). If there are not the only thing we can do
1561 * is returning an error. */
1562 if (server.maxmemory) {
1563 int retval = freeMemoryIfNeeded();
1564 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1565 addReply(c, shared.oomerr);
1566 return REDIS_OK;
1567 }
1568 }
1569
1570 /* Don't accept write commands if there are problems persisting on disk. */
1571 if (server.stop_writes_on_bgsave_err &&
1572 server.saveparamslen > 0
1573 && server.lastbgsave_status == REDIS_ERR &&
1574 c->cmd->flags & REDIS_CMD_WRITE)
1575 {
1576 addReply(c, shared.bgsaveerr);
1577 return REDIS_OK;
1578 }
1579
1580 /* Don't accept wirte commands if this is a read only slave. But
1581 * accept write commands if this is our master. */
1582 if (server.masterhost && server.repl_slave_ro &&
1583 !(c->flags & REDIS_MASTER) &&
1584 c->cmd->flags & REDIS_CMD_WRITE)
1585 {
1586 addReply(c, shared.roslaveerr);
1587 return REDIS_OK;
1588 }
1589
1590 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1591 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1592 &&
1593 c->cmd->proc != subscribeCommand &&
1594 c->cmd->proc != unsubscribeCommand &&
1595 c->cmd->proc != psubscribeCommand &&
1596 c->cmd->proc != punsubscribeCommand) {
1597 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1598 return REDIS_OK;
1599 }
1600
1601 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1602 * we are a slave with a broken link with master. */
1603 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1604 server.repl_serve_stale_data == 0 &&
1605 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1606 {
1607 addReply(c, shared.masterdownerr);
1608 return REDIS_OK;
1609 }
1610
1611 /* Loading DB? Return an error if the command is not INFO */
1612 if (server.loading && c->cmd->proc != infoCommand) {
1613 addReply(c, shared.loadingerr);
1614 return REDIS_OK;
1615 }
1616
1617 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1618 if (server.lua_timedout &&
1619 !(c->cmd->proc == shutdownCommand &&
1620 c->argc == 2 &&
1621 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1622 !(c->cmd->proc == scriptCommand &&
1623 c->argc == 2 &&
1624 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1625 {
1626 addReply(c, shared.slowscripterr);
1627 return REDIS_OK;
1628 }
1629
1630 /* Exec the command */
1631 if (c->flags & REDIS_MULTI &&
1632 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1633 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1634 {
1635 queueMultiCommand(c);
1636 addReply(c,shared.queued);
1637 } else {
1638 call(c,REDIS_CALL_FULL);
1639 }
1640 return REDIS_OK;
1641 }
1642
1643 /*================================== Shutdown =============================== */
1644
1645 int prepareForShutdown(int flags) {
1646 int save = flags & REDIS_SHUTDOWN_SAVE;
1647 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1648
1649 redisLog(REDIS_WARNING,"User requested shutdown...");
1650 /* Kill the saving child if there is a background saving in progress.
1651 We want to avoid race conditions, for instance our saving child may
1652 overwrite the synchronous saving did by SHUTDOWN. */
1653 if (server.rdb_child_pid != -1) {
1654 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1655 kill(server.rdb_child_pid,SIGKILL);
1656 rdbRemoveTempFile(server.rdb_child_pid);
1657 }
1658 if (server.aof_state != REDIS_AOF_OFF) {
1659 /* Kill the AOF saving child as the AOF we already have may be longer
1660 * but contains the full dataset anyway. */
1661 if (server.aof_child_pid != -1) {
1662 redisLog(REDIS_WARNING,
1663 "There is a child rewriting the AOF. Killing it!");
1664 kill(server.aof_child_pid,SIGKILL);
1665 }
1666 /* Append only file: fsync() the AOF and exit */
1667 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1668 aof_fsync(server.aof_fd);
1669 }
1670 if ((server.saveparamslen > 0 && !nosave) || save) {
1671 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1672 /* Snapshotting. Perform a SYNC SAVE and exit */
1673 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1674 /* Ooops.. error saving! The best we can do is to continue
1675 * operating. Note that if there was a background saving process,
1676 * in the next cron() Redis will be notified that the background
1677 * saving aborted, handling special stuff like slaves pending for
1678 * synchronization... */
1679 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1680 return REDIS_ERR;
1681 }
1682 }
1683 if (server.daemonize) {
1684 redisLog(REDIS_NOTICE,"Removing the pid file.");
1685 unlink(server.pidfile);
1686 }
1687 /* Close the listening sockets. Apparently this allows faster restarts. */
1688 if (server.ipfd != -1) close(server.ipfd);
1689 if (server.sofd != -1) close(server.sofd);
1690 if (server.unixsocket) {
1691 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1692 unlink(server.unixsocket); /* don't care if this fails */
1693 }
1694
1695 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1696 return REDIS_OK;
1697 }
1698
1699 /*================================== Commands =============================== */
1700
1701 /* Return zero if strings are the same, non-zero if they are not.
