]> git.saurik.com Git - redis.git/blob - src/redis.c
0e80b3e7232a87cbad73a1efaeee60c699929ba8
[redis.git] / src / redis.c
1 /*
2 * Copyright (c) 2009-2010, Salvatore Sanfilippo <antirez at gmail dot com>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are met:
7 *
8 * * Redistributions of source code must retain the above copyright notice,
9 * this list of conditions and the following disclaimer.
10 * * Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * * Neither the name of Redis nor the names of its contributors may be used
14 * to endorse or promote products derived from this software without
15 * specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27 * POSSIBILITY OF SUCH DAMAGE.
28 */
29
30 #include "redis.h"
31 #include "slowlog.h"
32 #include "bio.h"
33
34 #include <time.h>
35 #include <signal.h>
36 #include <sys/wait.h>
37 #include <errno.h>
38 #include <assert.h>
39 #include <ctype.h>
40 #include <stdarg.h>
41 #include <arpa/inet.h>
42 #include <sys/stat.h>
43 #include <fcntl.h>
44 #include <sys/time.h>
45 #include <sys/resource.h>
46 #include <sys/uio.h>
47 #include <limits.h>
48 #include <float.h>
49 #include <math.h>
50 #include <sys/resource.h>
51 #include <sys/utsname.h>
52
53 /* Our shared "common" objects */
54
55 struct sharedObjectsStruct shared;
56
57 /* Global vars that are actually used as constants. The following double
58 * values are used for double on-disk serialization, and are initialized
59 * at runtime to avoid strange compiler optimizations. */
60
61 double R_Zero, R_PosInf, R_NegInf, R_Nan;
62
63 /*================================= Globals ================================= */
64
65 /* Global vars */
66 struct redisServer server; /* server global state */
67 struct redisCommand *commandTable;
68
69 /* Our command table.
70 *
71 * Every entry is composed of the following fields:
72 *
73 * name: a string representing the command name.
74 * function: pointer to the C function implementing the command.
75 * arity: number of arguments, it is possible to use -N to say >= N
76 * sflags: command flags as string. See below for a table of flags.
77 * flags: flags as bitmask. Computed by Redis using the 'sflags' field.
78 * get_keys_proc: an optional function to get key arguments from a command.
79 * This is only used when the following three fields are not
80 * enough to specify what arguments are keys.
81 * first_key_index: first argument that is a key
82 * last_key_index: last argument that is a key
83 * key_step: step to get all the keys from first to last argument. For instance
84 * in MSET the step is two since arguments are key,val,key,val,...
85 * microseconds: microseconds of total execution time for this command.
86 * calls: total number of calls of this command.
87 *
88 * The flags, microseconds and calls fields are computed by Redis and should
89 * always be set to zero.
90 *
91 * Command flags are expressed using strings where every character represents
92 * a flag. Later the populateCommandTable() function will take care of
93 * populating the real 'flags' field using this characters.
94 *
95 * This is the meaning of the flags:
96 *
97 * w: write command (may modify the key space).
98 * r: read command (will never modify the key space).
99 * m: may increase memory usage once called. Don't allow if out of memory.
100 * a: admin command, like SAVE or SHUTDOWN.
101 * p: Pub/Sub related command.
102 * f: force replication of this command, regarless of server.dirty.
103 * s: command not allowed in scripts.
104 * R: random command. Command is not deterministic, that is, the same command
105 * with the same arguments, with the same key space, may have different
106 * results. For instance SPOP and RANDOMKEY are two random commands.
107 * S: Sort command output array if called from script, so that the output
108 * is deterministic.
109 */
110 struct redisCommand redisCommandTable[] = {
111 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
112 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
113 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
114 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
115 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
116 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
117 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
118 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
119 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
120 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
121 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
122 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
123 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
124 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
125 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
126 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
127 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
128 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
129 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
130 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
131 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
132 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
133 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
134 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
135 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
136 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
137 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
138 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
139 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
140 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
141 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
142 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
143 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
144 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
145 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
146 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
147 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
148 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
149 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
150 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
151 {"srandmember",srandmemberCommand,2,"rR",0,NULL,1,1,1,0,0},
152 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
153 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
154 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
155 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
156 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
157 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
158 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
159 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
160 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
161 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
162 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
163 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
164 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
165 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
166 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
167 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
168 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
169 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
170 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
171 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
172 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
173 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
174 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
175 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
176 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
177 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
178 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
179 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
180 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
182 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
183 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
184 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
185 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
186 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
187 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
188 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
189 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
190 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
191 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
192 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
193 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
194 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
195 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
196 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
197 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
198 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
199 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
200 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
201 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
202 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
203 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
204 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
205 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
206 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
207 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
208 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
209 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
210 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
211 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
212 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
213 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
214 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
215 {"exec",execCommand,1,"s",0,NULL,0,0,0,0,0},
216 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
217 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
218 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
219 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
220 {"sort",sortCommand,-2,"wmS",0,NULL,1,1,1,0,0},
221 {"info",infoCommand,-1,"r",0,NULL,0,0,0,0,0},
222 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
223 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
224 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
225 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
226 {"slaveof",slaveofCommand,3,"as",0,NULL,0,0,0,0,0},
227 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
228 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
229 {"subscribe",subscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
230 {"unsubscribe",unsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
231 {"psubscribe",psubscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
232 {"punsubscribe",punsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
233 {"publish",publishCommand,3,"pf",0,NULL,0,0,0,0,0},
234 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
235 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
236 {"cluster",clusterCommand,-2,"ar",0,NULL,0,0,0,0,0},
237 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
238 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
239 {"asking",askingCommand,1,"r",0,NULL,0,0,0,0,0},
240 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
241 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
242 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
243 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
244 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
245 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
246 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
247 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0},
248 {"bitop",bitopCommand,-4,"wm",0,NULL,2,-1,1,0,0},
249 {"bitcount",bitcountCommand,-2,"r",0,NULL,1,1,1,0,0}
250 };
251
252 /*============================ Utility functions ============================ */
253
254 /* Low level logging. To use only for very big messages, otherwise
255 * redisLog() is to prefer. */
256 void redisLogRaw(int level, const char *msg) {
257 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
258 const char *c = ".-*#";
259 FILE *fp;
260 char buf[64];
261 int rawmode = (level & REDIS_LOG_RAW);
262
263 level &= 0xff; /* clear flags */
264 if (level < server.verbosity) return;
265
266 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
267 if (!fp) return;
268
269 if (rawmode) {
270 fprintf(fp,"%s",msg);
271 } else {
272 int off;
273 struct timeval tv;
274
275 gettimeofday(&tv,NULL);
276 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
277 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
278 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
279 }
280 fflush(fp);
281
282 if (server.logfile) fclose(fp);
283
284 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
285 }
286
287 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
288 * is used across the code. The raw version is only used in order to dump
289 * the INFO output on crash. */
290 void redisLog(int level, const char *fmt, ...) {
291 va_list ap;
292 char msg[REDIS_MAX_LOGMSG_LEN];
293
294 if ((level&0xff) < server.verbosity) return;
295
296 va_start(ap, fmt);
297 vsnprintf(msg, sizeof(msg), fmt, ap);
298 va_end(ap);
299
300 redisLogRaw(level,msg);
301 }
302
303 /* Log a fixed message without printf-alike capabilities, in a way that is
304 * safe to call from a signal handler.
305 *
306 * We actually use this only for signals that are not fatal from the point
307 * of view of Redis. Signals that are going to kill the server anyway and
308 * where we need printf-alike features are served by redisLog(). */
309 void redisLogFromHandler(int level, const char *msg) {
310 int fd;
311 char buf[64];
312
313 if ((level&0xff) < server.verbosity ||
314 (server.logfile == NULL && server.daemonize)) return;
315 fd = server.logfile ?
316 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
317 STDOUT_FILENO;
318 if (fd == -1) return;
319 ll2string(buf,sizeof(buf),getpid());
320 if (write(fd,"[",1) == -1) goto err;
321 if (write(fd,buf,strlen(buf)) == -1) goto err;
322 if (write(fd," | signal handler] (",20) == -1) goto err;
323 ll2string(buf,sizeof(buf),time(NULL));
324 if (write(fd,buf,strlen(buf)) == -1) goto err;
325 if (write(fd,") ",2) == -1) goto err;
326 if (write(fd,msg,strlen(msg)) == -1) goto err;
327 if (write(fd,"\n",1) == -1) goto err;
328 err:
329 if (server.logfile) close(fd);
330 }
331
332 /* Redis generally does not try to recover from out of memory conditions
333 * when allocating objects or strings, it is not clear if it will be possible
334 * to report this condition to the client since the networking layer itself
335 * is based on heap allocation for send buffers, so we simply abort.
