]> git.saurik.com Git - redis.git/blob - src/redis.c
RDB files now embed a crc64 checksum. Version of RDB bumped to 5.
[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 write(fd,"[",1);
320 write(fd,buf,strlen(buf));
321 write(fd," | signal handler] (",20);
322 ll2string(buf,sizeof(buf),time(NULL));
323 write(fd,buf,strlen(buf));
324 write(fd,") ",2);
325 write(fd,msg,strlen(msg));
326 write(fd,"\n",1);
327 if (server.logfile) close(fd);
328 }
329
330 /* Redis generally does not try to recover from out of memory conditions
331 * when allocating objects or strings, it is not clear if it will be possible
332 * to report this condition to the client since the networking layer itself
333 * is based on heap allocation for send buffers, so we simply abort.
334 * At least the code will be simpler to read... */
335 void oom(const char *msg) {
336 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
337 sleep(1);
338 abort();
339 }
340
341 /* Return the UNIX time in microseconds */
342 long long ustime(void) {
343 struct timeval tv;
344 long long ust;
345
346 gettimeofday(&tv, NULL);
347 ust = ((long long)tv.tv_sec)*1000000;
348 ust += tv.tv_usec;
349 return ust;
350 }
351
352 /* Return the UNIX time in milliseconds */
353 long long mstime(void) {
354 return ustime()/1000;
355 }
356
357 /* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
358 * exit(), because the latter may interact with the same file objects used by
359 * the parent process. However if we are testing the coverage normal exit() is
360 * used in order to obtain the right coverage information. */
361 void exitFromChild(int retcode) {
362 #ifdef COVERAGE_TEST
363 exit(retcode);
364 #else
365 _exit(retcode);
366 #endif
367 }
368
369 /*====================== Hash table type implementation ==================== */
370
371 /* This is an hash table type that uses the SDS dynamic strings libary as
372 * keys and radis objects as values (objects can hold SDS strings,
373 * lists, sets). */
374
375 void dictVanillaFree(void *privdata, void *val)
376 {
377 DICT_NOTUSED(privdata);
378 zfree(val);
379 }
380
381 void dictListDestructor(void *privdata, void *val)
382 {
383 DICT_NOTUSED(privdata);
384 listRelease((list*)val);
385 }
386
387 int dictSdsKeyCompare(void *privdata, const void *key1,
388 const void *key2)
389 {
390 int l1,l2;
391 DICT_NOTUSED(privdata);
392
393 l1 = sdslen((sds)key1);
394 l2 = sdslen((sds)key2);
395 if (l1 != l2) return 0;
396 return memcmp(key1, key2, l1) == 0;
397 }
398
399 /* A case insensitive version used for the command lookup table. */
400 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
401 const void *key2)
402 {
403 DICT_NOTUSED(privdata);
404
405 return strcasecmp(key1, key2) == 0;
406 }
407
408 void dictRedisObjectDestructor(void *privdata, void *val)
409 {
410 DICT_NOTUSED(privdata);
411
412 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
413 decrRefCount(val);
414 }
415
416 void dictSdsDestructor(void *privdata, void *val)
417 {
418 DICT_NOTUSED(privdata);
419
420 sdsfree(val);
421 }
422
423 int dictObjKeyCompare(void *privdata, const void *key1,
424 const void *key2)
425 {
426 const robj *o1 = key1, *o2 = key2;
427 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
428 }
429
430 unsigned int dictObjHash(const void *key) {
431 const robj *o = key;
432 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
433 }
434
435 unsigned int dictSdsHash(const void *key) {
436 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
437 }
438
439 unsigned int dictSdsCaseHash(const void *key) {
440 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
441 }
442
443 int dictEncObjKeyCompare(void *privdata, const void *key1,
444 const void *key2)
445 {
446 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
447 int cmp;
448
449 if (o1->encoding == REDIS_ENCODING_INT &&
450 o2->encoding == REDIS_ENCODING_INT)
451 return o1->ptr == o2->ptr;
452
453 o1 = getDecodedObject(o1);
454 o2 = getDecodedObject(o2);
455 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
456 decrRefCount(o1);
457 decrRefCount(o2);
458 return cmp;
459 }
460
461 unsigned int dictEncObjHash(const void *key) {
462 robj *o = (robj*) key;
463
464 if (o->encoding == REDIS_ENCODING_RAW) {
465 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
466 } else {
467 if (o->encoding == REDIS_ENCODING_INT) {
468 char buf[32];
469 int len;
470
471 len = ll2string(buf,32,(long)o->ptr);
472 return dictGenHashFunction((unsigned char*)buf, len);
473 } else {
474 unsigned int hash;
475
476 o = getDecodedObject(o);
477 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
478 decrRefCount(o);
479 return hash;
480 }
481 }
482 }
483
484 /* Sets type hash table */
485 dictType setDictType = {
486 dictEncObjHash, /* hash function */
487 NULL, /* key dup */
488 NULL, /* val dup */
489 dictEncObjKeyCompare, /* key compare */
490 dictRedisObjectDestructor, /* key destructor */
491 NULL /* val destructor */
492 };
493
494 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
495 dictType zsetDictType = {
496 dictEncObjHash, /* hash function */
497 NULL, /* key dup */
498 NULL, /* val dup */
499 dictEncObjKeyCompare, /* key compare */
500 dictRedisObjectDestructor, /* key destructor */
501 NULL /* val destructor */
502 };
503
504 /* Db->dict, keys are sds strings, vals are Redis objects. */
505 dictType dbDictType = {
506 dictSdsHash, /* hash function */
507 NULL, /* key dup */
508 NULL, /* val dup */
509 dictSdsKeyCompare, /* key compare */
510 dictSdsDestructor, /* key destructor */
511 dictRedisObjectDestructor /* val destructor */
512 };
513
514 /* Db->expires */
515 dictType keyptrDictType = {
516 dictSdsHash, /* hash function */
517 NULL, /* key dup */
518 NULL, /* val dup */
519 dictSdsKeyCompare, /* key compare */
520 NULL, /* key destructor */
521 NULL /* val destructor */
522 };
523
524 /* Command table. sds string -> command struct pointer. */
525 dictType commandTableDictType = {
526 dictSdsCaseHash, /* hash function */
527 NULL, /* key dup */
528 NULL, /* val dup */
529 dictSdsKeyCaseCompare, /* key compare */
530 dictSdsDestructor, /* key destructor */
531 NULL /* val destructor */
532 };
533
534 /* Hash type hash table (note that small hashes are represented with zimpaps) */
535 dictType hashDictType = {
536 dictEncObjHash, /* hash function */
537 NULL, /* key dup */
538 NULL, /* val dup */
539 dictEncObjKeyCompare, /* key compare */
540 dictRedisObjectDestructor, /* key destructor */
541 dictRedisObjectDestructor /* val destructor */
542 };
543
544 /* Keylist hash table type has unencoded redis objects as keys and
545 * lists as values. It's used for blocking operations (BLPOP) and to
546 * map swapped keys to a list of clients waiting for this keys to be loaded. */
547 dictType keylistDictType = {
548 dictObjHash, /* hash function */
549 NULL, /* key dup */
550 NULL, /* val dup */
551 dictObjKeyCompare, /* key compare */
552 dictRedisObjectDestructor, /* key destructor */
553 dictListDestructor /* val destructor */
554 };
555
556 int htNeedsResize(dict *dict) {
557 long long size, used;
558
559 size = dictSlots(dict);
560 used = dictSize(dict);
561 return (size && used && size > DICT_HT_INITIAL_SIZE &&
562 (used*100/size < REDIS_HT_MINFILL));
563 }
564
565 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
566 * we resize the hash table to save memory */
567 void tryResizeHashTables(void) {
568 int j;
569
570 for (j = 0; j < server.dbnum; j++) {
571 if (htNeedsResize(server.db[j].dict))
572 dictResize(server.db[j].dict);
573 if (htNeedsResize(server.db[j].expires))
574 dictResize(server.db[j].expires);
575 }
576 }
577
578 /* Our hash table implementation performs rehashing incrementally while
579 * we write/read from the hash table. Still if the server is idle, the hash
580 * table will use two tables for a long time. So we try to use 1 millisecond
581 * of CPU time at every serverCron() loop in order to rehash some key. */
582 void incrementallyRehash(void) {
583 int j;
584
585 for (j = 0; j < server.dbnum; j++) {
586 if (dictIsRehashing(server.db[j].dict)) {
587 dictRehashMilliseconds(server.db[j].dict,1);
588 break; /* already used our millisecond for this loop... */
589 }
590 }
591 }
592
593 /* This function is called once a background process of some kind terminates,
594 * as we want to avoid resizing the hash tables when there is a child in order
595 * to play well with copy-on-write (otherwise when a resize happens lots of
596 * memory pages are copied). The goal of this function is to update the ability
597 * for dict.c to resize the hash tables accordingly to the fact we have o not
598 * running childs. */
599 void updateDictResizePolicy(void) {
600 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
601 dictEnableResize();
602 else
603 dictDisableResize();
604 }
605
606 /* ======================= Cron: called every 100 ms ======================== */
607
608 /* Try to expire a few timed out keys. The algorithm used is adaptive and
609 * will use few CPU cycles if there are few expiring keys, otherwise
610 * it will get more aggressive to avoid that too much memory is used by
611 * keys that can be removed from the keyspace. */
612 void activeExpireCycle(void) {
613 int j;
614
615 for (j = 0; j < server.dbnum; j++) {
616 int expired;
617 redisDb *db = server.db+j;
618
619 /* Continue to expire if at the end of the cycle more than 25%