1702 * The comparison is performed in a way that prevents an attacker to obtain
1703 * information about the nature of the strings just monitoring the execution
1704 * time of the function.
1705 *
1706 * Note that limiting the comparison length to strings up to 512 bytes we
1707 * can avoid leaking any information about the password length and any
1708 * possible branch misprediction related leak.
1709 */
1710 int time_independent_strcmp(char *a, char *b) {
1711 char bufa[REDIS_AUTHPASS_MAX_LEN], bufb[REDIS_AUTHPASS_MAX_LEN];
1712 /* The above two strlen perform len(a) + len(b) operations where either
1713 * a or b are fixed (our password) length, and the difference is only
1714 * relative to the length of the user provided string, so no information
1715 * leak is possible in the following two lines of code. */
1716 int alen = strlen(a);
1717 int blen = strlen(b);
1718 int j;
1719 int diff = 0;
1720
1721 /* We can't compare strings longer than our static buffers.
1722 * Note that this will never pass the first test in practical circumstances
1723 * so there is no info leak. */
1724 if (alen > sizeof(bufa) || blen > sizeof(bufb)) return 1;
1725
1726 memset(bufa,0,sizeof(bufa)); /* Constant time. */
1727 memset(bufb,0,sizeof(bufb)); /* Constant time. */
1728 /* Again the time of the following two copies is proportional to
1729 * len(a) + len(b) so no info is leaked. */
1730 memcpy(bufa,a,alen);
1731 memcpy(bufb,b,blen);
1732
1733 /* Always compare all the chars in the two buffers without
1734 * conditional expressions. */
1735 for (j = 0; j < sizeof(bufa); j++) {
1736 diff |= (bufa[j] ^ bufb[j]);
1737 }
1738 /* Length must be equal as well. */
1739 diff |= alen ^ blen;
1740 return diff; /* If zero strings are the same. */
1741 }
1742
1743 void authCommand(redisClient *c) {
1744 if (!server.requirepass) {
1745 addReplyError(c,"Client sent AUTH, but no password is set");
1746 } else if (!time_independent_strcmp(c->argv[1]->ptr, server.requirepass)) {
1747 c->authenticated = 1;
1748 addReply(c,shared.ok);
1749 } else {
1750 c->authenticated = 0;
1751 addReplyError(c,"invalid password");
1752 }
1753 }
1754
1755 void pingCommand(redisClient *c) {
1756 addReply(c,shared.pong);
1757 }
1758
1759 void echoCommand(redisClient *c) {
1760 addReplyBulk(c,c->argv[1]);
1761 }
1762
1763 void timeCommand(redisClient *c) {
1764 struct timeval tv;
1765
1766 /* gettimeofday() can only fail if &tv is a bad addresss so we
1767 * don't check for errors. */
1768 gettimeofday(&tv,NULL);
1769 addReplyMultiBulkLen(c,2);
1770 addReplyBulkLongLong(c,tv.tv_sec);
1771 addReplyBulkLongLong(c,tv.tv_usec);
1772 }
1773
1774 /* Convert an amount of bytes into a human readable string in the form
1775 * of 100B, 2G, 100M, 4K, and so forth. */
1776 void bytesToHuman(char *s, unsigned long long n) {
1777 double d;
1778
1779 if (n < 1024) {
1780 /* Bytes */
1781 sprintf(s,"%lluB",n);
1782 return;
1783 } else if (n < (1024*1024)) {
1784 d = (double)n/(1024);
1785 sprintf(s,"%.2fK",d);
1786 } else if (n < (1024LL*1024*1024)) {
1787 d = (double)n/(1024*1024);
1788 sprintf(s,"%.2fM",d);
1789 } else if (n < (1024LL*1024*1024*1024)) {
1790 d = (double)n/(1024LL*1024*1024);
1791 sprintf(s,"%.2fG",d);
1792 }
1793 }
1794
1795 /* Create the string returned by the INFO command. This is decoupled
1796 * by the INFO command itself as we need to report the same information
1797 * on memory corruption problems. */
1798 sds genRedisInfoString(char *section) {
1799 sds info = sdsempty();
1800 time_t uptime = server.unixtime-server.stat_starttime;
1801 int j, numcommands;
