]> git.saurik.com Git - redis.git/blob - src/redis.c
Check write(2) return value to avoid warnings, because in this context failing write...
[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
1071 updateLRUClock();
1072 resetServerSaveParams();
1073
1074 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1075 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1076 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1077 /* Replication related */
1078 server.masterauth = NULL;
1079 server.masterhost = NULL;
1080 server.masterport = 6379;
1081 server.master = NULL;
1082 server.repl_state = REDIS_REPL_NONE;
1083 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1084 server.repl_serve_stale_data = 1;
1085 server.repl_slave_ro = 1;
1086 server.repl_down_since = time(NULL);
1087
1088 /* Client output buffer limits */
1089 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1090 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1091 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1092 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1093 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1094 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1095 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1096 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1097 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1098
1099 /* Double constants initialization */
1100 R_Zero = 0.0;
1101 R_PosInf = 1.0/R_Zero;
1102 R_NegInf = -1.0/R_Zero;
1103 R_Nan = R_Zero/R_Zero;
1104
1105 /* Command table -- we intiialize it here as it is part of the
1106 * initial configuration, since command names may be changed via
1107 * redis.conf using the rename-command directive. */
1108 server.commands = dictCreate(&commandTableDictType,NULL);
1109 populateCommandTable();
1110 server.delCommand = lookupCommandByCString("del");
1111 server.multiCommand = lookupCommandByCString("multi");
1112 server.lpushCommand = lookupCommandByCString("lpush");
1113
1114 /* Slow log */
1115 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1116 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1117
1118 /* Debugging */
1119 server.assert_failed = "<no assertion failed>";
1120 server.assert_file = "<no file>";
1121 server.assert_line = 0;
1122 server.bug_report_start = 0;
1123 server.watchdog_period = 0;
1124 }
1125
1126 /* This function will try to raise the max number of open files accordingly to
1127 * the configured max number of clients. It will also account for 32 additional
1128 * file descriptors as we need a few more for persistence, listening
1129 * sockets, log files and so forth.
1130 *
1131 * If it will not be possible to set the limit accordingly to the configured
1132 * max number of clients, the function will do the reverse setting
1133 * server.maxclients to the value that we can actually handle. */
1134 void adjustOpenFilesLimit(void) {
1135 rlim_t maxfiles = server.maxclients+32;
1136 struct rlimit limit;
1137
1138 if (maxfiles < 1024) maxfiles = 1024;
1139 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1140 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1141 strerror(errno));
1142 server.maxclients = 1024-32;
1143 } else {
1144 rlim_t oldlimit = limit.rlim_cur;
1145
1146 /* Set the max number of files if the current limit is not enough
1147 * for our needs. */
1148 if (oldlimit < maxfiles) {
1149 rlim_t f;
1150
1151 f = maxfiles;
1152 while(f > oldlimit) {
1153 limit.rlim_cur = f;
1154 limit.rlim_max = f;
1155 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1156 f -= 128;
1157 }
1158 if (f < oldlimit) f = oldlimit;
1159 if (f != maxfiles) {
1160 server.maxclients = f-32;
1161 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1162 (int) maxfiles, strerror(errno), (int) server.maxclients);
1163 } else {
1164 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1165 (int) maxfiles);
1166 }
1167 }
1168 }
1169 }
1170
1171 void initServer() {
1172 int j;
1173
1174 signal(SIGHUP, SIG_IGN);
1175 signal(SIGPIPE, SIG_IGN);
1176 setupSignalHandlers();
1177
1178 if (server.syslog_enabled) {
1179 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1180 server.syslog_facility);
1181 }
1182
1183 server.current_client = NULL;
1184 server.clients = listCreate();
1185 server.clients_to_close = listCreate();
1186 server.slaves = listCreate();
1187 server.monitors = listCreate();
1188 server.unblocked_clients = listCreate();
1189
1190 createSharedObjects();
1191 adjustOpenFilesLimit();
1192 server.el = aeCreateEventLoop(server.maxclients+1024);
1193 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1194
1195 if (server.port != 0) {
1196 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1197 if (server.ipfd == ANET_ERR) {
1198 redisLog(REDIS_WARNING, "Opening port %d: %s",
1199 server.port, server.neterr);
1200 exit(1);
1201 }
1202 }
1203 if (server.unixsocket != NULL) {
1204 unlink(server.unixsocket); /* don't care if this fails */
1205 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1206 if (server.sofd == ANET_ERR) {
1207 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1208 exit(1);
1209 }
1210 }
1211 if (server.ipfd < 0 && server.sofd < 0) {
1212 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1213 exit(1);
1214 }
1215 for (j = 0; j < server.dbnum; j++) {
1216 server.db[j].dict = dictCreate(&dbDictType,NULL);
1217 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1218 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1219 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1220 server.db[j].id = j;
1221 }
1222 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1223 server.pubsub_patterns = listCreate();
1224 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1225 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1226 server.cronloops = 0;
1227 server.rdb_child_pid = -1;
1228 server.aof_child_pid = -1;
1229 server.aof_rewrite_buf = sdsempty();
1230 server.aof_buf = sdsempty();
1231 server.lastsave = time(NULL);
1232 server.dirty = 0;
1233 server.stat_numcommands = 0;
1234 server.stat_numconnections = 0;
1235 server.stat_expiredkeys = 0;
1236 server.stat_evictedkeys = 0;
1237 server.stat_starttime = time(NULL);
1238 server.stat_keyspace_misses = 0;
1239 server.stat_keyspace_hits = 0;
1240 server.stat_peak_memory = 0;
1241 server.stat_fork_time = 0;
1242 server.stat_rejected_conn = 0;
1243 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1244 server.ops_sec_idx = 0;
1245 server.ops_sec_last_sample_time = mstime();
1246 server.ops_sec_last_sample_ops = 0;
1247 server.unixtime = time(NULL);
1248 server.lastbgsave_status = REDIS_OK;
1249 server.stop_writes_on_bgsave_err = 1;
1250 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1251 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1252 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1253 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1254 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1255