620 * of the keys were expired. */
621 do {
622 long num = dictSize(db->expires);
623 long long now = mstime();
624
625 expired = 0;
626 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
627 num = REDIS_EXPIRELOOKUPS_PER_CRON;
628 while (num--) {
629 dictEntry *de;
630 long long t;
631
632 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
633 t = dictGetSignedIntegerVal(de);
634 if (now > t) {
635 sds key = dictGetKey(de);
636 robj *keyobj = createStringObject(key,sdslen(key));
637
638 propagateExpire(db,keyobj);
639 dbDelete(db,keyobj);
640 decrRefCount(keyobj);
641 expired++;
642 server.stat_expiredkeys++;
643 }
644 }
645 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
646 }
647 }
648
649 void updateLRUClock(void) {
650 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
651 REDIS_LRU_CLOCK_MAX;
652 }
653
654
655 /* Add a sample to the operations per second array of samples. */
656 void trackOperationsPerSecond(void) {
657 long long t = mstime() - server.ops_sec_last_sample_time;
658 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
659 long long ops_sec;
660
661 ops_sec = t > 0 ? (ops*1000/t) : 0;
662
663 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
664 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
665 server.ops_sec_last_sample_time = mstime();
666 server.ops_sec_last_sample_ops = server.stat_numcommands;
667 }
668
669 /* Return the mean of all the samples. */
670 long long getOperationsPerSecond(void) {
671 int j;
672 long long sum = 0;
673
674 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
675 sum += server.ops_sec_samples[j];
676 return sum / REDIS_OPS_SEC_SAMPLES;
677 }
678
679 /* Check for timeouts. Returns non-zero if the client was terminated */
680 int clientsCronHandleTimeout(redisClient *c) {
681 time_t now = server.unixtime;
682
683 if (server.maxidletime &&
684 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
685 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
686 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
687 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
688 listLength(c->pubsub_patterns) == 0 &&
689 (now - c->lastinteraction > server.maxidletime))
690 {
691 redisLog(REDIS_VERBOSE,"Closing idle client");
692 freeClient(c);
693 return 1;
694 } else if (c->flags & REDIS_BLOCKED) {
695 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
696 addReply(c,shared.nullmultibulk);
697 unblockClientWaitingData(c);
698 }
699 }
700 return 0;
701 }
702
703 /* The client query buffer is an sds.c string that can end with a lot of
704 * free space not used, this function reclaims space if needed.
705 *
706 * The funciton always returns 0 as it never terminates the client. */
707 int clientsCronResizeQueryBuffer(redisClient *c) {
708 size_t querybuf_size = sdsAllocSize(c->querybuf);
709 time_t idletime = server.unixtime - c->lastinteraction;
710
711 /* There are two conditions to resize the query buffer:
712 * 1) Query buffer is > BIG_ARG and too big for latest peak.
713 * 2) Client is inactive and the buffer is bigger than 1k. */
714 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
715 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
716 (querybuf_size > 1024 && idletime > 2))
717 {
718 /* Only resize the query buffer if it is actually wasting space. */
719 if (sdsavail(c->querybuf) > 1024) {
720 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
721 }
722 }
723 /* Reset the peak again to capture the peak memory usage in the next
724 * cycle. */
725 c->querybuf_peak = 0;
726 return 0;
727 }
728
729 void clientsCron(void) {
730 /* Make sure to process at least 1/100 of clients per call.
731 * Since this function is called 10 times per second we are sure that
732 * in the worst case we process all the clients in 10 seconds.
733 * In normal conditions (a reasonable number of clients) we process
734 * all the clients in a shorter time. */
735 int numclients = listLength(server.clients);
736 int iterations = numclients/100;
737
738 if (iterations < 50)
739 iterations = (numclients < 50) ? numclients : 50;
740 while(listLength(server.clients) && iterations--) {
741 redisClient *c;
742 listNode *head;
743
744 /* Rotate the list, take the current head, process.
745 * This way if the client must be removed from the list it's the
746 * first element and we don't incur into O(N) computation. */
747 listRotate(server.clients);
748 head = listFirst(server.clients);
749 c = listNodeValue(head);
750 /* The following functions do different service checks on the client.
751 * The protocol is that they return non-zero if the client was
752 * terminated. */
753 if (clientsCronHandleTimeout(c)) continue;
754 if (clientsCronResizeQueryBuffer(c)) continue;
755 }
756 }
757
758 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
759 int j, loops = server.cronloops;
760 REDIS_NOTUSED(eventLoop);
761 REDIS_NOTUSED(id);
762 REDIS_NOTUSED(clientData);
763
764 /* Software watchdog: deliver the SIGALRM that will reach the signal
765 * handler if we don't return here fast enough. */
766 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
767
768 /* We take a cached value of the unix time in the global state because
769 * with virtual memory and aging there is to store the current time
770 * in objects at every object access, and accuracy is not needed.
771 * To access a global var is faster than calling time(NULL) */
772 server.unixtime = time(NULL);
773
774 trackOperationsPerSecond();
775
776 /* We have just 22 bits per object for LRU information.
777 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
778 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
779 *
780 * Note that even if this will wrap after 1.5 years it's not a problem,
781 * everything will still work but just some object will appear younger
782 * to Redis. But for this to happen a given object should never be touched
783 * for 1.5 years.
784 *
785 * Note that you can change the resolution altering the
786 * REDIS_LRU_CLOCK_RESOLUTION define.
787 */
788 updateLRUClock();
789
790 /* Record the max memory used since the server was started. */
791 if (zmalloc_used_memory() > server.stat_peak_memory)
792 server.stat_peak_memory = zmalloc_used_memory();
793
794 /* We received a SIGTERM, shutting down here in a safe way, as it is
795 * not ok doing so inside the signal handler. */
796 if (server.shutdown_asap) {
797 if (prepareForShutdown(0) == REDIS_OK) exit(0);
798 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
799 }
800
801 /* Show some info about non-empty databases */
802 for (j = 0; j < server.dbnum; j++) {
803 long long size, used, vkeys;
804
805 size = dictSlots(server.db[j].dict);
806 used = dictSize(server.db[j].dict);
807 vkeys = dictSize(server.db[j].expires);
808 if (!(loops % 50) && (used || vkeys)) {
809 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
810 /* dictPrintStats(server.dict); */
811 }
812 }
813
814 /* We don't want to resize the hash tables while a bacground saving
815 * is in progress: the saving child is created using fork() that is
816 * implemented with a copy-on-write semantic in most modern systems, so
817 * if we resize the HT while there is the saving child at work actually
818 * a lot of memory movements in the parent will cause a lot of pages
819 * copied. */
820 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
821 if (!(loops % 10)) tryResizeHashTables();
822 if (server.activerehashing) incrementallyRehash();
823 }
824
825 /* Show information about connected clients */
826 if (!(loops % 50)) {
827 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
828 listLength(server.clients)-listLength(server.slaves),
829 listLength(server.slaves),
830 zmalloc_used_memory());
831 }
832
833 /* We need to do a few operations on clients asynchronously. */
834 clientsCron();
835
836 /* Start a scheduled AOF rewrite if this was requested by the user while
837 * a BGSAVE was in progress. */
838 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
839 server.aof_rewrite_scheduled)
840 {
841 rewriteAppendOnlyFileBackground();
842 }
843
844 /* Check if a background saving or AOF rewrite in progress terminated. */
845 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
846 int statloc;
847 pid_t pid;
848
849 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
850 int exitcode = WEXITSTATUS(statloc);
851 int bysignal = 0;
852
853 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
854
855 if (pid == server.rdb_child_pid) {
856 backgroundSaveDoneHandler(exitcode,bysignal);
857 } else {
858 backgroundRewriteDoneHandler(exitcode,bysignal);
859 }
860 updateDictResizePolicy();
861 }
862 } else {
863 /* If there is not a background saving/rewrite in progress check if
864 * we have to save/rewrite now */
865 for (j = 0; j < server.saveparamslen; j++) {
866 struct saveparam *sp = server.saveparams+j;
867
868 if (server.dirty >= sp->changes &&
869 server.unixtime-server.lastsave > sp->seconds) {
870 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
871 sp->changes, sp->seconds);
872 rdbSaveBackground(server.rdb_filename);
873 break;
874 }
875 }
876
877 /* Trigger an AOF rewrite if needed */
878 if (server.rdb_child_pid == -1 &&
879 server.aof_child_pid == -1 &&
880 server.aof_rewrite_perc &&
881 server.aof_current_size > server.aof_rewrite_min_size)
882 {
883 long long base = server.aof_rewrite_base_size ?