336 * At least the code will be simpler to read... */
337 void oom(const char *msg) {
338 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
339 sleep(1);
340 abort();
341 }
342
343 /* Return the UNIX time in microseconds */
344 long long ustime(void) {
345 struct timeval tv;
346 long long ust;
347
348 gettimeofday(&tv, NULL);
349 ust = ((long long)tv.tv_sec)*1000000;
350 ust += tv.tv_usec;
351 return ust;
352 }
353
354 /* Return the UNIX time in milliseconds */
355 long long mstime(void) {
356 return ustime()/1000;
357 }
358
359 /* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
360 * exit(), because the latter may interact with the same file objects used by
361 * the parent process. However if we are testing the coverage normal exit() is
362 * used in order to obtain the right coverage information. */
363 void exitFromChild(int retcode) {
364 #ifdef COVERAGE_TEST
365 exit(retcode);
366 #else
367 _exit(retcode);
368 #endif
369 }
370
371 /*====================== Hash table type implementation ==================== */
372
373 /* This is an hash table type that uses the SDS dynamic strings libary as
374 * keys and radis objects as values (objects can hold SDS strings,
375 * lists, sets). */
376
377 void dictVanillaFree(void *privdata, void *val)
378 {
379 DICT_NOTUSED(privdata);
380 zfree(val);
381 }
382
383 void dictListDestructor(void *privdata, void *val)
384 {
385 DICT_NOTUSED(privdata);
386 listRelease((list*)val);
387 }
388
389 int dictSdsKeyCompare(void *privdata, const void *key1,
390 const void *key2)
391 {
392 int l1,l2;
393 DICT_NOTUSED(privdata);
394
395 l1 = sdslen((sds)key1);
396 l2 = sdslen((sds)key2);
397 if (l1 != l2) return 0;
398 return memcmp(key1, key2, l1) == 0;
399 }
400
401 /* A case insensitive version used for the command lookup table. */
402 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
403 const void *key2)
404 {
405 DICT_NOTUSED(privdata);
406
407 return strcasecmp(key1, key2) == 0;
408 }
409
410 void dictRedisObjectDestructor(void *privdata, void *val)
411 {
412 DICT_NOTUSED(privdata);
413
414 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
415 decrRefCount(val);
416 }
417
418 void dictSdsDestructor(void *privdata, void *val)
419 {
420 DICT_NOTUSED(privdata);
421
422 sdsfree(val);
423 }
424
425 int dictObjKeyCompare(void *privdata, const void *key1,
426 const void *key2)
427 {
428 const robj *o1 = key1, *o2 = key2;
429 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
430 }
431
432 unsigned int dictObjHash(const void *key) {
433 const robj *o = key;
434 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
435 }
436
437 unsigned int dictSdsHash(const void *key) {
438 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
439 }
440
441 unsigned int dictSdsCaseHash(const void *key) {
442 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
443 }
444
445 int dictEncObjKeyCompare(void *privdata, const void *key1,
446 const void *key2)
447 {
448 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
449 int cmp;
450
451 if (o1->encoding == REDIS_ENCODING_INT &&
452 o2->encoding == REDIS_ENCODING_INT)
453 return o1->ptr == o2->ptr;
454
455 o1 = getDecodedObject(o1);
456 o2 = getDecodedObject(o2);
457 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
458 decrRefCount(o1);
459 decrRefCount(o2);
460 return cmp;
461 }
462
463 unsigned int dictEncObjHash(const void *key) {
464 robj *o = (robj*) key;
465
466 if (o->encoding == REDIS_ENCODING_RAW) {
467 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
468 } else {
469 if (o->encoding == REDIS_ENCODING_INT) {
470 char buf[32];
471 int len;
472
473 len = ll2string(buf,32,(long)o->ptr);
474 return dictGenHashFunction((unsigned char*)buf, len);
475 } else {
476 unsigned int hash;
477
478 o = getDecodedObject(o);
479 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
480 decrRefCount(o);
481 return hash;
482 }
483 }
484 }
485
486 /* Sets type hash table */
487 dictType setDictType = {
488 dictEncObjHash, /* hash function */
489 NULL, /* key dup */
490 NULL, /* val dup */
491 dictEncObjKeyCompare, /* key compare */
492 dictRedisObjectDestructor, /* key destructor */
493 NULL /* val destructor */
494 };
495
496 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
497 dictType zsetDictType = {
498 dictEncObjHash, /* hash function */
499 NULL, /* key dup */
500 NULL, /* val dup */
501 dictEncObjKeyCompare, /* key compare */
502 dictRedisObjectDestructor, /* key destructor */
503 NULL /* val destructor */
504 };
505
506 /* Db->dict, keys are sds strings, vals are Redis objects. */
507 dictType dbDictType = {
508 dictSdsHash, /* hash function */
509 NULL, /* key dup */
510 NULL, /* val dup */
511 dictSdsKeyCompare, /* key compare */
512 dictSdsDestructor, /* key destructor */
513 dictRedisObjectDestructor /* val destructor */
514 };
515
516 /* Db->expires */
517 dictType keyptrDictType = {
518 dictSdsHash, /* hash function */
519 NULL, /* key dup */
520 NULL, /* val dup */
521 dictSdsKeyCompare, /* key compare */
522 NULL, /* key destructor */
523 NULL /* val destructor */
524 };
525
526 /* Command table. sds string -> command struct pointer. */
527 dictType commandTableDictType = {
528 dictSdsCaseHash, /* hash function */
529 NULL, /* key dup */
530 NULL, /* val dup */
531 dictSdsKeyCaseCompare, /* key compare */
532 dictSdsDestructor, /* key destructor */
533 NULL /* val destructor */
534 };
535
536 /* Hash type hash table (note that small hashes are represented with zimpaps) */
537 dictType hashDictType = {
538 dictEncObjHash, /* hash function */
539 NULL, /* key dup */
540 NULL, /* val dup */
541 dictEncObjKeyCompare, /* key compare */
542 dictRedisObjectDestructor, /* key destructor */
543 dictRedisObjectDestructor /* val destructor */
544 };
545
546 /* Keylist hash table type has unencoded redis objects as keys and
547 * lists as values. It's used for blocking operations (BLPOP) and to
548 * map swapped keys to a list of clients waiting for this keys to be loaded. */
549 dictType keylistDictType = {
550 dictObjHash, /* hash function */
551 NULL, /* key dup */
552 NULL, /* val dup */
553 dictObjKeyCompare, /* key compare */
554 dictRedisObjectDestructor, /* key destructor */
555 dictListDestructor /* val destructor */
556 };
557
558 /* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
559 * clusterNode structures. */
560 dictType clusterNodesDictType = {
561 dictSdsHash, /* hash function */
562 NULL, /* key dup */
563 NULL, /* val dup */
564 dictSdsKeyCompare, /* key compare */
565 dictSdsDestructor, /* key destructor */
566 NULL /* val destructor */
567 };
568
569 int htNeedsResize(dict *dict) {
570 long long size, used;
571
572 size = dictSlots(dict);
573 used = dictSize(dict);
574 return (size && used && size > DICT_HT_INITIAL_SIZE &&
575 (used*100/size < REDIS_HT_MINFILL));
576 }
577
578 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
579 * we resize the hash table to save memory */
580 void tryResizeHashTables(void) {
581 int j;
582
583 for (j = 0; j < server.dbnum; j++) {
584 if (htNeedsResize(server.db[j].dict))
585 dictResize(server.db[j].dict);
586 if (htNeedsResize(server.db[j].expires))
587 dictResize(server.db[j].expires);
588 }
589 }
590
591 /* Our hash table implementation performs rehashing incrementally while
592 * we write/read from the hash table. Still if the server is idle, the hash
593 * table will use two tables for a long time. So we try to use 1 millisecond
594 * of CPU time at every serverCron() loop in order to rehash some key. */
595 void incrementallyRehash(void) {
596 int j;
597
598 for (j = 0; j < server.dbnum; j++) {
599 /* Keys dictionary */
600 if (dictIsRehashing(server.db[j].dict)) {
601 dictRehashMilliseconds(server.db[j].dict,1);
602 break; /* already used our millisecond for this loop... */
603 }
604 /* Expires */
605 if (dictIsRehashing(server.db[j].expires)) {
606 dictRehashMilliseconds(server.db[j].expires,1);
607 break; /* already used our millisecond for this loop... */
608 }
609 }
610 }
611
612 /* This function is called once a background process of some kind terminates,
613 * as we want to avoid resizing the hash tables when there is a child in order
614 * to play well with copy-on-write (otherwise when a resize happens lots of
615 * memory pages are copied). The goal of this function is to update the ability
616 * for dict.c to resize the hash tables accordingly to the fact we have o not
617 * running childs. */
618 void updateDictResizePolicy(void) {
619 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
620 dictEnableResize();
621 else
622 dictDisableResize();
623 }
624
625 /* ======================= Cron: called every 100 ms ======================== */
626
627 /* Try to expire a few timed out keys. The algorithm used is adaptive and
628 * will use few CPU cycles if there are few expiring keys, otherwise
629 * it will get more aggressive to avoid that too much memory is used by
630 * keys that can be removed from the keyspace. */
631 void activeExpireCycle(void) {
632 int j, iteration = 0;
633 long long start = ustime(), timelimit;
634
635 /* We can use at max REDIS_EXPIRELOOKUPS_TIME_PERC percentage of CPU time
636 * per iteration. Since this function gets called with a frequency of
637 * REDIS_HZ times per second, the following is the max amount of
638 * microseconds we can spend in this function. */
639 timelimit = 1000000*REDIS_EXPIRELOOKUPS_TIME_PERC/REDIS_HZ/100;
640 if (timelimit <= 0) timelimit = 1;
641
642 for (j = 0; j < server.dbnum; j++) {
643 int expired;
644 redisDb *db = server.db+j;
645
646 /* Continue to expire if at the end of the cycle more than 25%
647 * of the keys were expired. */
648 do {
649 unsigned long num = dictSize(db->expires);
650 unsigned long slots = dictSlots(db->expires);
651 long long now = mstime();
652
653 /* When there are less than 1% filled slots getting random
654 * keys is expensive, so stop here waiting for better times...
655 * The dictionary will be resized asap. */
656 if (num && slots > DICT_HT_INITIAL_SIZE &&
657 (num*100/slots < 1)) break;
658
659 /* The main collection cycle. Sample random keys among keys
660 * with an expire set, checking for expired ones. */
661 expired = 0;
662 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
663 num = REDIS_EXPIRELOOKUPS_PER_CRON;
664 while (num--) {
665 dictEntry *de;
666 long long t;
667
668 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
669 t = dictGetSignedIntegerVal(de);
670 if (now > t) {
671 sds key = dictGetKey(de);
672 robj *keyobj = createStringObject(key,sdslen(key));
673
674 propagateExpire(db,keyobj);
675 dbDelete(db,keyobj);
676 decrRefCount(keyobj);
677 expired++;
678 server.stat_expiredkeys++;
679 }
680 }
681 /* We can't block forever here even if there are many keys to
682 * expire. So after a given amount of milliseconds return to the
683 * caller waiting for the other active expire cycle. */
684 iteration++;
685 if ((iteration & 0xf) == 0 && /* check once every 16 cycles. */
686 (ustime()-start) > timelimit) return;
687 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
688 }
689 }
690
691 void updateLRUClock(void) {
692 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
693 REDIS_LRU_CLOCK_MAX;
694 }
695
696
697 /* Add a sample to the operations per second array of samples. */
698 void trackOperationsPerSecond(void) {
699 long long t = mstime() - server.ops_sec_last_sample_time;
700 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
701 long long ops_sec;
702
703 ops_sec = t > 0 ? (ops*1000/t) : 0;
704
705 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
706 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
707 server.ops_sec_last_sample_time = mstime();
708 server.ops_sec_last_sample_ops = server.stat_numcommands;
709 }
710
711 /* Return the mean of all the samples. */
712 long long getOperationsPerSecond(void) {
713 int j;
714 long long sum = 0;
715
716 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
717 sum += server.ops_sec_samples[j];
718 return sum / REDIS_OPS_SEC_SAMPLES;
719 }
720
721 /* Check for timeouts. Returns non-zero if the client was terminated */
722 int clientsCronHandleTimeout(redisClient *c) {
723 time_t now = server.unixtime;
724
725 if (server.maxidletime &&
726 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
727 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
728 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
729 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
730 listLength(c->pubsub_patterns) == 0 &&
731 (now - c->lastinteraction > server.maxidletime))
732 {
733 redisLog(REDIS_VERBOSE,"Closing idle client");
734 freeClient(c);
735 return 1;
736 } else if (c->flags & REDIS_BLOCKED) {
737 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
738 addReply(c,shared.nullmultibulk);
739 unblockClientWaitingData(c);
740 }
741 }
742 return 0;
743 }
744
745 /* The client query buffer is an sds.c string that can end with a lot of
746 * free space not used, this function reclaims space if needed.