1802 struct rusage self_ru, c_ru;
1803 unsigned long lol, bib;
1804 int allsections = 0, defsections = 0;
1805 int sections = 0;
1806
1807 if (section) {
1808 allsections = strcasecmp(section,"all") == 0;
1809 defsections = strcasecmp(section,"default") == 0;
1810 }
1811
1812 getrusage(RUSAGE_SELF, &self_ru);
1813 getrusage(RUSAGE_CHILDREN, &c_ru);
1814 getClientsMaxBuffers(&lol,&bib);
1815
1816 /* Server */
1817 if (allsections || defsections || !strcasecmp(section,"server")) {
1818 struct utsname name;
1819
1820 if (sections++) info = sdscat(info,"\r\n");
1821 uname(&name);
1822 info = sdscatprintf(info,
1823 "# Server\r\n"
1824 "redis_version:%s\r\n"
1825 "redis_git_sha1:%s\r\n"
1826 "redis_git_dirty:%d\r\n"
1827 "os:%s %s %s\r\n"
1828 "arch_bits:%d\r\n"
1829 "multiplexing_api:%s\r\n"
1830 "gcc_version:%d.%d.%d\r\n"
1831 "process_id:%ld\r\n"
1832 "run_id:%s\r\n"
1833 "tcp_port:%d\r\n"
1834 "uptime_in_seconds:%ld\r\n"
1835 "uptime_in_days:%ld\r\n"
1836 "lru_clock:%ld\r\n",
1837 REDIS_VERSION,
1838 redisGitSHA1(),
1839 strtol(redisGitDirty(),NULL,10) > 0,
1840 name.sysname, name.release, name.machine,
1841 server.arch_bits,
1842 aeGetApiName(),
1843 #ifdef __GNUC__
1844 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1845 #else
1846 0,0,0,
1847 #endif
1848 (long) getpid(),
1849 server.runid,
1850 server.port,
1851 uptime,
1852 uptime/(3600*24),
1853 (unsigned long) server.lruclock);
1854 }
1855
1856 /* Clients */
1857 if (allsections || defsections || !strcasecmp(section,"clients")) {
1858 if (sections++) info = sdscat(info,"\r\n");
1859 info = sdscatprintf(info,
1860 "# Clients\r\n"
1861 "connected_clients:%lu\r\n"
1862 "client_longest_output_list:%lu\r\n"
1863 "client_biggest_input_buf:%lu\r\n"
1864 "blocked_clients:%d\r\n",
1865 listLength(server.clients)-listLength(server.slaves),
1866 lol, bib,
1867 server.bpop_blocked_clients);
1868 }
1869
1870 /* Memory */
1871 if (allsections || defsections || !strcasecmp(section,"memory")) {
1872 char hmem[64];
1873 char peak_hmem[64];
1874
1875 bytesToHuman(hmem,zmalloc_used_memory());
1876 bytesToHuman(peak_hmem,server.stat_peak_memory);
1877 if (sections++) info = sdscat(info,"\r\n");
1878 info = sdscatprintf(info,
1879 "# Memory\r\n"
1880 "used_memory:%zu\r\n"
1881 "used_memory_human:%s\r\n"
1882 "used_memory_rss:%zu\r\n"
1883 "used_memory_peak:%zu\r\n"
1884 "used_memory_peak_human:%s\r\n"
1885 "used_memory_lua:%lld\r\n"
1886 "mem_fragmentation_ratio:%.2f\r\n"
1887 "mem_allocator:%s\r\n",
1888 zmalloc_used_memory(),
1889 hmem,
1890 zmalloc_get_rss(),
1891 server.stat_peak_memory,
1892 peak_hmem,
1893 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1894 zmalloc_get_fragmentation_ratio(),
1895 ZMALLOC_LIB
1896 );
1897 }
1898
1899 /* Persistence */
1900 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1901 if (sections++) info = sdscat(info,"\r\n");
1902 info = sdscatprintf(info,
1903 "# Persistence\r\n"
1904 "loading:%d\r\n"
1905 "rdb_changes_since_last_save:%lld\r\n"
1906 "rdb_bgsave_in_progress:%d\r\n"
1907 "rdb_last_save_time:%ld\r\n"
1908 "rdb_last_bgsave_status:%s\r\n"
1909 "rdb_last_bgsave_time_sec:%ld\r\n"
1910 "rdb_current_bgsave_time_sec:%ld\r\n"
1911 "aof_enabled:%d\r\n"
1912 "aof_rewrite_in_progress:%d\r\n"
1913 "aof_rewrite_scheduled:%d\r\n"
1914 "aof_last_rewrite_time_sec:%ld\r\n"
1915 "aof_current_rewrite_time_sec:%ld\r\n",
1916 server.loading,
1917 server.dirty,
1918 server.rdb_child_pid != -1,
1919 server.lastsave,
1920 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1921 server.rdb_save_time_last,
1922 (server.rdb_child_pid == -1) ?