884 server.aof_rewrite_base_size : 1;
885 long long growth = (server.aof_current_size*100/base) - 100;
886 if (growth >= server.aof_rewrite_perc) {
887 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
888 rewriteAppendOnlyFileBackground();
889 }
890 }
891 }
892
893
894 /* If we postponed an AOF buffer flush, let's try to do it every time the
895 * cron function is called. */
896 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
897
898 /* Expire a few keys per cycle, only if this is a master.
899 * On slaves we wait for DEL operations synthesized by the master
900 * in order to guarantee a strict consistency. */
901 if (server.masterhost == NULL) activeExpireCycle();
902
903 /* Close clients that need to be closed asynchronous */
904 freeClientsInAsyncFreeQueue();
905
906 /* Replication cron function -- used to reconnect to master and
907 * to detect transfer failures. */
908 if (!(loops % 10)) replicationCron();
909
910 server.cronloops++;
911 return 100;
912 }
913
914 /* This function gets called every time Redis is entering the
915 * main loop of the event driven library, that is, before to sleep
916 * for ready file descriptors. */
917 void beforeSleep(struct aeEventLoop *eventLoop) {
918 REDIS_NOTUSED(eventLoop);
919 listNode *ln;
920 redisClient *c;
921
922 /* Try to process pending commands for clients that were just unblocked. */
923 while (listLength(server.unblocked_clients)) {
924 ln = listFirst(server.unblocked_clients);
925 redisAssert(ln != NULL);
926 c = ln->value;
927 listDelNode(server.unblocked_clients,ln);
928 c->flags &= ~REDIS_UNBLOCKED;
929
930 /* Process remaining data in the input buffer. */
931 if (c->querybuf && sdslen(c->querybuf) > 0) {
932 server.current_client = c;
933 processInputBuffer(c);
934 server.current_client = NULL;
935 }
936 }
937
938 /* Write the AOF buffer on disk */
939 flushAppendOnlyFile(0);
940 }
941
942 /* =========================== Server initialization ======================== */
943
944 void createSharedObjects(void) {
945 int j;
946
947 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
948 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
949 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
950 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
951 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
952 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
953 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
954 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
955 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
956 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
957 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
958 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
959 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
960 "-ERR Operation against a key holding the wrong kind of value\r\n"));
961 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
962 "-ERR no such key\r\n"));
963 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
964 "-ERR syntax error\r\n"));
965 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
966 "-ERR source and destination objects are the same\r\n"));
967 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
968 "-ERR index out of range\r\n"));
969 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
970 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
971 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
972 "-LOADING Redis is loading the dataset in memory\r\n"));
973 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
974 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
975 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
976 "-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"));
977 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
978 "-READONLY You can't write against a read only slave.\r\n"));
979 shared.oomerr = createObject(REDIS_STRING,sdsnew(
980 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
981 shared.space = createObject(REDIS_STRING,sdsnew(" "));
982 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
983 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
984
985 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
986 shared.select[j] = createObject(REDIS_STRING,
987 sdscatprintf(sdsempty(),"select %d\r\n", j));
988 }
989 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
990 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
991 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
992 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
993 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
994 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
995 shared.del = createStringObject("DEL",3);
996 shared.rpop = createStringObject("RPOP",4);
997 shared.lpop = createStringObject("LPOP",4);
998 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
999 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1000 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1001 }
1002 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1003 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1004 sdscatprintf(sdsempty(),"*%d\r\n",j));
1005 shared.bulkhdr[j] = createObject(REDIS_STRING,
1006 sdscatprintf(sdsempty(),"$%d\r\n",j));
1007 }
1008 }
1009
1010 void initServerConfig() {
1011 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1012 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1013 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1014 server.port = REDIS_SERVERPORT;
1015 server.bindaddr = NULL;
1016 server.unixsocket = NULL;
1017 server.unixsocketperm = 0;
1018 server.ipfd = -1;
1019 server.sofd = -1;
1020 server.dbnum = REDIS_DEFAULT_DBNUM;
1021 server.verbosity = REDIS_NOTICE;
1022 server.maxidletime = REDIS_MAXIDLETIME;
1023 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1024 server.saveparams = NULL;
1025 server.loading = 0;
1026 server.logfile = NULL; /* NULL = log on standard output */
1027 server.syslog_enabled = 0;
1028 server.syslog_ident = zstrdup("redis");
1029 server.syslog_facility = LOG_LOCAL0;
1030 server.daemonize = 0;
1031 server.aof_state = REDIS_AOF_OFF;
1032 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1033 server.aof_no_fsync_on_rewrite = 0;
1034 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1035 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1036 server.aof_rewrite_base_size = 0;
1037 server.aof_rewrite_scheduled = 0;
1038 server.aof_last_fsync = time(NULL);
1039 server.aof_delayed_fsync = 0;
1040 server.aof_fd = -1;
1041 server.aof_selected_db = -1; /* Make sure the first time will not match */
1042 server.aof_flush_postponed_start = 0;
1043 server.pidfile = zstrdup("/var/run/redis.pid");
1044 server.rdb_filename = zstrdup("dump.rdb");
1045 server.aof_filename = zstrdup("appendonly.aof");
1046 server.requirepass = NULL;
1047 server.rdb_compression = 1;
1048 server.activerehashing = 1;
1049 server.maxclients = REDIS_MAX_CLIENTS;
1050 server.bpop_blocked_clients = 0;
1051 server.maxmemory = 0;
1052 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1053 server.maxmemory_samples = 3;
1054 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1055 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1056 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1057 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1058 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1059 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1060 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1061 server.shutdown_asap = 0;
1062 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1063 server.repl_timeout = REDIS_REPL_TIMEOUT;
1064 server.lua_caller = NULL;
1065 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1066 server.lua_client = NULL;
1067 server.lua_timedout = 0;
1068
1069 updateLRUClock();
1070 resetServerSaveParams();
1071
1072 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1073 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1074 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1075 /* Replication related */
1076 server.masterauth = NULL;
1077 server.masterhost = NULL;
1078 server.masterport = 6379;
1079 server.master = NULL;
1080 server.repl_state = REDIS_REPL_NONE;
1081 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1082 server.repl_serve_stale_data = 1;
1083 server.repl_slave_ro = 1;
1084 server.repl_down_since = time(NULL);
1085
1086 /* Client output buffer limits */
1087 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1088 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1089 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1090 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1091 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1092 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1093 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1094 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1095 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1096
1097 /* Double constants initialization */
1098 R_Zero = 0.0;
1099 R_PosInf = 1.0/R_Zero;
1100 R_NegInf = -1.0/R_Zero;
1101 R_Nan = R_Zero/R_Zero;
1102
1103 /* Command table -- we intiialize it here as it is part of the
1104 * initial configuration, since command names may be changed via
1105 * redis.conf using the rename-command directive. */
1106 server.commands = dictCreate(&commandTableDictType,NULL);
1107 populateCommandTable();
1108 server.delCommand = lookupCommandByCString("del");
1109 server.multiCommand = lookupCommandByCString("multi");
1110 server.lpushCommand = lookupCommandByCString("lpush");
1111
1112 /* Slow log */
1113 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1114 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1115
1116 /* Debugging */
1117 server.assert_failed = "<no assertion failed>";
1118 server.assert_file = "<no file>";
1119 server.assert_line = 0;
1120 server.bug_report_start = 0;
1121 server.watchdog_period = 0;
1122 }
1123
1124 /* This function will try to raise the max number of open files accordingly to
1125 * the configured max number of clients. It will also account for 32 additional
1126 * file descriptors as we need a few more for persistence, listening
1127 * sockets, log files and so forth.