747 *
748 * The funciton always returns 0 as it never terminates the client. */
749 int clientsCronResizeQueryBuffer(redisClient *c) {
750 size_t querybuf_size = sdsAllocSize(c->querybuf);
751 time_t idletime = server.unixtime - c->lastinteraction;
752
753 /* There are two conditions to resize the query buffer:
754 * 1) Query buffer is > BIG_ARG and too big for latest peak.
755 * 2) Client is inactive and the buffer is bigger than 1k. */
756 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
757 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
758 (querybuf_size > 1024 && idletime > 2))
759 {
760 /* Only resize the query buffer if it is actually wasting space. */
761 if (sdsavail(c->querybuf) > 1024) {
762 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
763 }
764 }
765 /* Reset the peak again to capture the peak memory usage in the next
766 * cycle. */
767 c->querybuf_peak = 0;
768 return 0;
769 }
770
771 void clientsCron(void) {
772 /* Make sure to process at least 1/(REDIS_HZ*10) of clients per call.
773 * Since this function is called REDIS_HZ times per second we are sure that
774 * in the worst case we process all the clients in 10 seconds.
775 * In normal conditions (a reasonable number of clients) we process
776 * all the clients in a shorter time. */
777 int numclients = listLength(server.clients);
778 int iterations = numclients/(REDIS_HZ*10);
779
780 if (iterations < 50)
781 iterations = (numclients < 50) ? numclients : 50;
782 while(listLength(server.clients) && iterations--) {
783 redisClient *c;
784 listNode *head;
785
786 /* Rotate the list, take the current head, process.
787 * This way if the client must be removed from the list it's the
788 * first element and we don't incur into O(N) computation. */
789 listRotate(server.clients);
790 head = listFirst(server.clients);
791 c = listNodeValue(head);
792 /* The following functions do different service checks on the client.
793 * The protocol is that they return non-zero if the client was
794 * terminated. */
795 if (clientsCronHandleTimeout(c)) continue;
796 if (clientsCronResizeQueryBuffer(c)) continue;
797 }
798 }
799
800 /* This is our timer interrupt, called REDIS_HZ times per second.
801 * Here is where we do a number of things that need to be done asynchronously.
802 * For instance:
803 *
804 * - Active expired keys collection (it is also performed in a lazy way on
805 * lookup).
806 * - Software watchdong.
807 * - Update some statistic.
808 * - Incremental rehashing of the DBs hash tables.
809 * - Triggering BGSAVE / AOF rewrite, and handling of terminated children.
810 * - Clients timeout of differnet kinds.
811 * - Replication reconnection.
812 * - Many more...
813 *
814 * Everything directly called here will be called REDIS_HZ times per second,
815 * so in order to throttle execution of things we want to do less frequently
816 * a macro is used: run_with_period(milliseconds) { .... }
817 */
818
819 /* Using the following macro you can run code inside serverCron() with the
820 * specified period, specified in milliseconds.
821 * The actual resolution depends on REDIS_HZ. */
822 #define run_with_period(_ms_) if (!(loops % ((_ms_)/(1000/REDIS_HZ))))
823
824 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
825 int j, loops = server.cronloops;
826 REDIS_NOTUSED(eventLoop);
827 REDIS_NOTUSED(id);
828 REDIS_NOTUSED(clientData);
829
830 /* Software watchdog: deliver the SIGALRM that will reach the signal
831 * handler if we don't return here fast enough. */
832 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
833
834 /* We take a cached value of the unix time in the global state because
835 * with virtual memory and aging there is to store the current time
836 * in objects at every object access, and accuracy is not needed.
837 * To access a global var is faster than calling time(NULL) */
838 server.unixtime = time(NULL);
839
840 run_with_period(100) trackOperationsPerSecond();
841
842 /* We have just 22 bits per object for LRU information.
843 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
844 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
845 *
846 * Note that even if this will wrap after 1.5 years it's not a problem,
847 * everything will still work but just some object will appear younger
848 * to Redis. But for this to happen a given object should never be touched
849 * for 1.5 years.
850 *
851 * Note that you can change the resolution altering the
852 * REDIS_LRU_CLOCK_RESOLUTION define.
853 */
854 updateLRUClock();
855
856 /* Record the max memory used since the server was started. */
857 if (zmalloc_used_memory() > server.stat_peak_memory)
858 server.stat_peak_memory = zmalloc_used_memory();
859
860 /* We received a SIGTERM, shutting down here in a safe way, as it is
861 * not ok doing so inside the signal handler. */
862 if (server.shutdown_asap) {
863 if (prepareForShutdown(0) == REDIS_OK) exit(0);
864 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
865 }
866
867 /* Show some info about non-empty databases */
868 run_with_period(5000) {
869 for (j = 0; j < server.dbnum; j++) {
870 long long size, used, vkeys;
871
872 size = dictSlots(server.db[j].dict);
873 used = dictSize(server.db[j].dict);
874 vkeys = dictSize(server.db[j].expires);
875 if (used || vkeys) {
876 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
877 /* dictPrintStats(server.dict); */
878 }
879 }
880 }
881
882 /* We don't want to resize the hash tables while a bacground saving
883 * is in progress: the saving child is created using fork() that is
884 * implemented with a copy-on-write semantic in most modern systems, so
885 * if we resize the HT while there is the saving child at work actually
886 * a lot of memory movements in the parent will cause a lot of pages
887 * copied. */
888 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
889 tryResizeHashTables();
890 if (server.activerehashing) incrementallyRehash();
891 }
892
893 /* Show information about connected clients */
894 run_with_period(5000) {
895 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
896 listLength(server.clients)-listLength(server.slaves),
897 listLength(server.slaves),
898 zmalloc_used_memory());
899 }
900
901 /* We need to do a few operations on clients asynchronously. */
902 clientsCron();
903
904 /* Start a scheduled AOF rewrite if this was requested by the user while
905 * a BGSAVE was in progress. */
906 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
907 server.aof_rewrite_scheduled)
908 {
909 rewriteAppendOnlyFileBackground();
910 }
911
912 /* Check if a background saving or AOF rewrite in progress terminated. */
913 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
914 int statloc;
915 pid_t pid;
916
917 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
918 int exitcode = WEXITSTATUS(statloc);
919 int bysignal = 0;
920
921 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
922
923 if (pid == server.rdb_child_pid) {
924 backgroundSaveDoneHandler(exitcode,bysignal);
925 } else {
926 backgroundRewriteDoneHandler(exitcode,bysignal);
927 }
928 updateDictResizePolicy();
929 }
930 } else {
931 /* If there is not a background saving/rewrite in progress check if
932 * we have to save/rewrite now */
933 for (j = 0; j < server.saveparamslen; j++) {
934 struct saveparam *sp = server.saveparams+j;
935
936 if (server.dirty >= sp->changes &&
937 server.unixtime-server.lastsave > sp->seconds) {
938 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
939 sp->changes, sp->seconds);
940 rdbSaveBackground(server.rdb_filename);
941 break;
942 }
943 }
944
945 /* Trigger an AOF rewrite if needed */
946 if (server.rdb_child_pid == -1 &&
947 server.aof_child_pid == -1 &&
948 server.aof_rewrite_perc &&
949 server.aof_current_size > server.aof_rewrite_min_size)
950 {
951 long long base = server.aof_rewrite_base_size ?
952 server.aof_rewrite_base_size : 1;
953 long long growth = (server.aof_current_size*100/base) - 100;
954 if (growth >= server.aof_rewrite_perc) {
955 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
956 rewriteAppendOnlyFileBackground();
957 }
958 }
959 }
960
961
962 /* If we postponed an AOF buffer flush, let's try to do it every time the
963 * cron function is called. */
964 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
965
966 /* Expire a few keys per cycle, only if this is a master.