1923 -1 : time(NULL)-server.rdb_save_time_start,
1924 server.aof_state != REDIS_AOF_OFF,
1925 server.aof_child_pid != -1,
1926 server.aof_rewrite_scheduled,
1927 server.aof_rewrite_time_last,
1928 (server.aof_child_pid == -1) ?
1929 -1 : time(NULL)-server.aof_rewrite_time_start);
1930
1931 if (server.aof_state != REDIS_AOF_OFF) {
1932 info = sdscatprintf(info,
1933 "aof_current_size:%lld\r\n"
1934 "aof_base_size:%lld\r\n"
1935 "aof_pending_rewrite:%d\r\n"
1936 "aof_buffer_length:%zu\r\n"
1937 "aof_rewrite_buffer_length:%zu\r\n"
1938 "aof_pending_bio_fsync:%llu\r\n"
1939 "aof_delayed_fsync:%lu\r\n",
1940 (long long) server.aof_current_size,
1941 (long long) server.aof_rewrite_base_size,
1942 server.aof_rewrite_scheduled,
1943 sdslen(server.aof_buf),
1944 aofRewriteBufferSize(),
1945 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1946 server.aof_delayed_fsync);
1947 }
1948
1949 if (server.loading) {
1950 double perc;
1951 time_t eta, elapsed;
1952 off_t remaining_bytes = server.loading_total_bytes-
1953 server.loading_loaded_bytes;
1954
1955 perc = ((double)server.loading_loaded_bytes /
1956 server.loading_total_bytes) * 100;
1957
1958 elapsed = server.unixtime-server.loading_start_time;
1959 if (elapsed == 0) {
1960 eta = 1; /* A fake 1 second figure if we don't have
1961 enough info */
1962 } else {
1963 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1964 }
1965
1966 info = sdscatprintf(info,
1967 "loading_start_time:%ld\r\n"
1968 "loading_total_bytes:%llu\r\n"
1969 "loading_loaded_bytes:%llu\r\n"
1970 "loading_loaded_perc:%.2f\r\n"
1971 "loading_eta_seconds:%ld\r\n"
1972 ,(unsigned long) server.loading_start_time,
1973 (unsigned long long) server.loading_total_bytes,
1974 (unsigned long long) server.loading_loaded_bytes,
1975 perc,
1976 eta
1977 );
1978 }
1979 }
1980
1981 /* Stats */
1982 if (allsections || defsections || !strcasecmp(section,"stats")) {
1983 if (sections++) info = sdscat(info,"\r\n");
1984 info = sdscatprintf(info,
1985 "# Stats\r\n"
1986 "total_connections_received:%lld\r\n"
1987 "total_commands_processed:%lld\r\n"
1988 "instantaneous_ops_per_sec:%lld\r\n"
1989 "rejected_connections:%lld\r\n"
1990 "expired_keys:%lld\r\n"
1991 "evicted_keys:%lld\r\n"
1992 "keyspace_hits:%lld\r\n"
1993 "keyspace_misses:%lld\r\n"
1994 "pubsub_channels:%ld\r\n"
1995 "pubsub_patterns:%lu\r\n"
1996 "latest_fork_usec:%lld\r\n",
1997 server.stat_numconnections,
1998 server.stat_numcommands,
1999 getOperationsPerSecond(),
2000 server.stat_rejected_conn,
2001 server.stat_expiredkeys,
2002 server.stat_evictedkeys,
2003 server.stat_keyspace_hits,
2004 server.stat_keyspace_misses,
2005 dictSize(server.pubsub_channels),
2006 listLength(server.pubsub_patterns),
2007 server.stat_fork_time);
2008 }
2009
2010 /* Replication */
2011 if (allsections || defsections || !strcasecmp(section,"replication")) {
2012 if (sections++) info = sdscat(info,"\r\n");
2013 info = sdscatprintf(info,
2014 "# Replication\r\n"
2015 "role:%s\r\n",
2016 server.masterhost == NULL ? "master" : "slave");
2017 if (server.masterhost) {
2018 info = sdscatprintf(info,
2019 "master_host:%s\r\n"
2020 "master_port:%d\r\n"
2021 "master_link_status:%s\r\n"
2022 "master_last_io_seconds_ago:%d\r\n"
2023 "master_sync_in_progress:%d\r\n"
2024 ,server.masterhost,
2025 server.masterport,
2026 (server.repl_state == REDIS_REPL_CONNECTED) ?