1128 *
1129 * If it will not be possible to set the limit accordingly to the configured
1130 * max number of clients, the function will do the reverse setting
1131 * server.maxclients to the value that we can actually handle. */
1132 void adjustOpenFilesLimit(void) {
1133 rlim_t maxfiles = server.maxclients+32;
1134 struct rlimit limit;
1135
1136 if (maxfiles < 1024) maxfiles = 1024;
1137 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1138 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1139 strerror(errno));
1140 server.maxclients = 1024-32;
1141 } else {
1142 rlim_t oldlimit = limit.rlim_cur;
1143
1144 /* Set the max number of files if the current limit is not enough
1145 * for our needs. */
1146 if (oldlimit < maxfiles) {
1147 rlim_t f;
1148
1149 f = maxfiles;
1150 while(f > oldlimit) {
1151 limit.rlim_cur = f;
1152 limit.rlim_max = f;
1153 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1154 f -= 128;
1155 }
1156 if (f < oldlimit) f = oldlimit;
1157 if (f != maxfiles) {
1158 server.maxclients = f-32;
1159 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1160 (int) maxfiles, strerror(errno), (int) server.maxclients);
1161 } else {
1162 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1163 (int) maxfiles);
1164 }
1165 }
1166 }
1167 }
1168
1169 void initServer() {
1170 int j;
1171
1172 signal(SIGHUP, SIG_IGN);
1173 signal(SIGPIPE, SIG_IGN);
1174 setupSignalHandlers();
1175
1176 if (server.syslog_enabled) {
1177 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1178 server.syslog_facility);
1179 }
1180
1181 server.current_client = NULL;
1182 server.clients = listCreate();
1183 server.clients_to_close = listCreate();
1184 server.slaves = listCreate();
1185 server.monitors = listCreate();
1186 server.unblocked_clients = listCreate();
1187
1188 createSharedObjects();
1189 adjustOpenFilesLimit();
1190 server.el = aeCreateEventLoop(server.maxclients+1024);
1191 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1192
1193 if (server.port != 0) {
1194 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1195 if (server.ipfd == ANET_ERR) {
1196 redisLog(REDIS_WARNING, "Opening port %d: %s",
1197 server.port, server.neterr);
1198 exit(1);
1199 }
1200 }
1201 if (server.unixsocket != NULL) {
1202 unlink(server.unixsocket); /* don't care if this fails */
1203 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1204 if (server.sofd == ANET_ERR) {
1205 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1206 exit(1);
1207 }
1208 }
1209 if (server.ipfd < 0 && server.sofd < 0) {
1210 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1211 exit(1);
1212 }
1213 for (j = 0; j < server.dbnum; j++) {
1214 server.db[j].dict = dictCreate(&dbDictType,NULL);
1215 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1216 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1217 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1218 server.db[j].id = j;
1219 }
1220 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1221 server.pubsub_patterns = listCreate();
1222 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1223 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1224 server.cronloops = 0;
1225 server.rdb_child_pid = -1;
1226 server.aof_child_pid = -1;
1227 server.aof_rewrite_buf = sdsempty();
1228 server.aof_buf = sdsempty();
1229 server.lastsave = time(NULL);
1230 server.dirty = 0;
1231 server.stat_numcommands = 0;
1232 server.stat_numconnections = 0;
1233 server.stat_expiredkeys = 0;
1234 server.stat_evictedkeys = 0;
1235 server.stat_starttime = time(NULL);
1236 server.stat_keyspace_misses = 0;
1237 server.stat_keyspace_hits = 0;
1238 server.stat_peak_memory = 0;
1239 server.stat_fork_time = 0;
1240 server.stat_rejected_conn = 0;
1241 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1242 server.ops_sec_idx = 0;
1243 server.ops_sec_last_sample_time = mstime();
1244 server.ops_sec_last_sample_ops = 0;
1245 server.unixtime = time(NULL);
1246 server.lastbgsave_status = REDIS_OK;
1247 server.stop_writes_on_bgsave_err = 1;
1248 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1249 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1250 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1251 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1252 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1253
1254 if (server.aof_state == REDIS_AOF_ON) {
1255 server.aof_fd = open(server.aof_filename,
1256 O_WRONLY|O_APPEND|O_CREAT,0644);
1257 if (server.aof_fd == -1) {
1258 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1259 strerror(errno));
1260 exit(1);
1261 }
1262 }
1263
1264 /* 32 bit instances are limited to 4GB of address space, so if there is
1265 * no explicit limit in the user provided configuration we set a limit
1266 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1267 * useless crashes of the Redis instance. */
1268 if (server.arch_bits == 32 && server.maxmemory == 0) {
1269 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1270 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1271 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1272 }
1273
1274 scriptingInit();
1275 slowlogInit();
1276 bioInit();
1277 }
1278
1279 /* Populates the Redis Command Table starting from the hard coded list
1280 * we have on top of redis.c file. */
1281 void populateCommandTable(void) {
1282 int j;
1283 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1284
1285 for (j = 0; j < numcommands; j++) {
1286 struct redisCommand *c = redisCommandTable+j;
1287 char *f = c->sflags;
1288 int retval;
1289
1290 while(*f != '\0') {
1291 switch(*f) {
1292 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1293 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1294 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1295 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1296 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1297 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1298 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1299 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1300 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1301 default: redisPanic("Unsupported command flag"); break;
1302 }
1303 f++;
1304 }
1305
1306 retval = dictAdd(server.commands, sdsnew(c->name), c);
1307 assert(retval == DICT_OK);
1308 }
1309 }
1310
1311 void resetCommandTableStats(void) {
1312 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1313 int j;
1314
1315 for (j = 0; j < numcommands; j++) {
1316 struct redisCommand *c = redisCommandTable+j;
1317
1318 c->microseconds = 0;
1319 c->calls = 0;
1320 }
1321 }
1322
1323 /* ========================== Redis OP Array API ============================ */
1324
1325 void redisOpArrayInit(redisOpArray *oa) {
1326 oa->ops = NULL;
1327 oa->numops = 0;
1328 }
1329
1330 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1331 robj **argv, int argc, int target)
1332 {
1333 redisOp *op;
1334
1335 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1336 op = oa->ops+oa->numops;
1337 op->cmd = cmd;
1338 op->dbid = dbid;
1339 op->argv = argv;
1340 op->argc = argc;
1341 op->target = target;
1342 oa->numops++;
1343 return oa->numops;
1344 }
1345
1346 void redisOpArrayFree(redisOpArray *oa) {
1347 while(oa->numops) {
1348 int j;
1349 redisOp *op;
1350
1351 oa->numops--;
1352 op = oa->ops+oa->numops;
1353 for (j = 0; j < op->argc; j++)
1354 decrRefCount(op->argv[j]);
1355 zfree(op->argv);
1356 }
1357 zfree(oa->ops);
1358 }
1359
1360 /* ====================== Commands lookup and execution ===================== */
1361
1362 struct redisCommand *lookupCommand(sds name) {
1363 return dictFetchValue(server.commands, name);
1364 }
1365
1366 struct redisCommand *lookupCommandByCString(char *s) {
1367 struct redisCommand *cmd;
1368 sds name = sdsnew(s);
1369
1370 cmd = dictFetchValue(server.commands, name);
1371 sdsfree(name);
1372 return cmd;
1373 }
1374
1375 /* Propagate the specified command (in the context of the specified database id)
1376 * to AOF, Slaves and Monitors.
1377 *
1378 * flags are an xor between:
1379 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1380 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1381 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1382 */
1383 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1384 int flags)
1385 {
1386 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1387 feedAppendOnlyFile(cmd,dbid,argv,argc);
1388 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1389 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1390 }
1391
1392 /* Used inside commands to schedule the propagation of additional commands
1393 * after the current command is propagated to AOF / Replication. */
1394 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1395 int target)
1396 {
1397 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1398 }
1399
1400 /* Call() is the core of Redis execution of a command */
1401 void call(redisClient *c, int flags) {
1402 long long dirty, start = ustime(), duration;
1403
1404 /* Sent the command to clients in MONITOR mode, only if the commands are
1405 * not geneated from reading an AOF. */
1406 if (listLength(server.monitors) && !server.loading)
1407 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1408
1409 /* Call the command. */
1410 redisOpArrayInit(&server.also_propagate);
1411 dirty = server.dirty;
1412 c->cmd->proc(c);
1413 dirty = server.dirty-dirty;
1414 duration = ustime()-start;
1415
1416 /* When EVAL is called loading the AOF we don't want commands called
1417 * from Lua to go into the slowlog or to populate statistics. */
1418 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1419 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1420
1421 /* Log the command into the Slow log if needed, and populate the
1422 * per-command statistics that we show in INFO commandstats. */
1423 if (flags & REDIS_CALL_SLOWLOG)
1424 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1425 if (flags & REDIS_CALL_STATS) {
1426 c->cmd->microseconds += duration;
1427 c->cmd->calls++;
1428 }
1429
1430 /* Propagate the command into the AOF and replication link */
1431 if (flags & REDIS_CALL_PROPAGATE) {
1432 int flags = REDIS_PROPAGATE_NONE;
1433
1434 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1435 flags |= REDIS_PROPAGATE_REPL;
1436 if (dirty)
1437 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1438 if (flags != REDIS_PROPAGATE_NONE)
1439 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1440 }
1441 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1442 * PUSH command. */
1443 if (server.also_propagate.numops) {
1444 int j;
1445 redisOp *rop;
1446
1447 for (j = 0; j < server.also_propagate.numops; j++) {
1448 rop = &server.also_propagate.ops[j];
1449 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1450 }
1451 redisOpArrayFree(&server.also_propagate);
1452 }
1453 server.stat_numcommands++;
1454 }
1455
1456 /* If this function gets called we already read a whole
1457 * command, argments are in the client argv/argc fields.