967 * On slaves we wait for DEL operations synthesized by the master
968 * in order to guarantee a strict consistency. */
969 if (server.masterhost == NULL) activeExpireCycle();
970
971 /* Close clients that need to be closed asynchronous */
972 freeClientsInAsyncFreeQueue();
973
974 /* Replication cron function -- used to reconnect to master and
975 * to detect transfer failures. */
976 run_with_period(1000) replicationCron();
977
978 /* Run other sub-systems specific cron jobs */
979 run_with_period(1000) {
980 if (server.cluster_enabled) clusterCron();
981 }
982
983 server.cronloops++;
984 return 1000/REDIS_HZ;
985 }
986
987 /* This function gets called every time Redis is entering the
988 * main loop of the event driven library, that is, before to sleep
989 * for ready file descriptors. */
990 void beforeSleep(struct aeEventLoop *eventLoop) {
991 REDIS_NOTUSED(eventLoop);
992 listNode *ln;
993 redisClient *c;
994
995 /* Try to process pending commands for clients that were just unblocked. */
996 while (listLength(server.unblocked_clients)) {
997 ln = listFirst(server.unblocked_clients);
998 redisAssert(ln != NULL);
999 c = ln->value;
1000 listDelNode(server.unblocked_clients,ln);
1001 c->flags &= ~REDIS_UNBLOCKED;
1002
1003 /* Process remaining data in the input buffer. */
1004 if (c->querybuf && sdslen(c->querybuf) > 0) {
1005 server.current_client = c;
1006 processInputBuffer(c);
1007 server.current_client = NULL;
1008 }
1009 }
1010
1011 /* Write the AOF buffer on disk */
1012 flushAppendOnlyFile(0);
1013 }
1014
1015 /* =========================== Server initialization ======================== */
1016
1017 void createSharedObjects(void) {
1018 int j;
1019
1020 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
1021 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
1022 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
1023 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
1024 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
1025 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
1026 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
1027 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
1028 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
1029 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
1030 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
1031 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
1032 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
1033 "-ERR Operation against a key holding the wrong kind of value\r\n"));
1034 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
1035 "-ERR no such key\r\n"));
1036 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
1037 "-ERR syntax error\r\n"));
1038 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
1039 "-ERR source and destination objects are the same\r\n"));
1040 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
1041 "-ERR index out of range\r\n"));
1042 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
1043 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
1044 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
1045 "-LOADING Redis is loading the dataset in memory\r\n"));
1046 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
1047 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
1048 shared.masterdownerr = createObject(REDIS_STRING,sdsnew(
1049 "-MASTERDOWN Link with MASTER is down and slave-serve-stale-data is set to 'no'.\r\n"));
1050 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
1051 "-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"));
1052 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
1053 "-READONLY You can't write against a read only slave.\r\n"));
1054 shared.oomerr = createObject(REDIS_STRING,sdsnew(
1055 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
1056 shared.space = createObject(REDIS_STRING,sdsnew(" "));
1057 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1058 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1059
1060 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1061 shared.select[j] = createObject(REDIS_STRING,
1062 sdscatprintf(sdsempty(),"select %d\r\n", j));
1063 }
1064 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1065 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1066 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1067 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1068 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1069 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1070 shared.del = createStringObject("DEL",3);
1071 shared.rpop = createStringObject("RPOP",4);
1072 shared.lpop = createStringObject("LPOP",4);
1073 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1074 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1075 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1076 }
1077 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1078 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1079 sdscatprintf(sdsempty(),"*%d\r\n",j));
1080 shared.bulkhdr[j] = createObject(REDIS_STRING,
1081 sdscatprintf(sdsempty(),"$%d\r\n",j));
1082 }
1083 }
1084
1085 void initServerConfig() {
1086 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1087 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1088 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1089 server.port = REDIS_SERVERPORT;
1090 server.bindaddr = NULL;
1091 server.unixsocket = NULL;
1092 server.unixsocketperm = 0;
1093 server.ipfd = -1;
1094 server.sofd = -1;
1095 server.dbnum = REDIS_DEFAULT_DBNUM;
1096 server.verbosity = REDIS_NOTICE;
1097 server.maxidletime = REDIS_MAXIDLETIME;
1098 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1099 server.saveparams = NULL;
1100 server.loading = 0;
1101 server.logfile = NULL; /* NULL = log on standard output */
1102 server.syslog_enabled = 0;
1103 server.syslog_ident = zstrdup("redis");
1104 server.syslog_facility = LOG_LOCAL0;
1105 server.daemonize = 0;
1106 server.aof_state = REDIS_AOF_OFF;
1107 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1108 server.aof_no_fsync_on_rewrite = 0;
1109 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1110 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1111 server.aof_rewrite_base_size = 0;
1112 server.aof_rewrite_scheduled = 0;
1113 server.aof_last_fsync = time(NULL);
1114 server.aof_rewrite_time_last = -1;
1115 server.aof_rewrite_time_start = -1;
1116 server.aof_delayed_fsync = 0;
1117 server.aof_fd = -1;
1118 server.aof_selected_db = -1; /* Make sure the first time will not match */
1119 server.aof_flush_postponed_start = 0;
1120 server.pidfile = zstrdup("/var/run/redis.pid");
1121 server.rdb_filename = zstrdup("dump.rdb");
1122 server.aof_filename = zstrdup("appendonly.aof");
1123 server.requirepass = NULL;
1124 server.rdb_compression = 1;
1125 server.rdb_checksum = 1;
1126 server.activerehashing = 1;
1127 server.maxclients = REDIS_MAX_CLIENTS;
1128 server.bpop_blocked_clients = 0;
1129 server.maxmemory = 0;
1130 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1131 server.maxmemory_samples = 3;
1132 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1133 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1134 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1135 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1136 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1137 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1138 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1139 server.shutdown_asap = 0;
1140 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1141 server.repl_timeout = REDIS_REPL_TIMEOUT;
1142 server.cluster_enabled = 0;
1143 server.cluster.configfile = zstrdup("nodes.conf");
1144 server.lua_caller = NULL;
1145 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1146 server.lua_client = NULL;
1147 server.lua_timedout = 0;
1148
1149 updateLRUClock();
1150 resetServerSaveParams();
1151
1152 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1153 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1154 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1155 /* Replication related */
1156 server.masterauth = NULL;
1157 server.masterhost = NULL;
1158 server.masterport = 6379;
1159 server.master = NULL;
1160 server.repl_state = REDIS_REPL_NONE;
1161 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1162 server.repl_serve_stale_data = 1;
1163 server.repl_slave_ro = 1;
1164 server.repl_down_since = time(NULL);
1165
1166 /* Client output buffer limits */
1167 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1168 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1169 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1170 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1171 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1172 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1173 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1174 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1175 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1176
1177 /* Double constants initialization */
1178 R_Zero = 0.0;
1179 R_PosInf = 1.0/R_Zero;
1180 R_NegInf = -1.0/R_Zero;
1181 R_Nan = R_Zero/R_Zero;
1182
1183 /* Command table -- we intiialize it here as it is part of the
1184 * initial configuration, since command names may be changed via
1185 * redis.conf using the rename-command directive. */
1186 server.commands = dictCreate(&commandTableDictType,NULL);
1187 populateCommandTable();
1188 server.delCommand = lookupCommandByCString("del");
1189 server.multiCommand = lookupCommandByCString("multi");
1190 server.lpushCommand = lookupCommandByCString("lpush");
1191
1192 /* Slow log */
1193 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1194 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1195
1196 /* Debugging */
1197 server.assert_failed = "<no assertion failed>";
1198 server.assert_file = "<no file>";
1199 server.assert_line = 0;
1200 server.bug_report_start = 0;
1201 server.watchdog_period = 0;
1202 }
1203
1204 /* This function will try to raise the max number of open files accordingly to
1205 * the configured max number of clients. It will also account for 32 additional
1206 * file descriptors as we need a few more for persistence, listening
1207 * sockets, log files and so forth.
1208 *
1209 * If it will not be possible to set the limit accordingly to the configured
1210 * max number of clients, the function will do the reverse setting
1211 * server.maxclients to the value that we can actually handle. */
1212 void adjustOpenFilesLimit(void) {
1213 rlim_t maxfiles = server.maxclients+32;
1214 struct rlimit limit;
1215
1216 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1217 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1218 strerror(errno));
1219 server.maxclients = 1024-32;
1220 } else {
1221 rlim_t oldlimit = limit.rlim_cur;
1222
1223 /* Set the max number of files if the current limit is not enough
1224 * for our needs. */
1225 if (oldlimit < maxfiles) {
1226 rlim_t f;
1227
1228 f = maxfiles;
1229 while(f > oldlimit) {
1230 limit.rlim_cur = f;
1231 limit.rlim_max = f;
1232 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1233 f -= 128;
1234 }
1235 if (f < oldlimit) f = oldlimit;
1236 if (f != maxfiles) {
1237 server.maxclients = f-32;
1238 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1239 (int) maxfiles, strerror(errno), (int) server.maxclients);
1240 } else {
1241 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1242 (int) maxfiles);
1243 }
1244 }
1245 }
1246 }
1247
1248 void initServer() {
1249 int j;
1250
1251 signal(SIGHUP, SIG_IGN);
1252 signal(SIGPIPE, SIG_IGN);
1253 setupSignalHandlers();
1254
1255 if (server.syslog_enabled) {
1256 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1257 server.syslog_facility);
1258 }
1259
1260 server.current_client = NULL;
1261 server.clients = listCreate();
1262 server.clients_to_close = listCreate();
1263 server.slaves = listCreate();
1264 server.monitors = listCreate();
1265 server.unblocked_clients = listCreate();
1266
1267 createSharedObjects();
1268 adjustOpenFilesLimit();
1269 server.el = aeCreateEventLoop(server.maxclients+1024);
1270 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1271
1272 if (server.port != 0) {
1273 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1274 if (server.ipfd == ANET_ERR) {
1275 redisLog(REDIS_WARNING, "Opening port %d: %s",
1276 server.port, server.neterr);
1277 exit(1);
1278 }
1279 }
1280 if (server.unixsocket != NULL) {
1281 unlink(server.unixsocket); /* don't care if this fails */
1282 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1283 if (server.sofd == ANET_ERR) {
1284 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1285 exit(1);
1286 }
1287 }
1288 if (server.ipfd < 0 && server.sofd < 0) {
1289 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1290 exit(1);
1291 }
1292 for (j = 0; j < server.dbnum; j++) {
1293 server.db[j].dict = dictCreate(&dbDictType,NULL);
1294 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1295 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1296 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1297 server.db[j].id = j;
1298 }
1299 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1300 server.pubsub_patterns = listCreate();
1301 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1302 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1303 server.cronloops = 0;
1304 server.rdb_child_pid = -1;
1305 server.aof_child_pid = -1;
1306 aofRewriteBufferReset();
1307 server.aof_buf = sdsempty();
1308 server.lastsave = time(NULL);
1309 server.rdb_save_time_last = -1;
1310 server.rdb_save_time_start = -1;
1311 server.dirty = 0;
1312 server.stat_numcommands = 0;
1313 server.stat_numconnections = 0;
1314 server.stat_expiredkeys = 0;
1315 server.stat_evictedkeys = 0;
1316 server.stat_starttime = time(NULL);
1317 server.stat_keyspace_misses = 0;
1318 server.stat_keyspace_hits = 0;
1319 server.stat_peak_memory = 0;
1320 server.stat_fork_time = 0;
1321 server.stat_rejected_conn = 0;
1322 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1323 server.ops_sec_idx = 0;
1324 server.ops_sec_last_sample_time = mstime();
1325 server.ops_sec_last_sample_ops = 0;
1326 server.unixtime = time(NULL);
1327 server.lastbgsave_status = REDIS_OK;
1328 server.stop_writes_on_bgsave_err = 1;
1329 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1330 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1331 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1332 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1333 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1334
1335 if (server.aof_state == REDIS_AOF_ON) {
1336 server.aof_fd = open(server.aof_filename,
1337 O_WRONLY|O_APPEND|O_CREAT,0644);
1338 if (server.aof_fd == -1) {
1339 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1340 strerror(errno));
1341 exit(1);
1342 }
1343 }
1344
1345 /* 32 bit instances are limited to 4GB of address space, so if there is
1346 * no explicit limit in the user provided configuration we set a limit
1347 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1348 * useless crashes of the Redis instance. */
1349 if (server.arch_bits == 32 && server.maxmemory == 0) {
1350 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1351 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1352 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1353 }
1354
1355 if (server.cluster_enabled) clusterInit();
1356 scriptingInit();
1357 slowlogInit();
1358 bioInit();
1359 }
1360
1361 /* Populates the Redis Command Table starting from the hard coded list
1362 * we have on top of redis.c file. */
1363 void populateCommandTable(void) {
1364 int j;
1365 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1366
1367 for (j = 0; j < numcommands; j++) {
1368 struct redisCommand *c = redisCommandTable+j;
1369 char *f = c->sflags;
1370 int retval;
1371
1372 while(*f != '\0') {
1373 switch(*f) {
1374 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1375 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1376 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1377 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1378 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1379 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1380 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1381 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1382 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1383 default: redisPanic("Unsupported command flag"); break;
1384 }
1385 f++;
1386 }
1387
1388 retval = dictAdd(server.commands, sdsnew(c->name), c);
1389 assert(retval == DICT_OK);
1390 }
1391 }
1392
1393 void resetCommandTableStats(void) {
1394 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1395 int j;
1396
1397 for (j = 0; j < numcommands; j++) {
1398 struct redisCommand *c = redisCommandTable+j;
1399
1400 c->microseconds = 0;
1401 c->calls = 0;
1402 }
1403 }
1404
1405 /* ========================== Redis OP Array API ============================ */
1406
1407 void redisOpArrayInit(redisOpArray *oa) {
1408 oa->ops = NULL;
1409 oa->numops = 0;
1410 }
1411
1412 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1413 robj **argv, int argc, int target)
1414 {
1415 redisOp *op;
1416
1417 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1418 op = oa->ops+oa->numops;
1419 op->cmd = cmd;
1420 op->dbid = dbid;
1421 op->argv = argv;
1422 op->argc = argc;
1423 op->target = target;
1424 oa->numops++;
1425 return oa->numops;
1426 }
1427
1428 void redisOpArrayFree(redisOpArray *oa) {
1429 while(oa->numops) {
1430 int j;
1431 redisOp *op;
1432
1433 oa->numops--;
1434 op = oa->ops+oa->numops;
1435 for (j = 0; j < op->argc; j++)
1436 decrRefCount(op->argv[j]);
1437 zfree(op->argv);
1438 }
1439 zfree(oa->ops);
1440 }
1441
1442 /* ====================== Commands lookup and execution ===================== */
1443
1444 struct redisCommand *lookupCommand(sds name) {
1445 return dictFetchValue(server.commands, name);
1446 }
1447
1448 struct redisCommand *lookupCommandByCString(char *s) {
1449 struct redisCommand *cmd;
1450 sds name = sdsnew(s);
1451
1452 cmd = dictFetchValue(server.commands, name);
1453 sdsfree(name);
1454 return cmd;
1455 }
1456
1457 /* Propagate the specified command (in the context of the specified database id)
1458 * to AOF, Slaves and Monitors.