2027 "up" : "down",
2028 server.master ?
2029 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
2030 server.repl_state == REDIS_REPL_TRANSFER
2031 );
2032
2033 if (server.repl_state == REDIS_REPL_TRANSFER) {
2034 info = sdscatprintf(info,
2035 "master_sync_left_bytes:%ld\r\n"
2036 "master_sync_last_io_seconds_ago:%d\r\n"
2037 ,(long)server.repl_transfer_left,
2038 (int)(server.unixtime-server.repl_transfer_lastio)
2039 );
2040 }
2041
2042 if (server.repl_state != REDIS_REPL_CONNECTED) {
2043 info = sdscatprintf(info,
2044 "master_link_down_since_seconds:%ld\r\n",
2045 (long)server.unixtime-server.repl_down_since);
2046 }
2047 }
2048 info = sdscatprintf(info,
2049 "connected_slaves:%lu\r\n",
2050 listLength(server.slaves));
2051 if (listLength(server.slaves)) {
2052 int slaveid = 0;
2053 listNode *ln;
2054 listIter li;
2055
2056 listRewind(server.slaves,&li);
2057 while((ln = listNext(&li))) {
2058 redisClient *slave = listNodeValue(ln);
2059 char *state = NULL;
2060 char ip[32];
2061 int port;
2062
2063 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
2064 switch(slave->replstate) {
2065 case REDIS_REPL_WAIT_BGSAVE_START:
2066 case REDIS_REPL_WAIT_BGSAVE_END:
2067 state = "wait_bgsave";
2068 break;
2069 case REDIS_REPL_SEND_BULK:
2070 state = "send_bulk";
2071 break;
2072 case REDIS_REPL_ONLINE:
2073 state = "online";
2074 break;
2075 }
2076 if (state == NULL) continue;
2077 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2078 slaveid,ip,port,state);
2079 slaveid++;
2080 }
2081 }
2082 }
2083
2084 /* CPU */
2085 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2086 if (sections++) info = sdscat(info,"\r\n");
2087 info = sdscatprintf(info,
2088 "# CPU\r\n"
2089 "used_cpu_sys:%.2f\r\n"
2090 "used_cpu_user:%.2f\r\n"
2091 "used_cpu_sys_children:%.2f\r\n"
2092 "used_cpu_user_children:%.2f\r\n",
2093 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2094 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2095 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2096 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2097 }
2098
2099 /* cmdtime */
2100 if (allsections || !strcasecmp(section,"commandstats")) {
2101 if (sections++) info = sdscat(info,"\r\n");
2102 info = sdscatprintf(info, "# Commandstats\r\n");
2103 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2104 for (j = 0; j < numcommands; j++) {
2105 struct redisCommand *c = redisCommandTable+j;
2106
2107 if (!c->calls) continue;
2108 info = sdscatprintf(info,
2109 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2110 c->name, c->calls, c->microseconds,
2111 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2112 }
2113 }
2114
2115 /* Key space */
2116 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2117 if (sections++) info = sdscat(info,"\r\n");
2118 info = sdscatprintf(info, "# Keyspace\r\n");
2119 for (j = 0; j < server.dbnum; j++) {
2120 long long keys, vkeys;
2121
2122 keys = dictSize(server.db[j].dict);
2123 vkeys = dictSize(server.db[j].expires);
2124 if (keys || vkeys) {
2125 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2126 j, keys, vkeys);
2127 }
2128 }
2129 }
2130 return info;
2131 }
2132
2133 void infoCommand(redisClient *c) {
2134 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2135
2136 if (c->argc > 2) {
2137 addReply(c,shared.syntaxerr);
2138 return;
2139 }
2140 sds info = genRedisInfoString(section);
2141 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2142 (unsigned long)sdslen(info)));
2143 addReplySds(c,info);
2144 addReply(c,shared.crlf);
2145 }
2146
2147 void monitorCommand(redisClient *c) {
2148 /* ignore MONITOR if aleady slave or in monitor mode */
2149 if (c->flags & REDIS_SLAVE) return;
2150
2151 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2152 c->slaveseldb = 0;
2153 listAddNodeTail(server.monitors,c);
2154 addReply(c,shared.ok);
2155 }
2156
2157 /* ============================ Maxmemory directive ======================== */
2158
2159 /* This function gets called when 'maxmemory' is set on the config file to limit
2160 * the max memory used by the server, before processing a command.