1458 * processCommand() execute the command or prepare the
1459 * server for a bulk read from the client.
1460 *
1461 * If 1 is returned the client is still alive and valid and
1462 * and other operations can be performed by the caller. Otherwise
1463 * if 0 is returned the client was destroied (i.e. after QUIT). */
1464 int processCommand(redisClient *c) {
1465 /* The QUIT command is handled separately. Normal command procs will
1466 * go through checking for replication and QUIT will cause trouble
1467 * when FORCE_REPLICATION is enabled and would be implemented in
1468 * a regular command proc. */
1469 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1470 addReply(c,shared.ok);
1471 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1472 return REDIS_ERR;
1473 }
1474
1475 /* Now lookup the command and check ASAP about trivial error conditions
1476 * such as wrong arity, bad command name and so forth. */
1477 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1478 if (!c->cmd) {
1479 addReplyErrorFormat(c,"unknown command '%s'",
1480 (char*)c->argv[0]->ptr);
1481 return REDIS_OK;
1482 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1483 (c->argc < -c->cmd->arity)) {
1484 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1485 c->cmd->name);
1486 return REDIS_OK;
1487 }
1488
1489 /* Check if the user is authenticated */
1490 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1491 {
1492 addReplyError(c,"operation not permitted");
1493 return REDIS_OK;
1494 }
1495
1496 /* Handle the maxmemory directive.
1497 *
1498 * First we try to free some memory if possible (if there are volatile
1499 * keys in the dataset). If there are not the only thing we can do
1500 * is returning an error. */
1501 if (server.maxmemory) {
1502 int retval = freeMemoryIfNeeded();
1503 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1504 addReply(c, shared.oomerr);
1505 return REDIS_OK;
1506 }
1507 }
1508
1509 /* Don't accept write commands if there are problems persisting on disk. */
1510 if (server.stop_writes_on_bgsave_err &&
1511 server.saveparamslen > 0
1512 && server.lastbgsave_status == REDIS_ERR &&
1513 c->cmd->flags & REDIS_CMD_WRITE)
1514 {
1515 addReply(c, shared.bgsaveerr);
1516 return REDIS_OK;
1517 }
1518
1519 /* Don't accept wirte commands if this is a read only slave. But
1520 * accept write commands if this is our master. */
1521 if (server.masterhost && server.repl_slave_ro &&
1522 !(c->flags & REDIS_MASTER) &&
1523 c->cmd->flags & REDIS_CMD_WRITE)
1524 {
1525 addReply(c, shared.roslaveerr);
1526 return REDIS_OK;
1527 }
1528
1529 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1530 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1531 &&
1532 c->cmd->proc != subscribeCommand &&
1533 c->cmd->proc != unsubscribeCommand &&
1534 c->cmd->proc != psubscribeCommand &&
1535 c->cmd->proc != punsubscribeCommand) {
1536 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1537 return REDIS_OK;
1538 }
1539
1540 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1541 * we are a slave with a broken link with master. */
1542 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1543 server.repl_serve_stale_data == 0 &&
1544 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1545 {
1546 addReplyError(c,
1547 "link with MASTER is down and slave-serve-stale-data is set to no");
1548 return REDIS_OK;
1549 }
1550
1551 /* Loading DB? Return an error if the command is not INFO */
1552 if (server.loading && c->cmd->proc != infoCommand) {
1553 addReply(c, shared.loadingerr);
1554 return REDIS_OK;
1555 }
1556
1557 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1558 if (server.lua_timedout &&
1559 !(c->cmd->proc != shutdownCommand &&
1560 c->argc == 2 &&
1561 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1562 !(c->cmd->proc == scriptCommand &&
1563 c->argc == 2 &&
1564 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1565 {
1566 addReply(c, shared.slowscripterr);
1567 return REDIS_OK;
1568 }
1569
1570 /* Exec the command */
1571 if (c->flags & REDIS_MULTI &&
1572 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1573 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1574 {
1575 queueMultiCommand(c);
1576 addReply(c,shared.queued);
1577 } else {
1578 call(c,REDIS_CALL_FULL);
1579 }
1580 return REDIS_OK;
1581 }
1582
1583 /*================================== Shutdown =============================== */
1584
1585 int prepareForShutdown(int flags) {
1586 int save = flags & REDIS_SHUTDOWN_SAVE;
1587 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1588
1589 redisLog(REDIS_WARNING,"User requested shutdown...");
1590 /* Kill the saving child if there is a background saving in progress.
1591 We want to avoid race conditions, for instance our saving child may
1592 overwrite the synchronous saving did by SHUTDOWN. */
1593 if (server.rdb_child_pid != -1) {
1594 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1595 kill(server.rdb_child_pid,SIGKILL);
1596 rdbRemoveTempFile(server.rdb_child_pid);
1597 }
1598 if (server.aof_state != REDIS_AOF_OFF) {
1599 /* Kill the AOF saving child as the AOF we already have may be longer
1600 * but contains the full dataset anyway. */
1601 if (server.aof_child_pid != -1) {
1602 redisLog(REDIS_WARNING,
1603 "There is a child rewriting the AOF. Killing it!");
1604 kill(server.aof_child_pid,SIGKILL);
1605 }
1606 /* Append only file: fsync() the AOF and exit */
1607 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1608 aof_fsync(server.aof_fd);
1609 }
1610 if ((server.saveparamslen > 0 && !nosave) || save) {
1611 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1612 /* Snapshotting. Perform a SYNC SAVE and exit */
1613 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1614 /* Ooops.. error saving! The best we can do is to continue
1615 * operating. Note that if there was a background saving process,
1616 * in the next cron() Redis will be notified that the background
1617 * saving aborted, handling special stuff like slaves pending for
1618 * synchronization... */
1619 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1620 return REDIS_ERR;
1621 }
1622 }
1623 if (server.daemonize) {
1624 redisLog(REDIS_NOTICE,"Removing the pid file.");
1625 unlink(server.pidfile);
1626 }
1627 /* Close the listening sockets. Apparently this allows faster restarts. */
1628 if (server.ipfd != -1) close(server.ipfd);
1629 if (server.sofd != -1) close(server.sofd);
1630 if (server.unixsocket) {
1631 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1632 unlink(server.unixsocket); /* don't care if this fails */
1633 }
1634
1635 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1636 return REDIS_OK;
1637 }
1638
1639 /*================================== Commands =============================== */
1640
1641 void authCommand(redisClient *c) {
1642 if (!server.requirepass) {
1643 addReplyError(c,"Client sent AUTH, but no password is set");
1644 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1645 c->authenticated = 1;
1646 addReply(c,shared.ok);
1647 } else {
1648 c->authenticated = 0;
1649 addReplyError(c,"invalid password");
1650 }
1651 }
1652
1653 void pingCommand(redisClient *c) {
1654 addReply(c,shared.pong);
1655 }
1656
1657 void echoCommand(redisClient *c) {
1658 addReplyBulk(c,c->argv[1]);
1659 }
1660
1661 void timeCommand(redisClient *c) {
1662 struct timeval tv;
1663
1664 /* gettimeofday() can only fail if &tv is a bad addresss so we
1665 * don't check for errors. */
1666 gettimeofday(&tv,NULL);