1459 *
1460 * flags are an xor between:
1461 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1462 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1463 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1464 */
1465 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1466 int flags)
1467 {
1468 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1469 feedAppendOnlyFile(cmd,dbid,argv,argc);
1470 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1471 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1472 }
1473
1474 /* Used inside commands to schedule the propagation of additional commands
1475 * after the current command is propagated to AOF / Replication. */
1476 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1477 int target)
1478 {
1479 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1480 }
1481
1482 /* Call() is the core of Redis execution of a command */
1483 void call(redisClient *c, int flags) {
1484 long long dirty, start = ustime(), duration;
1485
1486 /* Sent the command to clients in MONITOR mode, only if the commands are
1487 * not geneated from reading an AOF. */
1488 if (listLength(server.monitors) && !server.loading)
1489 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1490
1491 /* Call the command. */
1492 redisOpArrayInit(&server.also_propagate);
1493 dirty = server.dirty;
1494 c->cmd->proc(c);
1495 dirty = server.dirty-dirty;
1496 duration = ustime()-start;
1497
1498 /* When EVAL is called loading the AOF we don't want commands called
1499 * from Lua to go into the slowlog or to populate statistics. */
1500 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1501 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1502
1503 /* Log the command into the Slow log if needed, and populate the
1504 * per-command statistics that we show in INFO commandstats. */
1505 if (flags & REDIS_CALL_SLOWLOG)
1506 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1507 if (flags & REDIS_CALL_STATS) {
1508 c->cmd->microseconds += duration;
1509 c->cmd->calls++;
1510 }
1511
1512 /* Propagate the command into the AOF and replication link */
1513 if (flags & REDIS_CALL_PROPAGATE) {
1514 int flags = REDIS_PROPAGATE_NONE;
1515
1516 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1517 flags |= REDIS_PROPAGATE_REPL;
1518 if (dirty)
1519 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1520 if (flags != REDIS_PROPAGATE_NONE)
1521 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1522 }
1523 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1524 * PUSH command. */
1525 if (server.also_propagate.numops) {
1526 int j;
1527 redisOp *rop;
1528
1529 for (j = 0; j < server.also_propagate.numops; j++) {
1530 rop = &server.also_propagate.ops[j];
1531 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1532 }
1533 redisOpArrayFree(&server.also_propagate);
1534 }
1535 server.stat_numcommands++;
1536 }
1537
1538 /* If this function gets called we already read a whole
1539 * command, argments are in the client argv/argc fields.
1540 * processCommand() execute the command or prepare the
1541 * server for a bulk read from the client.
1542 *
1543 * If 1 is returned the client is still alive and valid and
1544 * and other operations can be performed by the caller. Otherwise
1545 * if 0 is returned the client was destroied (i.e. after QUIT). */
1546 int processCommand(redisClient *c) {
1547 /* The QUIT command is handled separately. Normal command procs will
1548 * go through checking for replication and QUIT will cause trouble
1549 * when FORCE_REPLICATION is enabled and would be implemented in
1550 * a regular command proc. */
1551 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1552 addReply(c,shared.ok);
1553 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1554 return REDIS_ERR;
1555 }
1556
1557 /* Now lookup the command and check ASAP about trivial error conditions
1558 * such as wrong arity, bad command name and so forth. */
1559 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1560 if (!c->cmd) {
1561 addReplyErrorFormat(c,"unknown command '%s'",
1562 (char*)c->argv[0]->ptr);
1563 return REDIS_OK;
1564 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1565 (c->argc < -c->cmd->arity)) {
1566 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1567 c->cmd->name);
1568 return REDIS_OK;
1569 }
1570
1571 /* Check if the user is authenticated */
1572 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1573 {
1574 addReplyError(c,"operation not permitted");
1575 return REDIS_OK;
1576 }
1577
1578 /* If cluster is enabled, redirect here */
1579 if (server.cluster_enabled &&
1580 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1581 int hashslot;
1582
1583 if (server.cluster.state != REDIS_CLUSTER_OK) {
1584 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1585 return REDIS_OK;
1586 } else {
1587 int ask;
1588 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1589 if (n == NULL) {
1590 addReplyError(c,"Multi keys request invalid in cluster");
1591 return REDIS_OK;
1592 } else if (n != server.cluster.myself) {
1593 addReplySds(c,sdscatprintf(sdsempty(),
1594 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1595 hashslot,n->ip,n->port));
1596 return REDIS_OK;
1597 }
1598 }
1599 }
1600
1601 /* Handle the maxmemory directive.
1602 *
1603 * First we try to free some memory if possible (if there are volatile
1604 * keys in the dataset). If there are not the only thing we can do
1605 * is returning an error. */
1606 if (server.maxmemory) {
1607 int retval = freeMemoryIfNeeded();
1608 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1609 addReply(c, shared.oomerr);
1610 return REDIS_OK;
1611 }
1612 }
1613
1614 /* Don't accept write commands if there are problems persisting on disk. */
1615 if (server.stop_writes_on_bgsave_err &&
1616 server.saveparamslen > 0
1617 && server.lastbgsave_status == REDIS_ERR &&
1618 c->cmd->flags & REDIS_CMD_WRITE)
1619 {
1620 addReply(c, shared.bgsaveerr);
1621 return REDIS_OK;
1622 }
1623
1624 /* Don't accept wirte commands if this is a read only slave. But
1625 * accept write commands if this is our master. */
1626 if (server.masterhost && server.repl_slave_ro &&
1627 !(c->flags & REDIS_MASTER) &&
1628 c->cmd->flags & REDIS_CMD_WRITE)
1629 {
1630 addReply(c, shared.roslaveerr);
1631 return REDIS_OK;
1632 }
1633
1634 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1635 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1636 &&
1637 c->cmd->proc != subscribeCommand &&
1638 c->cmd->proc != unsubscribeCommand &&
1639 c->cmd->proc != psubscribeCommand &&
1640 c->cmd->proc != punsubscribeCommand) {
1641 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1642 return REDIS_OK;
1643 }
1644
1645 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1646 * we are a slave with a broken link with master. */
1647 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1648 server.repl_serve_stale_data == 0 &&
1649 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1650 {
1651 addReply(c, shared.masterdownerr);
1652 return REDIS_OK;
1653 }
1654
1655 /* Loading DB? Return an error if the command is not INFO */
1656 if (server.loading && c->cmd->proc != infoCommand) {
1657 addReply(c, shared.loadingerr);
1658 return REDIS_OK;
1659 }
1660
1661 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1662 if (server.lua_timedout &&
1663 !(c->cmd->proc == shutdownCommand &&
1664 c->argc == 2 &&
1665 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1666 !(c->cmd->proc == scriptCommand &&
1667 c->argc == 2 &&
1668 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1669 {
1670 addReply(c, shared.slowscripterr);
1671 return REDIS_OK;
1672 }
1673
1674 /* Exec the command */
1675 if (c->flags & REDIS_MULTI &&
1676 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1677 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1678 {
1679 queueMultiCommand(c);
1680 addReply(c,shared.queued);
1681 } else {
1682 call(c,REDIS_CALL_FULL);
1683 }
1684 return REDIS_OK;
1685 }
1686
1687 /*================================== Shutdown =============================== */
1688
1689 int prepareForShutdown(int flags) {
1690 int save = flags & REDIS_SHUTDOWN_SAVE;
1691 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1692
1693 redisLog(REDIS_WARNING,"User requested shutdown...");
1694 /* Kill the saving child if there is a background saving in progress.