2161 *
2162 * The goal of the function is to free enough memory to keep Redis under the
2163 * configured memory limit.
2164 *
2165 * The function starts calculating how many bytes should be freed to keep
2166 * Redis under the limit, and enters a loop selecting the best keys to
2167 * evict accordingly to the configured policy.
2168 *
2169 * If all the bytes needed to return back under the limit were freed the
2170 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2171 * should block the execution of commands that will result in more memory
2172 * used by the server.
2173 */
2174 int freeMemoryIfNeeded(void) {
2175 size_t mem_used, mem_tofree, mem_freed;
2176 int slaves = listLength(server.slaves);
2177
2178 /* Remove the size of slaves output buffers and AOF buffer from the
2179 * count of used memory. */
2180 mem_used = zmalloc_used_memory();
2181 if (slaves) {
2182 listIter li;
2183 listNode *ln;
2184
2185 listRewind(server.slaves,&li);
2186 while((ln = listNext(&li))) {
2187 redisClient *slave = listNodeValue(ln);
2188 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2189 if (obuf_bytes > mem_used)
2190 mem_used = 0;
2191 else
2192 mem_used -= obuf_bytes;
2193 }
2194 }
2195 if (server.aof_state != REDIS_AOF_OFF) {
2196 mem_used -= sdslen(server.aof_buf);
2197 mem_used -= aofRewriteBufferSize();
2198 }
2199
2200 /* Check if we are over the memory limit. */
2201 if (mem_used <= server.maxmemory) return REDIS_OK;
2202
2203 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2204 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2205
2206 /* Compute how much memory we need to free. */
2207 mem_tofree = mem_used - server.maxmemory;
2208 mem_freed = 0;
2209 while (mem_freed < mem_tofree) {
2210 int j, k, keys_freed = 0;
2211
2212 for (j = 0; j < server.dbnum; j++) {
2213 long bestval = 0; /* just to prevent warning */
2214 sds bestkey = NULL;
2215 struct dictEntry *de;
2216 redisDb *db = server.db+j;
2217 dict *dict;
2218
2219 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2220 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2221 {
2222 dict = server.db[j].dict;
2223 } else {
2224 dict = server.db[j].expires;
2225 }
2226 if (dictSize(dict) == 0) continue;
2227
2228 /* volatile-random and allkeys-random policy */
2229 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2230 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2231 {
2232 de = dictGetRandomKey(dict);
2233 bestkey = dictGetKey(de);
2234 }
2235
2236 /* volatile-lru and allkeys-lru policy */
2237 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2238 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2239 {
2240 for (k = 0; k < server.maxmemory_samples; k++) {
2241 sds thiskey;
2242 long thisval;
2243 robj *o;
2244
2245 de = dictGetRandomKey(dict);
2246 thiskey = dictGetKey(de);
2247 /* When policy is volatile-lru we need an additonal lookup
2248 * to locate the real key, as dict is set to db->expires. */
2249 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2250 de = dictFind(db->dict, thiskey);
2251 o = dictGetVal(de);
2252 thisval = estimateObjectIdleTime(o);
2253
2254 /* Higher idle time is better candidate for deletion */
2255 if (bestkey == NULL || thisval > bestval) {
2256 bestkey = thiskey;
2257 bestval = thisval;
2258 }
2259 }
2260 }
2261
2262 /* volatile-ttl */
2263 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2264 for (k = 0; k < server.maxmemory_samples; k++) {
2265 sds thiskey;
2266 long thisval;
2267
2268 de = dictGetRandomKey(dict);
2269 thiskey = dictGetKey(de);
2270 thisval = (long) dictGetVal(de);
2271
2272 /* Expire sooner (minor expire unix timestamp) is better
2273 * candidate for deletion */
2274 if (bestkey == NULL || thisval < bestval) {
2275 bestkey = thiskey;
2276 bestval = thisval;
2277 }
2278 }
2279 }
2280
2281 /* Finally remove the selected key. */
2282 if (bestkey) {
2283 long long delta;
2284
2285 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2286 propagateExpire(db,keyobj);
2287 /* We compute the amount of memory freed by dbDelete() alone.