1667 addReplyMultiBulkLen(c,2);
1668 addReplyBulkLongLong(c,tv.tv_sec);
1669 addReplyBulkLongLong(c,tv.tv_usec);
1670 }
1671
1672 /* Convert an amount of bytes into a human readable string in the form
1673 * of 100B, 2G, 100M, 4K, and so forth. */
1674 void bytesToHuman(char *s, unsigned long long n) {
1675 double d;
1676
1677 if (n < 1024) {
1678 /* Bytes */
1679 sprintf(s,"%lluB",n);
1680 return;
1681 } else if (n < (1024*1024)) {
1682 d = (double)n/(1024);
1683 sprintf(s,"%.2fK",d);
1684 } else if (n < (1024LL*1024*1024)) {
1685 d = (double)n/(1024*1024);
1686 sprintf(s,"%.2fM",d);
1687 } else if (n < (1024LL*1024*1024*1024)) {
1688 d = (double)n/(1024LL*1024*1024);
1689 sprintf(s,"%.2fG",d);
1690 }
1691 }
1692
1693 /* Create the string returned by the INFO command. This is decoupled
1694 * by the INFO command itself as we need to report the same information
1695 * on memory corruption problems. */
1696 sds genRedisInfoString(char *section) {
1697 sds info = sdsempty();
1698 time_t uptime = server.unixtime-server.stat_starttime;
1699 int j, numcommands;
1700 struct rusage self_ru, c_ru;
1701 unsigned long lol, bib;
1702 int allsections = 0, defsections = 0;
1703 int sections = 0;
1704
1705 if (section) {
1706 allsections = strcasecmp(section,"all") == 0;
1707 defsections = strcasecmp(section,"default") == 0;
1708 }
1709
1710 getrusage(RUSAGE_SELF, &self_ru);
1711 getrusage(RUSAGE_CHILDREN, &c_ru);
1712 getClientsMaxBuffers(&lol,&bib);
1713
1714 /* Server */
1715 if (allsections || defsections || !strcasecmp(section,"server")) {
1716 struct utsname name;
1717
1718 if (sections++) info = sdscat(info,"\r\n");
1719 uname(&name);
1720 info = sdscatprintf(info,
1721 "# Server\r\n"
1722 "redis_version:%s\r\n"
1723 "redis_git_sha1:%s\r\n"
1724 "redis_git_dirty:%d\r\n"
1725 "os:%s %s %s\r\n"
1726 "arch_bits:%d\r\n"
1727 "multiplexing_api:%s\r\n"
1728 "gcc_version:%d.%d.%d\r\n"
1729 "process_id:%ld\r\n"
1730 "run_id:%s\r\n"
1731 "tcp_port:%d\r\n"
1732 "uptime_in_seconds:%ld\r\n"
1733 "uptime_in_days:%ld\r\n"
1734 "lru_clock:%ld\r\n",
1735 REDIS_VERSION,
1736 redisGitSHA1(),
1737 strtol(redisGitDirty(),NULL,10) > 0,
1738 name.sysname, name.release, name.machine,
1739 server.arch_bits,
1740 aeGetApiName(),
1741 #ifdef __GNUC__
1742 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1743 #else
1744 0,0,0,
1745 #endif
1746 (long) getpid(),
1747 server.runid,
1748 server.port,
1749 uptime,
1750 uptime/(3600*24),
1751 (unsigned long) server.lruclock);
1752 }
1753
1754 /* Clients */
1755 if (allsections || defsections || !strcasecmp(section,"clients")) {
1756 if (sections++) info = sdscat(info,"\r\n");
1757 info = sdscatprintf(info,
1758 "# Clients\r\n"
1759 "connected_clients:%lu\r\n"
1760 "client_longest_output_list:%lu\r\n"
1761 "client_biggest_input_buf:%lu\r\n"
1762 "blocked_clients:%d\r\n",
1763 listLength(server.clients)-listLength(server.slaves),
1764 lol, bib,
1765 server.bpop_blocked_clients);
1766 }
1767
1768 /* Memory */
1769 if (allsections || defsections || !strcasecmp(section,"memory")) {
1770 char hmem[64];
1771 char peak_hmem[64];
1772
1773 bytesToHuman(hmem,zmalloc_used_memory());
1774 bytesToHuman(peak_hmem,server.stat_peak_memory);
1775 if (sections++) info = sdscat(info,"\r\n");
1776 info = sdscatprintf(info,
1777 "# Memory\r\n"
1778 "used_memory:%zu\r\n"
1779 "used_memory_human:%s\r\n"
1780 "used_memory_rss:%zu\r\n"
1781 "used_memory_peak:%zu\r\n"
1782 "used_memory_peak_human:%s\r\n"
1783 "used_memory_lua:%lld\r\n"
1784 "mem_fragmentation_ratio:%.2f\r\n"
1785 "mem_allocator:%s\r\n",
1786 zmalloc_used_memory(),
1787 hmem,
1788 zmalloc_get_rss(),
1789 server.stat_peak_memory,
1790 peak_hmem,
1791 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1792 zmalloc_get_fragmentation_ratio(),
1793 ZMALLOC_LIB
1794 );
1795 }
1796
1797 /* Persistence */
1798 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1799 if (sections++) info = sdscat(info,"\r\n");
1800 info = sdscatprintf(info,
1801 "# Persistence\r\n"
1802 "loading:%d\r\n"
1803 "aof_enabled:%d\r\n"
1804 "changes_since_last_save:%lld\r\n"
1805 "bgsave_in_progress:%d\r\n"
1806 "last_save_time:%ld\r\n"
1807 "last_bgsave_status:%s\r\n"
1808 "bgrewriteaof_in_progress:%d\r\n"
1809 "bgrewriteaof_scheduled:%d\r\n",
1810 server.loading,
1811 server.aof_state != REDIS_AOF_OFF,
1812 server.dirty,
1813 server.rdb_child_pid != -1,
1814 server.lastsave,
1815 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1816 server.aof_child_pid != -1,
1817 server.aof_rewrite_scheduled);
1818
1819 if (server.aof_state != REDIS_AOF_OFF) {
1820 info = sdscatprintf(info,
1821 "aof_current_size:%lld\r\n"
1822 "aof_base_size:%lld\r\n"
1823 "aof_pending_rewrite:%d\r\n"
1824 "aof_buffer_length:%zu\r\n"
1825 "aof_pending_bio_fsync:%llu\r\n"
1826 "aof_delayed_fsync:%lu\r\n",
1827 (long long) server.aof_current_size,
1828 (long long) server.aof_rewrite_base_size,
1829 server.aof_rewrite_scheduled,
1830 sdslen(server.aof_buf),
1831 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1832 server.aof_delayed_fsync);
1833 }
1834
1835 if (server.loading) {
1836 double perc;
1837 time_t eta, elapsed;
1838 off_t remaining_bytes = server.loading_total_bytes-
1839 server.loading_loaded_bytes;
1840
1841 perc = ((double)server.loading_loaded_bytes /
1842 server.loading_total_bytes) * 100;
1843
1844 elapsed = server.unixtime-server.loading_start_time;
1845 if (elapsed == 0) {
1846 eta = 1; /* A fake 1 second figure if we don't have
1847 enough info */
1848 } else {
1849 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1850 }
1851
1852 info = sdscatprintf(info,
1853 "loading_start_time:%ld\r\n"
1854 "loading_total_bytes:%llu\r\n"
1855 "loading_loaded_bytes:%llu\r\n"
1856 "loading_loaded_perc:%.2f\r\n"
1857 "loading_eta_seconds:%ld\r\n"
1858 ,(unsigned long) server.loading_start_time,
1859 (unsigned long long) server.loading_total_bytes,
1860 (unsigned long long) server.loading_loaded_bytes,
1861 perc,
1862 eta
1863 );
1864 }
1865 }
1866
1867 /* Stats */
1868 if (allsections || defsections || !strcasecmp(section,"stats")) {
1869 if (sections++) info = sdscat(info,"\r\n");
1870 info = sdscatprintf(info,
1871 "# Stats\r\n"
1872 "total_connections_received:%lld\r\n"
1873 "total_commands_processed:%lld\r\n"
1874 "instantaneous_ops_per_sec:%lld\r\n"
1875 "rejected_connections:%lld\r\n"
1876 "expired_keys:%lld\r\n"
1877 "evicted_keys:%lld\r\n"
1878 "keyspace_hits:%lld\r\n"
1879 "keyspace_misses:%lld\r\n"
1880 "pubsub_channels:%ld\r\n"
1881 "pubsub_patterns:%lu\r\n"
1882 "latest_fork_usec:%lld\r\n",
1883 server.stat_numconnections,
1884 server.stat_numcommands,
1885 getOperationsPerSecond(),
1886 server.stat_rejected_conn,
1887 server.stat_expiredkeys,
1888 server.stat_evictedkeys,
1889 server.stat_keyspace_hits,
1890 server.stat_keyspace_misses,
1891 dictSize(server.pubsub_channels),
1892 listLength(server.pubsub_patterns),
1893 server.stat_fork_time);
1894 }
1895
1896 /* Replication */
1897 if (allsections || defsections || !strcasecmp(section,"replication")) {
1898 if (sections++) info = sdscat(info,"\r\n");
1899 info = sdscatprintf(info,
1900 "# Replication\r\n"
1901 "role:%s\r\n",
1902 server.masterhost == NULL ? "master" : "slave");
1903 if (server.masterhost) {
1904 info = sdscatprintf(info,
1905 "master_host:%s\r\n"
1906 "master_port:%d\r\n"
1907 "master_link_status:%s\r\n"
1908 "master_last_io_seconds_ago:%d\r\n"
1909 "master_sync_in_progress:%d\r\n"
1910 ,server.masterhost,
1911 server.masterport,
1912 (server.repl_state == REDIS_REPL_CONNECTED) ?