1256 if (server.aof_state == REDIS_AOF_ON) {
1257 server.aof_fd = open(server.aof_filename,
1258 O_WRONLY|O_APPEND|O_CREAT,0644);
1259 if (server.aof_fd == -1) {
1260 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1261 strerror(errno));
1262 exit(1);
1263 }
1264 }
1265
1266 /* 32 bit instances are limited to 4GB of address space, so if there is
1267 * no explicit limit in the user provided configuration we set a limit
1268 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1269 * useless crashes of the Redis instance. */
1270 if (server.arch_bits == 32 && server.maxmemory == 0) {
1271 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1272 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1273 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1274 }
1275
1276 scriptingInit();
1277 slowlogInit();
1278 bioInit();
1279 }
1280
1281 /* Populates the Redis Command Table starting from the hard coded list
1282 * we have on top of redis.c file. */
1283 void populateCommandTable(void) {
1284 int j;
1285 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1286
1287 for (j = 0; j < numcommands; j++) {
1288 struct redisCommand *c = redisCommandTable+j;
1289 char *f = c->sflags;
1290 int retval;
1291
1292 while(*f != '\0') {
1293 switch(*f) {
1294 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1295 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1296 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1297 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1298 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1299 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1300 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1301 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1302 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1303 default: redisPanic("Unsupported command flag"); break;
1304 }
1305 f++;
1306 }
1307
1308 retval = dictAdd(server.commands, sdsnew(c->name), c);
1309 assert(retval == DICT_OK);
1310 }
1311 }
1312
1313 void resetCommandTableStats(void) {
1314 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1315 int j;
1316
1317 for (j = 0; j < numcommands; j++) {
1318 struct redisCommand *c = redisCommandTable+j;
1319
1320 c->microseconds = 0;
1321 c->calls = 0;
1322 }
1323 }
1324
1325 /* ========================== Redis OP Array API ============================ */
1326
1327 void redisOpArrayInit(redisOpArray *oa) {
1328 oa->ops = NULL;
1329 oa->numops = 0;
1330 }
1331
1332 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1333 robj **argv, int argc, int target)
1334 {
1335 redisOp *op;
1336
1337 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1338 op = oa->ops+oa->numops;
1339 op->cmd = cmd;
1340 op->dbid = dbid;
1341 op->argv = argv;
1342 op->argc = argc;
1343 op->target = target;
1344 oa->numops++;
1345 return oa->numops;
1346 }
1347
1348 void redisOpArrayFree(redisOpArray *oa) {
1349 while(oa->numops) {
1350 int j;
1351 redisOp *op;
1352
1353 oa->numops--;
1354 op = oa->ops+oa->numops;
1355 for (j = 0; j < op->argc; j++)
1356 decrRefCount(op->argv[j]);
1357 zfree(op->argv);
1358 }
1359 zfree(oa->ops);
1360 }
1361
1362 /* ====================== Commands lookup and execution ===================== */
1363
1364 struct redisCommand *lookupCommand(sds name) {
1365 return dictFetchValue(server.commands, name);
1366 }
1367
1368 struct redisCommand *lookupCommandByCString(char *s) {
1369 struct redisCommand *cmd;
1370 sds name = sdsnew(s);
1371
1372 cmd = dictFetchValue(server.commands, name);
1373 sdsfree(name);
1374 return cmd;
1375 }
1376
1377 /* Propagate the specified command (in the context of the specified database id)
1378 * to AOF, Slaves and Monitors.
1379 *
1380 * flags are an xor between:
1381 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1382 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1383 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1384 */
1385 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1386 int flags)
1387 {
1388 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1389 feedAppendOnlyFile(cmd,dbid,argv,argc);
1390 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1391 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1392 }
1393
1394 /* Used inside commands to schedule the propagation of additional commands
1395 * after the current command is propagated to AOF / Replication. */
1396 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1397 int target)
1398 {
1399 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1400 }
1401
1402 /* Call() is the core of Redis execution of a command */
1403 void call(redisClient *c, int flags) {
1404 long long dirty, start = ustime(), duration;
1405
1406 /* Sent the command to clients in MONITOR mode, only if the commands are
1407 * not geneated from reading an AOF. */
1408 if (listLength(server.monitors) && !server.loading)
1409 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1410
1411 /* Call the command. */
1412 redisOpArrayInit(&server.also_propagate);
1413 dirty = server.dirty;
1414 c->cmd->proc(c);
1415 dirty = server.dirty-dirty;
1416 duration = ustime()-start;
1417
1418 /* When EVAL is called loading the AOF we don't want commands called
1419 * from Lua to go into the slowlog or to populate statistics. */
1420 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1421 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1422
1423 /* Log the command into the Slow log if needed, and populate the
1424 * per-command statistics that we show in INFO commandstats. */
1425 if (flags & REDIS_CALL_SLOWLOG)
1426 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1427 if (flags & REDIS_CALL_STATS) {
1428 c->cmd->microseconds += duration;
1429 c->cmd->calls++;
1430 }
1431
1432 /* Propagate the command into the AOF and replication link */
1433 if (flags & REDIS_CALL_PROPAGATE) {
1434 int flags = REDIS_PROPAGATE_NONE;
1435
1436 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1437 flags |= REDIS_PROPAGATE_REPL;
1438 if (dirty)
1439 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1440 if (flags != REDIS_PROPAGATE_NONE)
1441 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1442 }
1443 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1444 * PUSH command. */
1445 if (server.also_propagate.numops) {
1446 int j;
1447 redisOp *rop;
1448
1449 for (j = 0; j < server.also_propagate.numops; j++) {
1450 rop = &server.also_propagate.ops[j];
1451 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1452 }
1453 redisOpArrayFree(&server.also_propagate);
1454 }
1455 server.stat_numcommands++;
1456 }
1457
1458 /* If this function gets called we already read a whole
1459 * command, argments are in the client argv/argc fields.