2288 * It is possible that actually the memory needed to propagate
2289 * the DEL in AOF and replication link is greater than the one
2290 * we are freeing removing the key, but we can't account for
2291 * that otherwise we would never exit the loop.
2292 *
2293 * AOF and Output buffer memory will be freed eventually so
2294 * we only care about memory used by the key space. */
2295 delta = (long long) zmalloc_used_memory();
2296 dbDelete(db,keyobj);
2297 delta -= (long long) zmalloc_used_memory();
2298 mem_freed += delta;
2299 server.stat_evictedkeys++;
2300 decrRefCount(keyobj);
2301 keys_freed++;
2302
2303 /* When the memory to free starts to be big enough, we may
2304 * start spending so much time here that is impossible to
2305 * deliver data to the slaves fast enough, so we force the
2306 * transmission here inside the loop. */
2307 if (slaves) flushSlavesOutputBuffers();
2308 }
2309 }
2310 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2311 }
2312 return REDIS_OK;
2313 }
2314
2315 /* =================================== Main! ================================ */
2316
2317 #ifdef __linux__
2318 int linuxOvercommitMemoryValue(void) {
2319 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2320 char buf[64];
2321
2322 if (!fp) return -1;
2323 if (fgets(buf,64,fp) == NULL) {
2324 fclose(fp);
2325 return -1;
2326 }
2327 fclose(fp);
2328
2329 return atoi(buf);
2330 }
2331
2332 void linuxOvercommitMemoryWarning(void) {
2333 if (linuxOvercommitMemoryValue() == 0) {
2334 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.");
2335 }
2336 }
2337 #endif /* __linux__ */
2338
2339 void createPidFile(void) {
2340 /* Try to write the pid file in a best-effort way. */
2341 FILE *fp = fopen(server.pidfile,"w");
2342 if (fp) {
2343 fprintf(fp,"%d\n",(int)getpid());
2344 fclose(fp);
2345 }
2346 }
2347
2348 void daemonize(void) {
2349 int fd;
2350
2351 if (fork() != 0) exit(0); /* parent exits */
2352 setsid(); /* create a new session */
2353
2354 /* Every output goes to /dev/null. If Redis is daemonized but
2355 * the 'logfile' is set to 'stdout' in the configuration file
2356 * it will not log at all. */
2357 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2358 dup2(fd, STDIN_FILENO);
2359 dup2(fd, STDOUT_FILENO);
2360 dup2(fd, STDERR_FILENO);
2361 if (fd > STDERR_FILENO) close(fd);
2362 }
2363 }
2364
2365 void version() {
2366 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2367 REDIS_VERSION,
2368 redisGitSHA1(),
2369 atoi(redisGitDirty()) > 0,
2370 ZMALLOC_LIB,
2371 sizeof(long) == 4 ? 32 : 64);
2372 exit(0);
2373 }
2374
2375 void usage() {
2376 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2377 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2378 fprintf(stderr," ./redis-server -v or --version\n");
2379 fprintf(stderr," ./redis-server -h or --help\n");
2380 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2381 fprintf(stderr,"Examples:\n");
2382 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2383 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2384 fprintf(stderr," ./redis-server --port 7777\n");
2385 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2386 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2387 exit(1);
2388 }
2389
2390 void redisAsciiArt(void) {
2391 #include "asciilogo.h"
2392 char *buf = zmalloc(1024*16);
2393
2394 snprintf(buf,1024*16,ascii_logo,
2395 REDIS_VERSION,
2396 redisGitSHA1(),
2397 strtol(redisGitDirty(),NULL,10) > 0,
2398 (sizeof(long) == 8) ? "64" : "32",
2399 "stand alone",
2400 server.port,
2401 (long) getpid()
2402 );
2403 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2404 zfree(buf);
2405 }
2406
2407 static void sigtermHandler(int sig) {
2408 REDIS_NOTUSED(sig);
2409