1695 We want to avoid race conditions, for instance our saving child may
1696 overwrite the synchronous saving did by SHUTDOWN. */
1697 if (server.rdb_child_pid != -1) {
1698 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1699 kill(server.rdb_child_pid,SIGKILL);
1700 rdbRemoveTempFile(server.rdb_child_pid);
1701 }
1702 if (server.aof_state != REDIS_AOF_OFF) {
1703 /* Kill the AOF saving child as the AOF we already have may be longer
1704 * but contains the full dataset anyway. */
1705 if (server.aof_child_pid != -1) {
1706 redisLog(REDIS_WARNING,
1707 "There is a child rewriting the AOF. Killing it!");
1708 kill(server.aof_child_pid,SIGKILL);
1709 }
1710 /* Append only file: fsync() the AOF and exit */
1711 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1712 aof_fsync(server.aof_fd);
1713 }
1714 if ((server.saveparamslen > 0 && !nosave) || save) {
1715 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1716 /* Snapshotting. Perform a SYNC SAVE and exit */
1717 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1718 /* Ooops.. error saving! The best we can do is to continue
1719 * operating. Note that if there was a background saving process,
1720 * in the next cron() Redis will be notified that the background
1721 * saving aborted, handling special stuff like slaves pending for
1722 * synchronization... */
1723 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1724 return REDIS_ERR;
1725 }
1726 }
1727 if (server.daemonize) {
1728 redisLog(REDIS_NOTICE,"Removing the pid file.");
1729 unlink(server.pidfile);
1730 }
1731 /* Close the listening sockets. Apparently this allows faster restarts. */
1732 if (server.ipfd != -1) close(server.ipfd);
1733 if (server.sofd != -1) close(server.sofd);
1734 if (server.unixsocket) {
1735 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1736 unlink(server.unixsocket); /* don't care if this fails */
1737 }
1738
1739 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1740 return REDIS_OK;
1741 }
1742
1743 /*================================== Commands =============================== */
1744
1745 /* Return 0 if strings are the same, 1 if they are not.
1746 * The comparison is performed in a way that prevents an attacker to obtain
1747 * information about the nature of the strings just monitoring the execution
1748 * time of the function.
1749 *
1750 * Note that limiting the comparison length to strings up to 512 bytes we
1751 * can avoid leaking any information about the password length and any
1752 * possible branch misprediction related leak.
1753 */
1754 int time_independent_strcmp(char *a, char *b) {
1755 char bufa[REDIS_AUTHPASS_MAX_LEN], bufb[REDIS_AUTHPASS_MAX_LEN];
1756 /* The above two strlen perform len(a) + len(b) operations where either
1757 * a or b are fixed (our password) length, and the difference is only
1758 * relative to the length of the user provided string, so no information
1759 * leak is possible in the following two lines of code. */
1760 int alen = strlen(a);
1761 int blen = strlen(b);
1762 int j;
1763 int diff = 0;
1764
1765 /* We can't compare strings longer than our static buffers.
1766 * Note that this will never pass the first test in practical circumstances
1767 * so there is no info leak. */
1768 if (alen > sizeof(bufa) || blen > sizeof(bufb)) return 1;
1769
1770 memset(bufa,0,sizeof(bufa)); /* Constant time. */
1771 memset(bufb,0,sizeof(bufb)); /* Constant time. */
1772 /* Again the time of the following two copies is proportional to
1773 * len(a) + len(b) so no info is leaked. */
1774 memcpy(bufa,a,alen);
1775 memcpy(bufb,b,blen);
1776
1777 /* Always compare all the chars in the two buffers without
1778 * conditional expressions. */
1779 for (j = 0; j < sizeof(bufa); j++) {
1780 diff |= (bufa[j] ^ bufb[j]);
1781 }
1782 /* Length must be equal as well. */
1783 diff |= alen ^ blen;
1784 return diff; /* If zero strings are the same. */
1785 }
1786
1787 void authCommand(redisClient *c) {
1788 if (!server.requirepass) {
1789 addReplyError(c,"Client sent AUTH, but no password is set");
1790 } else if (!time_independent_strcmp(c->argv[1]->ptr, server.requirepass)) {
1791 c->authenticated = 1;
1792 addReply(c,shared.ok);
1793 } else {
1794 c->authenticated = 0;
1795 addReplyError(c,"invalid password");
1796 }
1797 }
1798
1799 void pingCommand(redisClient *c) {
1800 addReply(c,shared.pong);
1801 }
1802
1803 void echoCommand(redisClient *c) {
1804 addReplyBulk(c,c->argv[1]);
1805 }
1806
1807 void timeCommand(redisClient *c) {
1808 struct timeval tv;
1809
1810 /* gettimeofday() can only fail if &tv is a bad addresss so we
1811 * don't check for errors. */
1812 gettimeofday(&tv,NULL);
1813 addReplyMultiBulkLen(c,2);
1814 addReplyBulkLongLong(c,tv.tv_sec);
1815 addReplyBulkLongLong(c,tv.tv_usec);
1816 }
1817
1818 /* Convert an amount of bytes into a human readable string in the form
1819 * of 100B, 2G, 100M, 4K, and so forth. */
1820 void bytesToHuman(char *s, unsigned long long n) {
1821 double d;
1822
1823 if (n < 1024) {
1824 /* Bytes */
1825 sprintf(s,"%lluB",n);
1826 return;
1827 } else if (n < (1024*1024)) {
1828 d = (double)n/(1024);
1829 sprintf(s,"%.2fK",d);
1830 } else if (n < (1024LL*1024*1024)) {
1831 d = (double)n/(1024*1024);
1832 sprintf(s,"%.2fM",d);
1833 } else if (n < (1024LL*1024*1024*1024)) {
1834 d = (double)n/(1024LL*1024*1024);
1835 sprintf(s,"%.2fG",d);
1836 }
1837 }
1838
1839 /* Create the string returned by the INFO command. This is decoupled
1840 * by the INFO command itself as we need to report the same information
1841 * on memory corruption problems. */
1842 sds genRedisInfoString(char *section) {
1843 sds info = sdsempty();
1844 time_t uptime = server.unixtime-server.stat_starttime;
1845 int j, numcommands;
1846 struct rusage self_ru, c_ru;
1847 unsigned long lol, bib;
1848 int allsections = 0, defsections = 0;
1849 int sections = 0;
1850
1851 if (section) {
1852 allsections = strcasecmp(section,"all") == 0;
1853 defsections = strcasecmp(section,"default") == 0;
1854 }
1855
1856 getrusage(RUSAGE_SELF, &self_ru);
1857 getrusage(RUSAGE_CHILDREN, &c_ru);
1858 getClientsMaxBuffers(&lol,&bib);
1859
1860 /* Server */
1861 if (allsections || defsections || !strcasecmp(section,"server")) {
1862 struct utsname name;
1863
1864 if (sections++) info = sdscat(info,"\r\n");
1865 uname(&name);
1866 info = sdscatprintf(info,
1867 "# Server\r\n"
1868 "redis_version:%s\r\n"
1869 "redis_git_sha1:%s\r\n"
1870 "redis_git_dirty:%d\r\n"
1871 "os:%s %s %s\r\n"
1872 "arch_bits:%d\r\n"
1873 "multiplexing_api:%s\r\n"
1874 "gcc_version:%d.%d.%d\r\n"
1875 "process_id:%ld\r\n"
1876 "run_id:%s\r\n"
1877 "tcp_port:%d\r\n"
1878 "uptime_in_seconds:%ld\r\n"
1879 "uptime_in_days:%ld\r\n"
1880 "lru_clock:%ld\r\n",
1881 REDIS_VERSION,
1882 redisGitSHA1(),
1883 strtol(redisGitDirty(),NULL,10) > 0,
1884 name.sysname, name.release, name.machine,
1885 server.arch_bits,
1886 aeGetApiName(),
1887 #ifdef __GNUC__
1888 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1889 #else
1890 0,0,0,
1891 #endif
1892 (long) getpid(),
1893 server.runid,
1894 server.port,
1895 uptime,
1896 uptime/(3600*24),
1897 (unsigned long) server.lruclock);
1898 }
1899
1900 /* Clients */
1901 if (allsections || defsections || !strcasecmp(section,"clients")) {
1902 if (sections++) info = sdscat(info,"\r\n");
1903 info = sdscatprintf(info,
1904 "# Clients\r\n"
1905 "connected_clients:%lu\r\n"
1906 "client_longest_output_list:%lu\r\n"
1907 "client_biggest_input_buf:%lu\r\n"
1908 "blocked_clients:%d\r\n",
1909 listLength(server.clients)-listLength(server.slaves),
1910 lol, bib,
1911 server.bpop_blocked_clients);
1912 }
1913
1914 /* Memory */
1915 if (allsections || defsections || !strcasecmp(section,"memory")) {
1916 char hmem[64];
1917 char peak_hmem[64];
1918
1919 bytesToHuman(hmem,zmalloc_used_memory());
1920 bytesToHuman(peak_hmem,server.stat_peak_memory);
1921 if (sections++) info = sdscat(info,"\r\n");
1922 info = sdscatprintf(info,
1923 "# Memory\r\n"
1924 "used_memory:%zu\r\n"
1925 "used_memory_human:%s\r\n"
1926 "used_memory_rss:%zu\r\n"
1927 "used_memory_peak:%zu\r\n"
1928 "used_memory_peak_human:%s\r\n"
1929 "used_memory_lua:%lld\r\n"
1930 "mem_fragmentation_ratio:%.2f\r\n"
1931 "mem_allocator:%s\r\n",
1932 zmalloc_used_memory(),
1933 hmem,
1934 zmalloc_get_rss(),
1935 server.stat_peak_memory,
1936 peak_hmem,
1937 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1938 zmalloc_get_fragmentation_ratio(),
1939 ZMALLOC_LIB
1940 );
1941 }
1942
1943 /* Persistence */
1944 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1945 if (sections++) info = sdscat(info,"\r\n");
1946 info = sdscatprintf(info,
1947 "# Persistence\r\n"
1948 "loading:%d\r\n"
1949 "rdb_changes_since_last_save:%lld\r\n"
1950 "rdb_bgsave_in_progress:%d\r\n"
1951 "rdb_last_save_time:%ld\r\n"
1952 "rdb_last_bgsave_status:%s\r\n"
1953 "rdb_last_bgsave_time_sec:%ld\r\n"
1954 "rdb_current_bgsave_time_sec:%ld\r\n"
1955 "aof_enabled:%d\r\n"
1956 "aof_rewrite_in_progress:%d\r\n"
1957 "aof_rewrite_scheduled:%d\r\n"
1958 "aof_last_rewrite_time_sec:%ld\r\n"
1959 "aof_current_rewrite_time_sec:%ld\r\n",
1960 server.loading,
1961 server.dirty,
1962 server.rdb_child_pid != -1,
1963 server.lastsave,
1964 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1965 server.rdb_save_time_last,
1966 (server.rdb_child_pid == -1) ?