1460 * processCommand() execute the command or prepare the
1461 * server for a bulk read from the client.
1462 *
1463 * If 1 is returned the client is still alive and valid and
1464 * and other operations can be performed by the caller. Otherwise
1465 * if 0 is returned the client was destroied (i.e. after QUIT). */
1466 int processCommand(redisClient *c) {
1467 /* The QUIT command is handled separately. Normal command procs will
1468 * go through checking for replication and QUIT will cause trouble
1469 * when FORCE_REPLICATION is enabled and would be implemented in
1470 * a regular command proc. */
1471 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1472 addReply(c,shared.ok);
1473 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1474 return REDIS_ERR;
1475 }
1476
1477 /* Now lookup the command and check ASAP about trivial error conditions
1478 * such as wrong arity, bad command name and so forth. */
1479 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1480 if (!c->cmd) {
1481 addReplyErrorFormat(c,"unknown command '%s'",
1482 (char*)c->argv[0]->ptr);
1483 return REDIS_OK;
1484 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1485 (c->argc < -c->cmd->arity)) {
1486 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1487 c->cmd->name);
1488 return REDIS_OK;
1489 }
1490
1491 /* Check if the user is authenticated */
1492 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1493 {
1494 addReplyError(c,"operation not permitted");
1495 return REDIS_OK;
1496 }
1497
1498 /* Handle the maxmemory directive.
1499 *
1500 * First we try to free some memory if possible (if there are volatile
1501 * keys in the dataset). If there are not the only thing we can do
1502 * is returning an error. */
1503 if (server.maxmemory) {
1504 int retval = freeMemoryIfNeeded();
1505 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1506 addReply(c, shared.oomerr);
1507 return REDIS_OK;
1508 }
1509 }
1510
1511 /* Don't accept write commands if there are problems persisting on disk. */
1512 if (server.stop_writes_on_bgsave_err &&
1513 server.saveparamslen > 0
1514 && server.lastbgsave_status == REDIS_ERR &&
1515 c->cmd->flags & REDIS_CMD_WRITE)
1516 {
1517 addReply(c, shared.bgsaveerr);
1518 return REDIS_OK;
1519 }
1520
1521 /* Don't accept wirte commands if this is a read only slave. But
1522 * accept write commands if this is our master. */
1523 if (server.masterhost && server.repl_slave_ro &&
1524 !(c->flags & REDIS_MASTER) &&
1525 c->cmd->flags & REDIS_CMD_WRITE)
1526 {
1527 addReply(c, shared.roslaveerr);
1528 return REDIS_OK;
1529 }
1530
1531 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1532 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1533 &&
1534 c->cmd->proc != subscribeCommand &&
1535 c->cmd->proc != unsubscribeCommand &&
1536 c->cmd->proc != psubscribeCommand &&
1537 c->cmd->proc != punsubscribeCommand) {
1538 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1539 return REDIS_OK;
1540 }
1541
1542 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1543 * we are a slave with a broken link with master. */
1544 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1545 server.repl_serve_stale_data == 0 &&
1546 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1547 {
1548 addReplyError(c,
1549 "link with MASTER is down and slave-serve-stale-data is set to no");
1550 return REDIS_OK;
1551 }
1552
1553 /* Loading DB? Return an error if the command is not INFO */
1554 if (server.loading && c->cmd->proc != infoCommand) {
1555 addReply(c, shared.loadingerr);
1556 return REDIS_OK;
1557 }
1558
1559 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1560 if (server.lua_timedout &&
1561 !(c->cmd->proc != shutdownCommand &&
1562 c->argc == 2 &&
1563 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1564 !(c->cmd->proc == scriptCommand &&
1565 c->argc == 2 &&
1566 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1567 {
1568 addReply(c, shared.slowscripterr);
1569 return REDIS_OK;
1570 }
1571
1572 /* Exec the command */
1573 if (c->flags & REDIS_MULTI &&
1574 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1575 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1576 {
1577 queueMultiCommand(c);
1578 addReply(c,shared.queued);
1579 } else {
1580 call(c,REDIS_CALL_FULL);
1581 }
1582 return REDIS_OK;
1583 }
1584
1585 /*================================== Shutdown =============================== */
1586
1587 int prepareForShutdown(int flags) {
1588 int save = flags & REDIS_SHUTDOWN_SAVE;
1589 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1590
1591 redisLog(REDIS_WARNING,"User requested shutdown...");
1592 /* Kill the saving child if there is a background saving in progress.