1913 "up" : "down",
1914 server.master ?
1915 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
1916 server.repl_state == REDIS_REPL_TRANSFER
1917 );
1918
1919 if (server.repl_state == REDIS_REPL_TRANSFER) {
1920 info = sdscatprintf(info,
1921 "master_sync_left_bytes:%ld\r\n"
1922 "master_sync_last_io_seconds_ago:%d\r\n"
1923 ,(long)server.repl_transfer_left,
1924 (int)(server.unixtime-server.repl_transfer_lastio)
1925 );
1926 }
1927
1928 if (server.repl_state != REDIS_REPL_CONNECTED) {
1929 info = sdscatprintf(info,
1930 "master_link_down_since_seconds:%ld\r\n",
1931 (long)server.unixtime-server.repl_down_since);
1932 }
1933 }
1934 info = sdscatprintf(info,
1935 "connected_slaves:%lu\r\n",
1936 listLength(server.slaves));
1937 if (listLength(server.slaves)) {
1938 int slaveid = 0;
1939 listNode *ln;
1940 listIter li;
1941
1942 listRewind(server.slaves,&li);
1943 while((ln = listNext(&li))) {
1944 redisClient *slave = listNodeValue(ln);
1945 char *state = NULL;
1946 char ip[32];
1947 int port;
1948
1949 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
1950 switch(slave->replstate) {
1951 case REDIS_REPL_WAIT_BGSAVE_START:
1952 case REDIS_REPL_WAIT_BGSAVE_END:
1953 state = "wait_bgsave";
1954 break;
1955 case REDIS_REPL_SEND_BULK:
1956 state = "send_bulk";
1957 break;
1958 case REDIS_REPL_ONLINE:
1959 state = "online";
1960 break;
1961 }
1962 if (state == NULL) continue;
1963 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
1964 slaveid,ip,port,state);
1965 slaveid++;
1966 }
1967 }
1968 }
1969
1970 /* CPU */
1971 if (allsections || defsections || !strcasecmp(section,"cpu")) {
1972 if (sections++) info = sdscat(info,"\r\n");
1973 info = sdscatprintf(info,
1974 "# CPU\r\n"
1975 "used_cpu_sys:%.2f\r\n"
1976 "used_cpu_user:%.2f\r\n"
1977 "used_cpu_sys_children:%.2f\r\n"
1978 "used_cpu_user_children:%.2f\r\n",
1979 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
1980 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
1981 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
1982 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
1983 }
1984
1985 /* cmdtime */
1986 if (allsections || !strcasecmp(section,"commandstats")) {
1987 if (sections++) info = sdscat(info,"\r\n");
1988 info = sdscatprintf(info, "# Commandstats\r\n");
1989 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1990 for (j = 0; j < numcommands; j++) {
1991 struct redisCommand *c = redisCommandTable+j;
1992
1993 if (!c->calls) continue;
1994 info = sdscatprintf(info,
1995 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
1996 c->name, c->calls, c->microseconds,
1997 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
1998 }
1999 }
2000
2001 /* Key space */
2002 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2003 if (sections++) info = sdscat(info,"\r\n");
2004 info = sdscatprintf(info, "# Keyspace\r\n");
2005 for (j = 0; j < server.dbnum; j++) {
2006 long long keys, vkeys;
2007
2008 keys = dictSize(server.db[j].dict);
2009 vkeys = dictSize(server.db[j].expires);
2010 if (keys || vkeys) {
2011 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2012 j, keys, vkeys);
2013 }
2014 }
2015 }
2016 return info;
2017 }
2018
2019 void infoCommand(redisClient *c) {
2020 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2021
2022 if (c->argc > 2) {
2023 addReply(c,shared.syntaxerr);
2024 return;
2025 }
2026 sds info = genRedisInfoString(section);
2027 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2028 (unsigned long)sdslen(info)));
2029 addReplySds(c,info);
2030 addReply(c,shared.crlf);
2031 }
2032
2033 void monitorCommand(redisClient *c) {
2034 /* ignore MONITOR if aleady slave or in monitor mode */
2035 if (c->flags & REDIS_SLAVE) return;
2036
2037 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2038 c->slaveseldb = 0;
2039 listAddNodeTail(server.monitors,c);
2040 addReply(c,shared.ok);
2041 }
2042
2043 /* ============================ Maxmemory directive ======================== */
2044
2045 /* This function gets called when 'maxmemory' is set on the config file to limit
2046 * the max memory used by the server, before processing a command.
2047 *
2048 * The goal of the function is to free enough memory to keep Redis under the
2049 * configured memory limit.
2050 *
2051 * The function starts calculating how many bytes should be freed to keep
2052 * Redis under the limit, and enters a loop selecting the best keys to
2053 * evict accordingly to the configured policy.
2054 *
2055 * If all the bytes needed to return back under the limit were freed the
2056 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2057 * should block the execution of commands that will result in more memory
2058 * used by the server.
2059 */
2060 int freeMemoryIfNeeded(void) {
2061 size_t mem_used, mem_tofree, mem_freed;
2062 int slaves = listLength(server.slaves);
2063
2064 /* Remove the size of slaves output buffers and AOF buffer from the
2065 * count of used memory. */
2066 mem_used = zmalloc_used_memory();
2067 if (slaves) {
2068 listIter li;
2069 listNode *ln;
2070
2071 listRewind(server.slaves,&li);
2072 while((ln = listNext(&li))) {
2073 redisClient *slave = listNodeValue(ln);
2074 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2075 if (obuf_bytes > mem_used)
2076 mem_used = 0;
2077 else
2078 mem_used -= obuf_bytes;
2079 }
2080 }
2081 if (server.aof_state != REDIS_AOF_OFF) {
2082 mem_used -= sdslen(server.aof_buf);
2083 mem_used -= sdslen(server.aof_rewrite_buf);
2084 }
2085
2086 /* Check if we are over the memory limit. */
2087 if (mem_used <= server.maxmemory) return REDIS_OK;
2088
2089 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2090 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2091
2092 /* Compute how much memory we need to free. */
2093 mem_tofree = mem_used - server.maxmemory;
2094 mem_freed = 0;
2095 while (mem_freed < mem_tofree) {
2096 int j, k, keys_freed = 0;
2097
2098 for (j = 0; j < server.dbnum; j++) {
2099 long bestval = 0; /* just to prevent warning */
2100 sds bestkey = NULL;
2101 struct dictEntry *de;
2102 redisDb *db = server.db+j;
2103 dict *dict;
2104
2105 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2106 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2107 {
2108 dict = server.db[j].dict;
2109 } else {
2110 dict = server.db[j].expires;
2111 }
2112 if (dictSize(dict) == 0) continue;
2113
2114 /* volatile-random and allkeys-random policy */
2115 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2116 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2117 {
2118 de = dictGetRandomKey(dict);
2119 bestkey = dictGetKey(de);
2120 }
2121
2122 /* volatile-lru and allkeys-lru policy */
2123 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2124 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2125 {
2126 for (k = 0; k < server.maxmemory_samples; k++) {
2127 sds thiskey;
2128 long thisval;
2129 robj *o;
2130
2131 de = dictGetRandomKey(dict);
2132 thiskey = dictGetKey(de);
2133 /* When policy is volatile-lru we need an additonal lookup
2134 * to locate the real key, as dict is set to db->expires. */
2135 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2136 de = dictFind(db->dict, thiskey);
2137 o = dictGetVal(de);
2138 thisval = estimateObjectIdleTime(o);
2139
2140 /* Higher idle time is better candidate for deletion */
2141 if (bestkey == NULL || thisval > bestval) {
2142 bestkey = thiskey;
2143 bestval = thisval;
2144 }
2145 }
2146 }
2147
2148 /* volatile-ttl */
2149 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2150 for (k = 0; k < server.maxmemory_samples; k++) {
2151 sds thiskey;
2152 long thisval;
2153
2154 de = dictGetRandomKey(dict);
2155 thiskey = dictGetKey(de);
2156 thisval = (long) dictGetVal(de);
2157
2158 /* Expire sooner (minor expire unix timestamp) is better
2159 * candidate for deletion */
2160 if (bestkey == NULL || thisval < bestval) {
2161 bestkey = thiskey;
2162 bestval = thisval;
2163 }
2164 }
2165 }
2166
2167 /* Finally remove the selected key. */
2168 if (bestkey) {
2169 long long delta;
2170
2171 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2172 propagateExpire(db,keyobj);
2173 /* We compute the amount of memory freed by dbDelete() alone.
