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