1593 We want to avoid race conditions, for instance our saving child may
1594 overwrite the synchronous saving did by SHUTDOWN. */
1595 if (server.rdb_child_pid != -1) {
1596 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1597 kill(server.rdb_child_pid,SIGKILL);
1598 rdbRemoveTempFile(server.rdb_child_pid);
1599 }
1600 if (server.aof_state != REDIS_AOF_OFF) {
1601 /* Kill the AOF saving child as the AOF we already have may be longer
1602 * but contains the full dataset anyway. */
1603 if (server.aof_child_pid != -1) {
1604 redisLog(REDIS_WARNING,
1605 "There is a child rewriting the AOF. Killing it!");
1606 kill(server.aof_child_pid,SIGKILL);
1607 }
1608 /* Append only file: fsync() the AOF and exit */
1609 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1610 aof_fsync(server.aof_fd);
1611 }
1612 if ((server.saveparamslen > 0 && !nosave) || save) {
1613 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1614 /* Snapshotting. Perform a SYNC SAVE and exit */
1615 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1616 /* Ooops.. error saving! The best we can do is to continue
1617 * operating. Note that if there was a background saving process,
1618 * in the next cron() Redis will be notified that the background
1619 * saving aborted, handling special stuff like slaves pending for
1620 * synchronization... */
1621 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1622 return REDIS_ERR;
1623 }
1624 }
1625 if (server.daemonize) {
1626 redisLog(REDIS_NOTICE,"Removing the pid file.");
1627 unlink(server.pidfile);
1628 }
1629 /* Close the listening sockets. Apparently this allows faster restarts. */
1630 if (server.ipfd != -1) close(server.ipfd);
1631 if (server.sofd != -1) close(server.sofd);
1632 if (server.unixsocket) {
1633 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1634 unlink(server.unixsocket); /* don't care if this fails */
1635 }
1636
1637 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1638 return REDIS_OK;
1639 }
1640
1641 /*================================== Commands =============================== */
1642
1643 void authCommand(redisClient *c) {
1644 if (!server.requirepass) {
1645 addReplyError(c,"Client sent AUTH, but no password is set");
1646 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1647 c->authenticated = 1;
1648 addReply(c,shared.ok);
1649 } else {
1650 c->authenticated = 0;
1651 addReplyError(c,"invalid password");
1652 }
1653 }
1654
1655 void pingCommand(redisClient *c) {
1656 addReply(c,shared.pong);
1657 }
1658
1659 void echoCommand(redisClient *c) {
1660 addReplyBulk(c,c->argv[1]);
1661 }
1662
1663 void timeCommand(redisClient *c) {
1664 struct timeval tv;
1665
1666 /* gettimeofday() can only fail if &tv is a bad addresss so we
1667 * don't check for errors. */
1668 gettimeofday(&tv,NULL);
1669 addReplyMultiBulkLen(c,2);
1670 addReplyBulkLongLong(c,tv.tv_sec);
1671 addReplyBulkLongLong(c,tv.tv_usec);
1672 }
1673
1674 /* Convert an amount of bytes into a human readable string in the form
1675 * of 100B, 2G, 100M, 4K, and so forth. */
1676 void bytesToHuman(char *s, unsigned long long n) {
1677 double d;
1678
1679 if (n < 1024) {
1680 /* Bytes */
1681 sprintf(s,"%lluB",n);
1682 return;
1683 } else if (n < (1024*1024)) {
1684 d = (double)n/(1024);
1685 sprintf(s,"%.2fK",d);
1686 } else if (n < (1024LL*1024*1024)) {
1687 d = (double)n/(1024*1024);
1688 sprintf(s,"%.2fM",d);
1689 } else if (n < (1024LL*1024*1024*1024)) {
1690 d = (double)n/(1024LL*1024*1024);
1691 sprintf(s,"%.2fG",d);
1692 }
1693 }
1694
1695 /* Create the string returned by the INFO command. This is decoupled
1696 * by the INFO command itself as we need to report the same information
1697 * on memory corruption problems. */
1698 sds genRedisInfoString(char *section) {
1699 sds info = sdsempty();
1700 time_t uptime = server.unixtime-server.stat_starttime;
1701 int j, numcommands;
1702 struct rusage self_ru, c_ru;
1703 unsigned long lol, bib;
1704 int allsections = 0, defsections = 0;
1705 int sections = 0;
1706
1707 if (section) {
1708 allsections = strcasecmp(section,"all") == 0;
1709 defsections = strcasecmp(section,"default") == 0;
1710 }
1711
1712 getrusage(RUSAGE_SELF, &self_ru);
1713 getrusage(RUSAGE_CHILDREN, &c_ru);
1714 getClientsMaxBuffers(&lol,&bib);
1715
1716 /* Server */
1717 if (allsections || defsections || !strcasecmp(section,"server")) {
1718 struct utsname name;
1719
1720 if (sections++) info = sdscat(info,"\r\n");
1721 uname(&name);
1722 info = sdscatprintf(info,
1723 "# Server\r\n"
1724 "redis_version:%s\r\n"
1725 "redis_git_sha1:%s\r\n"
1726 "redis_git_dirty:%d\r\n"
1727 "os:%s %s %s\r\n"
1728 "arch_bits:%d\r\n"
1729 "multiplexing_api:%s\r\n"
1730 "gcc_version:%d.%d.%d\r\n"
1731 "process_id:%ld\r\n"
1732 "run_id:%s\r\n"
1733 "tcp_port:%d\r\n"
1734 "uptime_in_seconds:%ld\r\n"
1735 "uptime_in_days:%ld\r\n"
1736 "lru_clock:%ld\r\n",
1737 REDIS_VERSION,
1738 redisGitSHA1(),
1739 strtol(redisGitDirty(),NULL,10) > 0,
1740 name.sysname, name.release, name.machine,
1741 server.arch_bits,
1742 aeGetApiName(),
1743 #ifdef __GNUC__
1744 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1745 #else
1746 0,0,0,
1747 #endif