2174 * It is possible that actually the memory needed to propagate
2175 * the DEL in AOF and replication link is greater than the one
2176 * we are freeing removing the key, but we can't account for
2177 * that otherwise we would never exit the loop.
2178 *
2179 * AOF and Output buffer memory will be freed eventually so
2180 * we only care about memory used by the key space. */
2181 delta = (long long) zmalloc_used_memory();
2182 dbDelete(db,keyobj);
2183 delta -= (long long) zmalloc_used_memory();
2184 mem_freed += delta;
2185 server.stat_evictedkeys++;
2186 decrRefCount(keyobj);
2187 keys_freed++;
2188
2189 /* When the memory to free starts to be big enough, we may
2190 * start spending so much time here that is impossible to
2191 * deliver data to the slaves fast enough, so we force the
2192 * transmission here inside the loop. */
2193 if (slaves) flushSlavesOutputBuffers();
2194 }
2195 }
2196 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2197 }
2198 return REDIS_OK;
2199 }
2200
2201 /* =================================== Main! ================================ */
2202
2203 #ifdef __linux__
2204 int linuxOvercommitMemoryValue(void) {
2205 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2206 char buf[64];
2207
2208 if (!fp) return -1;
2209 if (fgets(buf,64,fp) == NULL) {
2210 fclose(fp);
2211 return -1;
2212 }
2213 fclose(fp);
2214
2215 return atoi(buf);
2216 }
2217
2218 void linuxOvercommitMemoryWarning(void) {
2219 if (linuxOvercommitMemoryValue() == 0) {
2220 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.");
2221 }
2222 }
2223 #endif /* __linux__ */
2224
2225 void createPidFile(void) {
2226 /* Try to write the pid file in a best-effort way. */
2227 FILE *fp = fopen(server.pidfile,"w");
2228 if (fp) {
2229 fprintf(fp,"%d\n",(int)getpid());
2230 fclose(fp);
2231 }
2232 }
2233
2234 void daemonize(void) {
2235 int fd;
2236
2237 if (fork() != 0) exit(0); /* parent exits */
2238 setsid(); /* create a new session */
2239
2240 /* Every output goes to /dev/null. If Redis is daemonized but
2241 * the 'logfile' is set to 'stdout' in the configuration file
2242 * it will not log at all. */
2243 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2244 dup2(fd, STDIN_FILENO);
2245 dup2(fd, STDOUT_FILENO);
2246 dup2(fd, STDERR_FILENO);
2247 if (fd > STDERR_FILENO) close(fd);
2248 }
2249 }
2250
2251 void version() {
2252 printf("Redis server v=%s sha=%s:%d malloc=%s\n", REDIS_VERSION,
2253 redisGitSHA1(), atoi(redisGitDirty()) > 0, ZMALLOC_LIB);
2254 exit(0);
2255 }
2256
2257 void usage() {
2258 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2259 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2260 fprintf(stderr," ./redis-server -v or --version\n");
2261 fprintf(stderr," ./redis-server -h or --help\n");
2262 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2263 fprintf(stderr,"Examples:\n");
2264 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2265 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2266 fprintf(stderr," ./redis-server --port 7777\n");
2267 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2268 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2269 exit(1);
2270 }
2271
2272 void redisAsciiArt(void) {
2273 #include "asciilogo.h"
2274 char *buf = zmalloc(1024*16);
2275
2276 snprintf(buf,1024*16,ascii_logo,
2277 REDIS_VERSION,
2278 redisGitSHA1(),
2279 strtol(redisGitDirty(),NULL,10) > 0,
2280 (sizeof(long) == 8) ? "64" : "32",
2281 "stand alone",
2282 server.port,
2283 (long) getpid()
2284 );
2285 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2286 zfree(buf);
2287 }
2288
2289 static void sigtermHandler(int sig) {
2290 REDIS_NOTUSED(sig);
2291
2292 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2293 server.shutdown_asap = 1;
2294 }
2295
2296 void setupSignalHandlers(void) {
2297 struct sigaction act;
2298 stack_t stack;
2299
2300 stack.ss_sp = altstack;
2301 stack.ss_flags = 0;
2302 stack.ss_size = SIGSTKSZ;
2303
2304 sigaltstack(&stack, NULL);
2305
2306 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2307 * Otherwise, sa_handler is used. */
2308 sigemptyset(&act.sa_mask);
2309 act.sa_flags = 0;
2310 act.sa_handler = sigtermHandler;
2311 sigaction(SIGTERM, &act, NULL);
2312
2313 #ifdef HAVE_BACKTRACE
2314 /* Use alternate stack so we don't clobber stack in case of segv, or when we run out of stack ..
2315 * also resethand & nodefer so we can get interrupted (and killed) if we cause SEGV during SEGV handler */
2316 sigemptyset(&act.sa_mask);
2317 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
2318 act.sa_sigaction = sigsegvHandler;
2319 sigaction(SIGSEGV, &act, NULL);
2320 sigaction(SIGBUS, &act, NULL);
2321 sigaction(SIGFPE, &act, NULL);
2322 sigaction(SIGILL, &act, NULL);
2323 #endif
2324 return;
2325 }
2326
2327 void memtest(size_t megabytes, int passes);
2328
2329 int main(int argc, char **argv) {
2330 long long start;
2331 struct timeval tv;
2332
2333 /* We need to initialize our libraries, and the server configuration. */
2334 zmalloc_enable_thread_safeness();
2335 srand(time(NULL)^getpid());
2336 gettimeofday(&tv,NULL);
2337 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2338 initServerConfig();
2339
2340 if (argc >= 2) {
2341 int j = 1; /* First option to parse in argv[] */
2342 sds options = sdsempty();
2343 char *configfile = NULL;
2344
2345 /* Handle special options --help and --version */
2346 if (strcmp(argv[1], "-v") == 0 ||
2347 strcmp(argv[1], "--version") == 0) version();
2348 if (strcmp(argv[1], "--help") == 0 ||
2349 strcmp(argv[1], "-h") == 0) usage();
2350 if (strcmp(argv[1], "--test-memory") == 0) {
2351 if (argc == 3) {
2352 memtest(atoi(argv[2]),50);
2353 exit(0);
2354 } else {
2355 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2356 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2357 exit(1);
2358 }
2359 }
2360
2361 /* First argument is the config file name? */
2362 if (argv[j][0] != '-' || argv[j][1] != '-')
2363 configfile = argv[j++];
2364 /* All the other options are parsed and conceptually appended to the
2365 * configuration file. For instance --port 6380 will generate the
2366 * string "port 6380\n" to be parsed after the actual file name
2367 * is parsed, if any. */
2368 while(j != argc) {
2369 if (argv[j][0] == '-' && argv[j][1] == '-') {
2370 /* Option name */
2371 if (sdslen(options)) options = sdscat(options,"\n");
2372 options = sdscat(options,argv[j]+2);
2373 options = sdscat(options," ");
2374 } else {
2375 /* Option argument */
2376 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2377 options = sdscat(options," ");
2378 }
2379 j++;
2380 }
2381 resetServerSaveParams();
2382 loadServerConfig(configfile,options);
2383 sdsfree(options);
2384 } else {
2385 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'");
2386 }
2387 if (server.daemonize) daemonize();
2388 initServer();
2389 if (server.daemonize) createPidFile();
2390 redisAsciiArt();
2391 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2392 #ifdef __linux__
2393 linuxOvercommitMemoryWarning();
2394 #endif
2395 start = ustime();
2396 if (server.aof_state == REDIS_AOF_ON) {
2397 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2398 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2399 } else {
2400 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2401 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2402 (float)(ustime()-start)/1000000);
2403 } else if (errno != ENOENT) {
2404 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2405 exit(1);
2406 }
2407 }
2408 if (server.ipfd > 0)
2409 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2410 if (server.sofd > 0)
2411 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2412 aeSetBeforeSleepProc(server.el,beforeSleep);
2413 aeMain(server.el);
2414 aeDeleteEventLoop(server.el);
2415 return 0;
2416 }
2417
2418 /* The End */