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