1748 (long) getpid(),
1749 server.runid,
1750 server.port,
1751 uptime,
1752 uptime/(3600*24),
1753 (unsigned long) server.lruclock);
1754 }
1755
1756 /* Clients */
1757 if (allsections || defsections || !strcasecmp(section,"clients")) {
1758 if (sections++) info = sdscat(info,"\r\n");
1759 info = sdscatprintf(info,
1760 "# Clients\r\n"
1761 "connected_clients:%lu\r\n"
1762 "client_longest_output_list:%lu\r\n"
1763 "client_biggest_input_buf:%lu\r\n"
1764 "blocked_clients:%d\r\n",
1765 listLength(server.clients)-listLength(server.slaves),
1766 lol, bib,
1767 server.bpop_blocked_clients);
1768 }
1769
1770 /* Memory */
1771 if (allsections || defsections || !strcasecmp(section,"memory")) {
1772 char hmem[64];
1773 char peak_hmem[64];
1774
1775 bytesToHuman(hmem,zmalloc_used_memory());
1776 bytesToHuman(peak_hmem,server.stat_peak_memory);
1777 if (sections++) info = sdscat(info,"\r\n");
1778 info = sdscatprintf(info,
1779 "# Memory\r\n"
1780 "used_memory:%zu\r\n"
1781 "used_memory_human:%s\r\n"
1782 "used_memory_rss:%zu\r\n"
1783 "used_memory_peak:%zu\r\n"
1784 "used_memory_peak_human:%s\r\n"
1785 "used_memory_lua:%lld\r\n"
1786 "mem_fragmentation_ratio:%.2f\r\n"
1787 "mem_allocator:%s\r\n",
1788 zmalloc_used_memory(),
1789 hmem,
1790 zmalloc_get_rss(),
1791 server.stat_peak_memory,
1792 peak_hmem,
1793 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1794 zmalloc_get_fragmentation_ratio(),
1795 ZMALLOC_LIB
1796 );
1797 }
1798
1799 /* Persistence */
1800 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1801 if (sections++) info = sdscat(info,"\r\n");
1802 info = sdscatprintf(info,
1803 "# Persistence\r\n"
1804 "loading:%d\r\n"
1805 "aof_enabled:%d\r\n"
1806 "changes_since_last_save:%lld\r\n"
1807 "bgsave_in_progress:%d\r\n"
1808 "last_save_time:%ld\r\n"
1809 "last_bgsave_status:%s\r\n"
1810 "bgrewriteaof_in_progress:%d\r\n"
1811 "bgrewriteaof_scheduled:%d\r\n",
1812 server.loading,
1813 server.aof_state != REDIS_AOF_OFF,
1814 server.dirty,
1815 server.rdb_child_pid != -1,
1816 server.lastsave,
1817 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1818 server.aof_child_pid != -1,
1819 server.aof_rewrite_scheduled);
1820
1821 if (server.aof_state != REDIS_AOF_OFF) {
1822 info = sdscatprintf(info,
1823 "aof_current_size:%lld\r\n"
1824 "aof_base_size:%lld\r\n"
1825 "aof_pending_rewrite:%d\r\n"
1826 "aof_buffer_length:%zu\r\n"
1827 "aof_pending_bio_fsync:%llu\r\n"
1828 "aof_delayed_fsync:%lu\r\n",
1829 (long long) server.aof_current_size,
1830 (long long) server.aof_rewrite_base_size,
1831 server.aof_rewrite_scheduled,
1832 sdslen(server.aof_buf),
1833 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1834 server.aof_delayed_fsync);
1835 }
1836
1837 if (server.loading) {
1838 double perc;
1839 time_t eta, elapsed;
1840 off_t remaining_bytes = server.loading_total_bytes-
1841 server.loading_loaded_bytes;
1842
1843 perc = ((double)server.loading_loaded_bytes /
1844 server.loading_total_bytes) * 100;
1845
1846 elapsed = server.unixtime-server.loading_start_time;
1847 if (elapsed == 0) {
1848 eta = 1; /* A fake 1 second figure if we don't have
1849 enough info */
1850 } else {
1851 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1852 }
1853
1854 info = sdscatprintf(info,
1855 "loading_start_time:%ld\r\n"
1856 "loading_total_bytes:%llu\r\n"
1857 "loading_loaded_bytes:%llu\r\n"
1858 "loading_loaded_perc:%.2f\r\n"
1859 "loading_eta_seconds:%ld\r\n"
1860 ,(unsigned long) server.loading_start_time,
1861 (unsigned long long) server.loading_total_bytes,
1862 (unsigned long long) server.loading_loaded_bytes,
1863 perc,
1864 eta
1865 );
1866 }
1867 }
1868
1869 /* Stats */
1870 if (allsections || defsections || !strcasecmp(section,"stats")) {
1871 if (sections++) info = sdscat(info,"\r\n");
1872 info = sdscatprintf(info,
1873 "# Stats\r\n"
1874 "total_connections_received:%lld\r\n"
1875 "total_commands_processed:%lld\r\n"
1876 "instantaneous_ops_per_sec:%lld\r\n"
1877 "rejected_connections:%lld\r\n"
1878 "expired_keys:%lld\r\n"
1879 "evicted_keys:%lld\r\n"
1880 "keyspace_hits:%lld\r\n"
1881 "keyspace_misses:%lld\r\n"
1882 "pubsub_channels:%ld\r\n"
1883 "pubsub_patterns:%lu\r\n"
1884 "latest_fork_usec:%lld\r\n",
1885 server.stat_numconnections,
1886 server.stat_numcommands,
1887 getOperationsPerSecond(),
1888 server.stat_rejected_conn,
1889 server.stat_expiredkeys,
1890 server.stat_evictedkeys,
1891 server.stat_keyspace_hits,
1892 server.stat_keyspace_misses,
1893 dictSize(server.pubsub_channels),
1894 listLength(server.pubsub_patterns),
1895 server.stat_fork_time);
1896 }
1897
1898 /* Replication */
1899 if (allsections || defsections || !strcasecmp(section,"replication")) {
1900 if (sections++) info = sdscat(info,"\r\n");
1901 info = sdscatprintf(info,
1902 "# Replication\r\n"
1903 "role:%s\r\n",
1904 server.masterhost == NULL ? "master" : "slave");
1905 if (server.masterhost) {
1906 info = sdscatprintf(info,
1907 "master_host:%s\r\n"
1908 "master_port:%d\r\n"
1909 "master_link_status:%s\r\n"
1910 "master_last_io_seconds_ago:%d\r\n"
1911 "master_sync_in_progress:%d\r\n"
1912 ,server.masterhost,
1913 server.masterport,
1914 (server.repl_state == REDIS_REPL_CONNECTED) ?