1967 -1 : time(NULL)-server.rdb_save_time_start,
1968 server.aof_state != REDIS_AOF_OFF,
1969 server.aof_child_pid != -1,
1970 server.aof_rewrite_scheduled,
1971 server.aof_rewrite_time_last,
1972 (server.aof_child_pid == -1) ?
1973 -1 : time(NULL)-server.aof_rewrite_time_start);
1974
1975 if (server.aof_state != REDIS_AOF_OFF) {
1976 info = sdscatprintf(info,
1977 "aof_current_size:%lld\r\n"
1978 "aof_base_size:%lld\r\n"
1979 "aof_pending_rewrite:%d\r\n"
1980 "aof_buffer_length:%zu\r\n"
1981 "aof_rewrite_buffer_length:%zu\r\n"
1982 "aof_pending_bio_fsync:%llu\r\n"
1983 "aof_delayed_fsync:%lu\r\n",
1984 (long long) server.aof_current_size,
1985 (long long) server.aof_rewrite_base_size,
1986 server.aof_rewrite_scheduled,
1987 sdslen(server.aof_buf),
1988 aofRewriteBufferSize(),
1989 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1990 server.aof_delayed_fsync);
1991 }
1992
1993 if (server.loading) {
1994 double perc;
1995 time_t eta, elapsed;
1996 off_t remaining_bytes = server.loading_total_bytes-
1997 server.loading_loaded_bytes;
1998
1999 perc = ((double)server.loading_loaded_bytes /
2000 server.loading_total_bytes) * 100;
2001
2002 elapsed = server.unixtime-server.loading_start_time;
2003 if (elapsed == 0) {
2004 eta = 1; /* A fake 1 second figure if we don't have
2005 enough info */
2006 } else {
2007 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
2008 }
2009
2010 info = sdscatprintf(info,
2011 "loading_start_time:%ld\r\n"
2012 "loading_total_bytes:%llu\r\n"
2013 "loading_loaded_bytes:%llu\r\n"
2014 "loading_loaded_perc:%.2f\r\n"
2015 "loading_eta_seconds:%ld\r\n"
2016 ,(unsigned long) server.loading_start_time,
2017 (unsigned long long) server.loading_total_bytes,
2018 (unsigned long long) server.loading_loaded_bytes,
2019 perc,
2020 eta
2021 );
2022 }
2023 }
2024
2025 /* Stats */
2026 if (allsections || defsections || !strcasecmp(section,"stats")) {
2027 if (sections++) info = sdscat(info,"\r\n");
2028 info = sdscatprintf(info,
2029 "# Stats\r\n"
2030 "total_connections_received:%lld\r\n"
2031 "total_commands_processed:%lld\r\n"
2032 "instantaneous_ops_per_sec:%lld\r\n"
2033 "rejected_connections:%lld\r\n"
2034 "expired_keys:%lld\r\n"
2035 "evicted_keys:%lld\r\n"
2036 "keyspace_hits:%lld\r\n"
2037 "keyspace_misses:%lld\r\n"
2038 "pubsub_channels:%ld\r\n"
2039 "pubsub_patterns:%lu\r\n"
2040 "latest_fork_usec:%lld\r\n",
2041 server.stat_numconnections,
2042 server.stat_numcommands,
2043 getOperationsPerSecond(),
2044 server.stat_rejected_conn,
2045 server.stat_expiredkeys,
2046 server.stat_evictedkeys,
2047 server.stat_keyspace_hits,
2048 server.stat_keyspace_misses,
2049 dictSize(server.pubsub_channels),
2050 listLength(server.pubsub_patterns),
2051 server.stat_fork_time);
2052 }
2053
2054 /* Replication */
2055 if (allsections || defsections || !strcasecmp(section,"replication")) {
2056 if (sections++) info = sdscat(info,"\r\n");
2057 info = sdscatprintf(info,
2058 "# Replication\r\n"
2059 "role:%s\r\n",
2060 server.masterhost == NULL ? "master" : "slave");
2061 if (server.masterhost) {
2062 info = sdscatprintf(info,
2063 "master_host:%s\r\n"
2064 "master_port:%d\r\n"
2065 "master_link_status:%s\r\n"
2066 "master_last_io_seconds_ago:%d\r\n"
2067 "master_sync_in_progress:%d\r\n"
2068 ,server.masterhost,
2069 server.masterport,
2070 (server.repl_state == REDIS_REPL_CONNECTED) ?
2071 "up" : "down",
2072 server.master ?
2073 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
2074 server.repl_state == REDIS_REPL_TRANSFER
2075 );
2076
2077 if (server.repl_state == REDIS_REPL_TRANSFER) {
2078 info = sdscatprintf(info,
2079 "master_sync_left_bytes:%ld\r\n"
2080 "master_sync_last_io_seconds_ago:%d\r\n"
2081 ,(long)server.repl_transfer_left,
2082 (int)(server.unixtime-server.repl_transfer_lastio)
2083 );
2084 }
2085
2086 if (server.repl_state != REDIS_REPL_CONNECTED) {
2087 info = sdscatprintf(info,
2088 "master_link_down_since_seconds:%ld\r\n",
2089 (long)server.unixtime-server.repl_down_since);
2090 }
2091 }
2092 info = sdscatprintf(info,
2093 "connected_slaves:%lu\r\n",
2094 listLength(server.slaves));
2095 if (listLength(server.slaves)) {
2096 int slaveid = 0;
2097 listNode *ln;
2098 listIter li;
2099
2100 listRewind(server.slaves,&li);
2101 while((ln = listNext(&li))) {
2102 redisClient *slave = listNodeValue(ln);
2103 char *state = NULL;
2104 char ip[32];
2105 int port;
2106
2107 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
2108 switch(slave->replstate) {
2109 case REDIS_REPL_WAIT_BGSAVE_START:
2110 case REDIS_REPL_WAIT_BGSAVE_END:
2111 state = "wait_bgsave";
2112 break;
2113 case REDIS_REPL_SEND_BULK:
2114 state = "send_bulk";
2115 break;
2116 case REDIS_REPL_ONLINE:
2117 state = "online";
2118 break;
2119 }
2120 if (state == NULL) continue;
2121 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2122 slaveid,ip,port,state);
2123 slaveid++;
2124 }
2125 }
2126 }
2127
2128 /* CPU */
2129 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2130 if (sections++) info = sdscat(info,"\r\n");
2131 info = sdscatprintf(info,
2132 "# CPU\r\n"
2133 "used_cpu_sys:%.2f\r\n"
2134 "used_cpu_user:%.2f\r\n"
2135 "used_cpu_sys_children:%.2f\r\n"
2136 "used_cpu_user_children:%.2f\r\n",
2137 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2138 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2139 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2140 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2141 }
2142
2143 /* cmdtime */
2144 if (allsections || !strcasecmp(section,"commandstats")) {
2145 if (sections++) info = sdscat(info,"\r\n");
2146 info = sdscatprintf(info, "# Commandstats\r\n");
2147 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2148 for (j = 0; j < numcommands; j++) {
2149 struct redisCommand *c = redisCommandTable+j;
2150
2151 if (!c->calls) continue;
2152 info = sdscatprintf(info,
2153 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2154 c->name, c->calls, c->microseconds,
2155 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2156 }
2157 }
2158
2159 /* Cluster */
2160 if (allsections || defsections || !strcasecmp(section,"cluster")) {
2161 if (sections++) info = sdscat(info,"\r\n");
2162 info = sdscatprintf(info,
2163 "# Cluster\r\n"
2164 "cluster_enabled:%d\r\n",
2165 server.cluster_enabled);
2166 }
2167
2168 /* Key space */
2169 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2170 if (sections++) info = sdscat(info,"\r\n");
2171 info = sdscatprintf(info, "# Keyspace\r\n");
2172 for (j = 0; j < server.dbnum; j++) {
2173 long long keys, vkeys;
2174
2175 keys = dictSize(server.db[j].dict);
2176 vkeys = dictSize(server.db[j].expires);
2177 if (keys || vkeys) {
2178 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2179 j, keys, vkeys);
2180 }
2181 }
2182 }
2183 return info;
2184 }
2185
2186 void infoCommand(redisClient *c) {
2187 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2188
2189 if (c->argc > 2) {
2190 addReply(c,shared.syntaxerr);
2191 return;
2192 }
2193 sds info = genRedisInfoString(section);
2194 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2195 (unsigned long)sdslen(info)));
2196 addReplySds(c,info);
2197 addReply(c,shared.crlf);
2198 }
2199
2200 void monitorCommand(redisClient *c) {
2201 /* ignore MONITOR if aleady slave or in monitor mode */
2202 if (c->flags & REDIS_SLAVE) return;
2203
2204 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2205 c->slaveseldb = 0;
2206 listAddNodeTail(server.monitors,c);
2207 addReply(c,shared.ok);
2208 }
2209
2210 /* ============================ Maxmemory directive ======================== */
2211
2212 /* This function gets called when 'maxmemory' is set on the config file to limit
2213 * the max memory used by the server, before processing a command.
2214 *
2215 * The goal of the function is to free enough memory to keep Redis under the
2216 * configured memory limit.
2217 *
2218 * The function starts calculating how many bytes should be freed to keep
2219 * Redis under the limit, and enters a loop selecting the best keys to
2220 * evict accordingly to the configured policy.
2221 *
2222 * If all the bytes needed to return back under the limit were freed the
2223 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2224 * should block the execution of commands that will result in more memory
2225 * used by the server.