2410 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2411 server.shutdown_asap = 1;
2412 }
2413
2414 void setupSignalHandlers(void) {
2415 struct sigaction act;
2416
2417 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2418 * Otherwise, sa_handler is used. */
2419 sigemptyset(&act.sa_mask);
2420 act.sa_flags = 0;
2421 act.sa_handler = sigtermHandler;
2422 sigaction(SIGTERM, &act, NULL);
2423
2424 #ifdef HAVE_BACKTRACE
2425 sigemptyset(&act.sa_mask);
2426 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2427 act.sa_sigaction = sigsegvHandler;
2428 sigaction(SIGSEGV, &act, NULL);
2429 sigaction(SIGBUS, &act, NULL);
2430 sigaction(SIGFPE, &act, NULL);
2431 sigaction(SIGILL, &act, NULL);
2432 #endif
2433 return;
2434 }
2435
2436 void memtest(size_t megabytes, int passes);
2437
2438 int main(int argc, char **argv) {
2439 long long start;
2440 struct timeval tv;
2441
2442 /* We need to initialize our libraries, and the server configuration. */
2443 zmalloc_enable_thread_safeness();
2444 srand(time(NULL)^getpid());
2445 gettimeofday(&tv,NULL);
2446 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2447 initServerConfig();
2448
2449 if (argc >= 2) {
2450 int j = 1; /* First option to parse in argv[] */
2451 sds options = sdsempty();
2452 char *configfile = NULL;
2453
2454 /* Handle special options --help and --version */
2455 if (strcmp(argv[1], "-v") == 0 ||
2456 strcmp(argv[1], "--version") == 0) version();
2457 if (strcmp(argv[1], "--help") == 0 ||
2458 strcmp(argv[1], "-h") == 0) usage();
2459 if (strcmp(argv[1], "--test-memory") == 0) {
2460 if (argc == 3) {
2461 memtest(atoi(argv[2]),50);
2462 exit(0);
2463 } else {
2464 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2465 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2466 exit(1);
2467 }
2468 }
2469
2470 /* First argument is the config file name? */
2471 if (argv[j][0] != '-' || argv[j][1] != '-')
2472 configfile = argv[j++];
2473 /* All the other options are parsed and conceptually appended to the
2474 * configuration file. For instance --port 6380 will generate the
2475 * string "port 6380\n" to be parsed after the actual file name
2476 * is parsed, if any. */
2477 while(j != argc) {
2478 if (argv[j][0] == '-' && argv[j][1] == '-') {
2479 /* Option name */
2480 if (sdslen(options)) options = sdscat(options,"\n");
2481 options = sdscat(options,argv[j]+2);
2482 options = sdscat(options," ");
2483 } else {
2484 /* Option argument */
2485 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2486 options = sdscat(options," ");
2487 }
2488 j++;
2489 }
2490 resetServerSaveParams();
2491 loadServerConfig(configfile,options);
2492 sdsfree(options);
2493 } else {
2494 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'");
2495 }
2496 if (server.daemonize) daemonize();
2497 initServer();
2498 if (server.daemonize) createPidFile();
2499 redisAsciiArt();
2500 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2501 #ifdef __linux__
2502 linuxOvercommitMemoryWarning();
2503 #endif
2504 start = ustime();
2505 if (server.aof_state == REDIS_AOF_ON) {
2506 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2507 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2508 } else {
2509 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2510 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2511 (float)(ustime()-start)/1000000);
2512 } else if (errno != ENOENT) {
2513 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2514 exit(1);
2515 }
2516 }
2517 if (server.ipfd > 0)
2518 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2519 if (server.sofd > 0)
2520 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2521 aeSetBeforeSleepProc(server.el,beforeSleep);
2522 aeMain(server.el);
2523 aeDeleteEventLoop(server.el);
2524 return 0;
2525 }
2526
2527 /* The End */