1915 "up" : "down",
1916 server.master ?
1917 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
1918 server.repl_state == REDIS_REPL_TRANSFER
1919 );
1920
1921 if (server.repl_state == REDIS_REPL_TRANSFER) {
1922 info = sdscatprintf(info,
1923 "master_sync_left_bytes:%ld\r\n"
1924 "master_sync_last_io_seconds_ago:%d\r\n"
1925 ,(long)server.repl_transfer_left,
1926 (int)(server.unixtime-server.repl_transfer_lastio)
1927 );
1928 }
1929
1930 if (server.repl_state != REDIS_REPL_CONNECTED) {
1931 info = sdscatprintf(info,
1932 "master_link_down_since_seconds:%ld\r\n",
1933 (long)server.unixtime-server.repl_down_since);
1934 }
1935 }
1936 info = sdscatprintf(info,
1937 "connected_slaves:%lu\r\n",
1938 listLength(server.slaves));
1939 if (listLength(server.slaves)) {
1940 int slaveid = 0;
1941 listNode *ln;
1942 listIter li;
1943
1944 listRewind(server.slaves,&li);
1945 while((ln = listNext(&li))) {
1946 redisClient *slave = listNodeValue(ln);
1947 char *state = NULL;
1948 char ip[32];
1949 int port;
1950
1951 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
1952 switch(slave->replstate) {
1953 case REDIS_REPL_WAIT_BGSAVE_START:
1954 case REDIS_REPL_WAIT_BGSAVE_END:
1955 state = "wait_bgsave";
1956 break;
1957 case REDIS_REPL_SEND_BULK:
1958 state = "send_bulk";
1959 break;
1960 case REDIS_REPL_ONLINE:
1961 state = "online";
1962 break;
1963 }
1964 if (state == NULL) continue;
1965 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
1966 slaveid,ip,port,state);
1967 slaveid++;
1968 }
1969 }
1970 }
1971
1972 /* CPU */
1973 if (allsections || defsections || !strcasecmp(section,"cpu")) {
1974 if (sections++) info = sdscat(info,"\r\n");
1975 info = sdscatprintf(info,
1976 "# CPU\r\n"
1977 "used_cpu_sys:%.2f\r\n"
1978 "used_cpu_user:%.2f\r\n"
1979 "used_cpu_sys_children:%.2f\r\n"
1980 "used_cpu_user_children:%.2f\r\n",
1981 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
1982 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
1983 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
1984 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
1985 }
1986
1987 /* cmdtime */
1988 if (allsections || !strcasecmp(section,"commandstats")) {
1989 if (sections++) info = sdscat(info,"\r\n");
1990 info = sdscatprintf(info, "# Commandstats\r\n");
1991 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1992 for (j = 0; j < numcommands; j++) {
1993 struct redisCommand *c = redisCommandTable+j;
1994
1995 if (!c->calls) continue;
1996 info = sdscatprintf(info,
1997 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
1998 c->name, c->calls, c->microseconds,
1999 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2000 }
2001 }
2002
2003 /* Key space */
2004 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2005 if (sections++) info = sdscat(info,"\r\n");
2006 info = sdscatprintf(info, "# Keyspace\r\n");
2007 for (j = 0; j < server.dbnum; j++) {
2008 long long keys, vkeys;
2009
2010 keys = dictSize(server.db[j].dict);
2011 vkeys = dictSize(server.db[j].expires);
2012 if (keys || vkeys) {
2013 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2014 j, keys, vkeys);
2015 }
2016 }
2017 }
2018 return info;
2019 }
2020
2021 void infoCommand(redisClient *c) {
2022 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2023
2024 if (c->argc > 2) {
2025 addReply(c,shared.syntaxerr);
2026 return;
2027 }
2028 sds info = genRedisInfoString(section);
2029 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2030 (unsigned long)sdslen(info)));
2031 addReplySds(c,info);
2032 addReply(c,shared.crlf);
2033 }
2034
2035 void monitorCommand(redisClient *c) {
2036 /* ignore MONITOR if aleady slave or in monitor mode */
2037 if (c->flags & REDIS_SLAVE) return;
2038
2039 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2040 c->slaveseldb = 0;
2041 listAddNodeTail(server.monitors,c);
2042 addReply(c,shared.ok);
2043 }
2044
2045 /* ============================ Maxmemory directive ======================== */
2046
2047 /* This function gets called when 'maxmemory' is set on the config file to limit
2048 * the max memory used by the server, before processing a command.
2049 *
2050 * The goal of the function is to free enough memory to keep Redis under the
2051 * configured memory limit.
2052 *
2053 * The function starts calculating how many bytes should be freed to keep
2054 * Redis under the limit, and enters a loop selecting the best keys to
2055 * evict accordingly to the configured policy.
2056 *
2057 * If all the bytes needed to return back under the limit were freed the
2058 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2059 * should block the execution of commands that will result in more memory
2060 * used by the server.