2226 */
2227 int freeMemoryIfNeeded(void) {
2228 size_t mem_used, mem_tofree, mem_freed;
2229 int slaves = listLength(server.slaves);
2230
2231 /* Remove the size of slaves output buffers and AOF buffer from the
2232 * count of used memory. */
2233 mem_used = zmalloc_used_memory();
2234 if (slaves) {
2235 listIter li;
2236 listNode *ln;
2237
2238 listRewind(server.slaves,&li);
2239 while((ln = listNext(&li))) {
2240 redisClient *slave = listNodeValue(ln);
2241 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2242 if (obuf_bytes > mem_used)
2243 mem_used = 0;
2244 else
2245 mem_used -= obuf_bytes;
2246 }
2247 }
2248 if (server.aof_state != REDIS_AOF_OFF) {
2249 mem_used -= sdslen(server.aof_buf);
2250 mem_used -= aofRewriteBufferSize();
2251 }
2252
2253 /* Check if we are over the memory limit. */
2254 if (mem_used <= server.maxmemory) return REDIS_OK;
2255
2256 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2257 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2258
2259 /* Compute how much memory we need to free. */
2260 mem_tofree = mem_used - server.maxmemory;
2261 mem_freed = 0;
2262 while (mem_freed < mem_tofree) {
2263 int j, k, keys_freed = 0;
2264
2265 for (j = 0; j < server.dbnum; j++) {
2266 long bestval = 0; /* just to prevent warning */
2267 sds bestkey = NULL;
2268 struct dictEntry *de;
2269 redisDb *db = server.db+j;
2270 dict *dict;
2271
2272 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2273 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2274 {
2275 dict = server.db[j].dict;
2276 } else {
2277 dict = server.db[j].expires;
2278 }
2279 if (dictSize(dict) == 0) continue;
2280
2281 /* volatile-random and allkeys-random policy */
2282 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2283 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2284 {
2285 de = dictGetRandomKey(dict);
2286 bestkey = dictGetKey(de);
2287 }
2288
2289 /* volatile-lru and allkeys-lru policy */
2290 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2291 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2292 {
2293 for (k = 0; k < server.maxmemory_samples; k++) {
2294 sds thiskey;
2295 long thisval;
2296 robj *o;
2297
2298 de = dictGetRandomKey(dict);
2299 thiskey = dictGetKey(de);
2300 /* When policy is volatile-lru we need an additonal lookup
2301 * to locate the real key, as dict is set to db->expires. */
2302 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2303 de = dictFind(db->dict, thiskey);
2304 o = dictGetVal(de);
2305 thisval = estimateObjectIdleTime(o);
2306
2307 /* Higher idle time is better candidate for deletion */
2308 if (bestkey == NULL || thisval > bestval) {
2309 bestkey = thiskey;
2310 bestval = thisval;
2311 }
2312 }
2313 }
2314
2315 /* volatile-ttl */
2316 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2317 for (k = 0; k < server.maxmemory_samples; k++) {
2318 sds thiskey;
2319 long thisval;
2320
2321 de = dictGetRandomKey(dict);
2322 thiskey = dictGetKey(de);
2323 thisval = (long) dictGetVal(de);
2324
2325 /* Expire sooner (minor expire unix timestamp) is better
2326 * candidate for deletion */
2327 if (bestkey == NULL || thisval < bestval) {
2328 bestkey = thiskey;
2329 bestval = thisval;
2330 }
2331 }
2332 }
2333
2334 /* Finally remove the selected key. */
2335 if (bestkey) {
2336 long long delta;
2337
2338 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2339 propagateExpire(db,keyobj);
2340 /* We compute the amount of memory freed by dbDelete() alone.
2341 * It is possible that actually the memory needed to propagate
2342 * the DEL in AOF and replication link is greater than the one
2343 * we are freeing removing the key, but we can't account for
2344 * that otherwise we would never exit the loop.
2345 *
2346 * AOF and Output buffer memory will be freed eventually so
2347 * we only care about memory used by the key space. */
2348 delta = (long long) zmalloc_used_memory();
2349 dbDelete(db,keyobj);
2350 delta -= (long long) zmalloc_used_memory();
2351 mem_freed += delta;
2352 server.stat_evictedkeys++;
2353 decrRefCount(keyobj);
2354 keys_freed++;
2355
2356 /* When the memory to free starts to be big enough, we may
2357 * start spending so much time here that is impossible to
2358 * deliver data to the slaves fast enough, so we force the
2359 * transmission here inside the loop. */
2360 if (slaves) flushSlavesOutputBuffers();
2361 }
2362 }
2363 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2364 }
2365 return REDIS_OK;
2366 }
2367
2368 /* =================================== Main! ================================ */
2369
2370 #ifdef __linux__
2371 int linuxOvercommitMemoryValue(void) {
2372 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2373 char buf[64];
2374
2375 if (!fp) return -1;
2376 if (fgets(buf,64,fp) == NULL) {
2377 fclose(fp);
2378 return -1;
2379 }
2380 fclose(fp);
2381
2382 return atoi(buf);
2383 }
2384
2385 void linuxOvercommitMemoryWarning(void) {
2386 if (linuxOvercommitMemoryValue() == 0) {
2387 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.");
2388 }
2389 }
2390 #endif /* __linux__ */
2391
2392 void createPidFile(void) {
2393 /* Try to write the pid file in a best-effort way. */
2394 FILE *fp = fopen(server.pidfile,"w");
2395 if (fp) {
2396 fprintf(fp,"%d\n",(int)getpid());
2397 fclose(fp);
2398 }
2399 }
2400
2401 void daemonize(void) {
2402 int fd;
2403
2404 if (fork() != 0) exit(0); /* parent exits */
2405 setsid(); /* create a new session */
2406
2407 /* Every output goes to /dev/null. If Redis is daemonized but
2408 * the 'logfile' is set to 'stdout' in the configuration file
2409 * it will not log at all. */
2410 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2411 dup2(fd, STDIN_FILENO);
2412 dup2(fd, STDOUT_FILENO);
2413 dup2(fd, STDERR_FILENO);
2414 if (fd > STDERR_FILENO) close(fd);
2415 }
2416 }
2417
2418 void version() {
2419 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2420 REDIS_VERSION,
2421 redisGitSHA1(),
2422 atoi(redisGitDirty()) > 0,
2423 ZMALLOC_LIB,
2424 sizeof(long) == 4 ? 32 : 64);
2425 exit(0);
2426 }
2427
2428 void usage() {
2429 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2430 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2431 fprintf(stderr," ./redis-server -v or --version\n");
2432 fprintf(stderr," ./redis-server -h or --help\n");
2433 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2434 fprintf(stderr,"Examples:\n");
2435 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2436 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2437 fprintf(stderr," ./redis-server --port 7777\n");
2438 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2439 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2440 exit(1);
2441 }
2442
2443 void redisAsciiArt(void) {
2444 #include "asciilogo.h"
2445 char *buf = zmalloc(1024*16);
2446
2447 snprintf(buf,1024*16,ascii_logo,
2448 REDIS_VERSION,
2449 redisGitSHA1(),
2450 strtol(redisGitDirty(),NULL,10) > 0,
2451 (sizeof(long) == 8) ? "64" : "32",
2452 server.cluster_enabled ? "cluster" : "stand alone",
2453 server.port,
2454 (long) getpid()
2455 );
2456 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2457 zfree(buf);
2458 }
2459
2460 static void sigtermHandler(int sig) {
2461 REDIS_NOTUSED(sig);
2462
2463 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2464 server.shutdown_asap = 1;
2465 }
2466
2467 void setupSignalHandlers(void) {
2468 struct sigaction act;
2469
2470 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2471 * Otherwise, sa_handler is used. */
2472 sigemptyset(&act.sa_mask);
2473 act.sa_flags = 0;
2474 act.sa_handler = sigtermHandler;
2475 sigaction(SIGTERM, &act, NULL);
2476
2477 #ifdef HAVE_BACKTRACE
2478 sigemptyset(&act.sa_mask);
2479 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2480 act.sa_sigaction = sigsegvHandler;
2481 sigaction(SIGSEGV, &act, NULL);
2482 sigaction(SIGBUS, &act, NULL);
2483 sigaction(SIGFPE, &act, NULL);
2484 sigaction(SIGILL, &act, NULL);
2485 #endif
2486 return;
2487 }
2488
2489 void memtest(size_t megabytes, int passes);
2490
2491 int main(int argc, char **argv) {
2492 long long start;
2493 struct timeval tv;
2494
2495 /* We need to initialize our libraries, and the server configuration. */
2496 zmalloc_enable_thread_safeness();
2497 srand(time(NULL)^getpid());
2498 gettimeofday(&tv,NULL);
2499 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2500 initServerConfig();
2501
2502 if (argc >= 2) {
2503 int j = 1; /* First option to parse in argv[] */
2504 sds options = sdsempty();
2505 char *configfile = NULL;
2506
2507 /* Handle special options --help and --version */
2508 if (strcmp(argv[1], "-v") == 0 ||
2509 strcmp(argv[1], "--version") == 0) version();
2510 if (strcmp(argv[1], "--help") == 0 ||
2511 strcmp(argv[1], "-h") == 0) usage();
2512 if (strcmp(argv[1], "--test-memory") == 0) {
2513 if (argc == 3) {
2514 memtest(atoi(argv[2]),50);
2515 exit(0);
2516 } else {
2517 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2518 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2519 exit(1);
2520 }
2521 }
2522
2523 /* First argument is the config file name? */
2524 if (argv[j][0] != '-' || argv[j][1] != '-')
2525 configfile = argv[j++];
2526 /* All the other options are parsed and conceptually appended to the
2527 * configuration file. For instance --port 6380 will generate the
2528 * string "port 6380\n" to be parsed after the actual file name
2529 * is parsed, if any. */
2530 while(j != argc) {
2531 if (argv[j][0] == '-' && argv[j][1] == '-') {
2532 /* Option name */
2533 if (sdslen(options)) options = sdscat(options,"\n");
2534 options = sdscat(options,argv[j]+2);
2535 options = sdscat(options," ");
2536 } else {
2537 /* Option argument */
2538 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2539 options = sdscat(options," ");
2540 }
2541 j++;
2542 }
2543 resetServerSaveParams();
2544 loadServerConfig(configfile,options);
2545 sdsfree(options);
2546 } else {
2547 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'");
2548 }
2549 if (server.daemonize) daemonize();
2550 initServer();
2551 if (server.daemonize) createPidFile();
2552 redisAsciiArt();
2553 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2554 #ifdef __linux__
2555 linuxOvercommitMemoryWarning();
2556 #endif
2557 start = ustime();
2558 if (server.aof_state == REDIS_AOF_ON) {
2559 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2560 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2561 } else {
2562 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2563 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2564 (float)(ustime()-start)/1000000);
2565 } else if (errno != ENOENT) {
2566 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2567 exit(1);
2568 }
2569 }
2570 if (server.ipfd > 0)
2571 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2572 if (server.sofd > 0)
2573 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2574 aeSetBeforeSleepProc(server.el,beforeSleep);
2575 aeMain(server.el);
2576 aeDeleteEventLoop(server.el);
2577 return 0;
2578 }
2579
2580 /* The End */