2061 */
2062 int freeMemoryIfNeeded(void) {
2063 size_t mem_used, mem_tofree, mem_freed;
2064 int slaves = listLength(server.slaves);
2065
2066 /* Remove the size of slaves output buffers and AOF buffer from the
2067 * count of used memory. */
2068 mem_used = zmalloc_used_memory();
2069 if (slaves) {
2070 listIter li;
2071 listNode *ln;
2072
2073 listRewind(server.slaves,&li);
2074 while((ln = listNext(&li))) {
2075 redisClient *slave = listNodeValue(ln);
2076 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2077 if (obuf_bytes > mem_used)
2078 mem_used = 0;
2079 else
2080 mem_used -= obuf_bytes;
2081 }
2082 }
2083 if (server.aof_state != REDIS_AOF_OFF) {
2084 mem_used -= sdslen(server.aof_buf);
2085 mem_used -= sdslen(server.aof_rewrite_buf);
2086 }
2087
2088 /* Check if we are over the memory limit. */
2089 if (mem_used <= server.maxmemory) return REDIS_OK;
2090
2091 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2092 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2093
2094 /* Compute how much memory we need to free. */
2095 mem_tofree = mem_used - server.maxmemory;
2096 mem_freed = 0;
2097 while (mem_freed < mem_tofree) {
2098 int j, k, keys_freed = 0;
2099
2100 for (j = 0; j < server.dbnum; j++) {
2101 long bestval = 0; /* just to prevent warning */
2102 sds bestkey = NULL;
2103 struct dictEntry *de;
2104 redisDb *db = server.db+j;
2105 dict *dict;
2106
2107 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2108 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2109 {
2110 dict = server.db[j].dict;
2111 } else {
2112 dict = server.db[j].expires;
2113 }
2114 if (dictSize(dict) == 0) continue;
2115
2116 /* volatile-random and allkeys-random policy */
2117 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2118 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2119 {
2120 de = dictGetRandomKey(dict);
2121 bestkey = dictGetKey(de);
2122 }
2123
2124 /* volatile-lru and allkeys-lru policy */
2125 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2126 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2127 {
2128 for (k = 0; k < server.maxmemory_samples; k++) {
2129 sds thiskey;
2130 long thisval;
2131 robj *o;
2132
2133 de = dictGetRandomKey(dict);
2134 thiskey = dictGetKey(de);
2135 /* When policy is volatile-lru we need an additonal lookup
2136 * to locate the real key, as dict is set to db->expires. */
2137 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2138 de = dictFind(db->dict, thiskey);
2139 o = dictGetVal(de);
2140 thisval = estimateObjectIdleTime(o);
2141
2142 /* Higher idle time is better candidate for deletion */
2143 if (bestkey == NULL || thisval > bestval) {
2144 bestkey = thiskey;
2145 bestval = thisval;
2146 }
2147 }
2148 }
2149
2150 /* volatile-ttl */
2151 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2152 for (k = 0; k < server.maxmemory_samples; k++) {
2153 sds thiskey;
2154 long thisval;
2155
2156 de = dictGetRandomKey(dict);
2157 thiskey = dictGetKey(de);
2158 thisval = (long) dictGetVal(de);
2159
2160 /* Expire sooner (minor expire unix timestamp) is better
2161 * candidate for deletion */
2162 if (bestkey == NULL || thisval < bestval) {
2163 bestkey = thiskey;
2164 bestval = thisval;
2165 }
2166 }
2167 }
2168
2169 /* Finally remove the selected key. */
2170 if (bestkey) {
2171 long long delta;
2172
2173 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2174 propagateExpire(db,keyobj);
2175 /* We compute the amount of memory freed by dbDelete() alone.
2176 * It is possible that actually the memory needed to propagate
2177 * the DEL in AOF and replication link is greater than the one
2178 * we are freeing removing the key, but we can't account for
2179 * that otherwise we would never exit the loop.
2180 *
2181 * AOF and Output buffer memory will be freed eventually so
2182 * we only care about memory used by the key space. */
2183 delta = (long long) zmalloc_used_memory();
2184 dbDelete(db,keyobj);
2185 delta -= (long long) zmalloc_used_memory();
2186 mem_freed += delta;
2187 server.stat_evictedkeys++;
2188 decrRefCount(keyobj);
2189 keys_freed++;
2190
2191 /* When the memory to free starts to be big enough, we may
2192 * start spending so much time here that is impossible to
2193 * deliver data to the slaves fast enough, so we force the
2194 * transmission here inside the loop. */
2195 if (slaves) flushSlavesOutputBuffers();
2196 }
2197 }
2198 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2199 }
2200 return REDIS_OK;
2201 }
2202
2203 /* =================================== Main! ================================ */
2204
2205 #ifdef __linux__
2206 int linuxOvercommitMemoryValue(void) {
2207 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2208 char buf[64];
2209
2210 if (!fp) return -1;
2211 if (fgets(buf,64,fp) == NULL) {
2212 fclose(fp);
2213 return -1;
2214 }
2215 fclose(fp);
2216
2217 return atoi(buf);
2218 }
2219
2220 void linuxOvercommitMemoryWarning(void) {
2221 if (linuxOvercommitMemoryValue() == 0) {
2222 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.");
2223 }
2224 }
2225 #endif /* __linux__ */
2226
2227 void createPidFile(void) {
2228 /* Try to write the pid file in a best-effort way. */
2229 FILE *fp = fopen(server.pidfile,"w");
2230 if (fp) {
2231 fprintf(fp,"%d\n",(int)getpid());
2232 fclose(fp);
2233 }
2234 }
2235
2236 void daemonize(void) {
2237 int fd;
2238
2239 if (fork() != 0) exit(0); /* parent exits */
2240 setsid(); /* create a new session */
2241
2242 /* Every output goes to /dev/null. If Redis is daemonized but
2243 * the 'logfile' is set to 'stdout' in the configuration file
2244 * it will not log at all. */
2245 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2246 dup2(fd, STDIN_FILENO);
2247 dup2(fd, STDOUT_FILENO);
2248 dup2(fd, STDERR_FILENO);
2249 if (fd > STDERR_FILENO) close(fd);
2250 }
2251 }
2252
2253 void version() {
2254 printf("Redis server v=%s sha=%s:%d malloc=%s\n", REDIS_VERSION,
2255 redisGitSHA1(), atoi(redisGitDirty()) > 0, ZMALLOC_LIB);
2256 exit(0);
2257 }
2258
2259 void usage() {
2260 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2261 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2262 fprintf(stderr," ./redis-server -v or --version\n");
2263 fprintf(stderr," ./redis-server -h or --help\n");
2264 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2265 fprintf(stderr,"Examples:\n");
2266 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2267 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2268 fprintf(stderr," ./redis-server --port 7777\n");
2269 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2270 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2271 exit(1);
2272 }
2273
2274 void redisAsciiArt(void) {
2275 #include "asciilogo.h"
2276 char *buf = zmalloc(1024*16);
2277
2278 snprintf(buf,1024*16,ascii_logo,
2279 REDIS_VERSION,
2280 redisGitSHA1(),
2281 strtol(redisGitDirty(),NULL,10) > 0,
2282 (sizeof(long) == 8) ? "64" : "32",
2283 "stand alone",
2284 server.port,
2285 (long) getpid()
2286 );
2287 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2288 zfree(buf);
2289 }
2290
2291 static void sigtermHandler(int sig) {
2292 REDIS_NOTUSED(sig);
2293
2294 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2295 server.shutdown_asap = 1;
2296 }
2297
2298 void setupSignalHandlers(void) {
2299 struct sigaction act;
2300 stack_t stack;
2301
2302 stack.ss_sp = altstack;
2303 stack.ss_flags = 0;
2304 stack.ss_size = SIGSTKSZ;
2305
2306 sigaltstack(&stack, NULL);
2307
2308 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2309 * Otherwise, sa_handler is used. */
2310 sigemptyset(&act.sa_mask);
2311 act.sa_flags = 0;
2312 act.sa_handler = sigtermHandler;
2313 sigaction(SIGTERM, &act, NULL);
2314
2315 #ifdef HAVE_BACKTRACE
2316 /* Use alternate stack so we don't clobber stack in case of segv, or when we run out of stack ..
2317 * also resethand & nodefer so we can get interrupted (and killed) if we cause SEGV during SEGV handler */
2318 sigemptyset(&act.sa_mask);
2319 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
2320 act.sa_sigaction = sigsegvHandler;
2321 sigaction(SIGSEGV, &act, NULL);
2322 sigaction(SIGBUS, &act, NULL);
2323 sigaction(SIGFPE, &act, NULL);
2324 sigaction(SIGILL, &act, NULL);
2325 #endif
2326 return;
2327 }
2328
2329 void memtest(size_t megabytes, int passes);
2330
2331 int main(int argc, char **argv) {
2332 long long start;
2333 struct timeval tv;
2334
2335 /* We need to initialize our libraries, and the server configuration. */
2336 zmalloc_enable_thread_safeness();
2337 srand(time(NULL)^getpid());
2338 gettimeofday(&tv,NULL);
2339 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2340 initServerConfig();
2341
2342 if (argc >= 2) {
2343 int j = 1; /* First option to parse in argv[] */
2344 sds options = sdsempty();
2345 char *configfile = NULL;
2346
2347 /* Handle special options --help and --version */
2348 if (strcmp(argv[1], "-v") == 0 ||
2349 strcmp(argv[1], "--version") == 0) version();
2350 if (strcmp(argv[1], "--help") == 0 ||
2351 strcmp(argv[1], "-h") == 0) usage();
2352 if (strcmp(argv[1], "--test-memory") == 0) {
2353 if (argc == 3) {
2354 memtest(atoi(argv[2]),50);
2355 exit(0);
2356 } else {
2357 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2358 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2359 exit(1);
2360 }
2361 }
2362
2363 /* First argument is the config file name? */
2364 if (argv[j][0] != '-' || argv[j][1] != '-')
2365 configfile = argv[j++];
2366 /* All the other options are parsed and conceptually appended to the
2367 * configuration file. For instance --port 6380 will generate the
2368 * string "port 6380\n" to be parsed after the actual file name
2369 * is parsed, if any. */
2370 while(j != argc) {
2371 if (argv[j][0] == '-' && argv[j][1] == '-') {
2372 /* Option name */
2373 if (sdslen(options)) options = sdscat(options,"\n");
2374 options = sdscat(options,argv[j]+2);
2375 options = sdscat(options," ");
2376 } else {
2377 /* Option argument */
2378 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2379 options = sdscat(options," ");
2380 }
2381 j++;
2382 }
2383 resetServerSaveParams();
2384 loadServerConfig(configfile,options);
2385 sdsfree(options);
2386 } else {
2387 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'");
2388 }
2389 if (server.daemonize) daemonize();
2390 initServer();
2391 if (server.daemonize) createPidFile();
2392 redisAsciiArt();
2393 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2394 #ifdef __linux__
2395 linuxOvercommitMemoryWarning();
2396 #endif
2397 start = ustime();
2398 if (server.aof_state == REDIS_AOF_ON) {
2399 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2400 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2401 } else {
2402 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2403 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2404 (float)(ustime()-start)/1000000);
2405 } else if (errno != ENOENT) {
2406 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2407 exit(1);
2408 }
2409 }
2410 if (server.ipfd > 0)
2411 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2412 if (server.sofd > 0)
2413 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2414 aeSetBeforeSleepProc(server.el,beforeSleep);
2415 aeMain(server.el);
2416 aeDeleteEventLoop(server.el);
2417 return 0;
2418 }
2419
2420 /* The End */