]> git.saurik.com Git - redis.git/blame_incremental - src/redis.c
Make clear that contributing code to the Redis project means to release it under...
[redis.git] / src / redis.c
... / ...
CommitLineData
1/*
2 * Copyright (c) 2009-2012, 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
55struct 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
61double R_Zero, R_PosInf, R_NegInf, R_Nan;
62
63/*================================= Globals ================================= */
64
65/* Global vars */
66struct redisServer server; /* server global state */
67struct redisCommand *commandTable;
68
69/* Our command table.
70 *
71 * Every entry is composed of the following fields:
72 *
73 * name: a string representing the command name.
74 * function: pointer to the C function implementing the command.
75 * arity: number of arguments, it is possible to use -N to say >= N
76 * sflags: command flags as string. See below for a table of flags.
77 * flags: flags as bitmask. Computed by Redis using the 'sflags' field.
78 * get_keys_proc: an optional function to get key arguments from a command.
79 * This is only used when the following three fields are not
80 * enough to specify what arguments are keys.
81 * first_key_index: first argument that is a key
82 * last_key_index: last argument that is a key
83 * key_step: step to get all the keys from first to last argument. For instance
84 * in MSET the step is two since arguments are key,val,key,val,...
85 * microseconds: microseconds of total execution time for this command.
86 * calls: total number of calls of this command.
87 *
88 * The flags, microseconds and calls fields are computed by Redis and should
89 * always be set to zero.
90 *
91 * Command flags are expressed using strings where every character represents
92 * a flag. Later the populateCommandTable() function will take care of
93 * populating the real 'flags' field using this characters.
94 *
95 * This is the meaning of the flags:
96 *
97 * w: write command (may modify the key space).
98 * r: read command (will never modify the key space).
99 * m: may increase memory usage once called. Don't allow if out of memory.
100 * a: admin command, like SAVE or SHUTDOWN.
101 * p: Pub/Sub related command.
102 * f: force replication of this command, regarless of server.dirty.
103 * s: command not allowed in scripts.
104 * R: random command. Command is not deterministic, that is, the same command
105 * with the same arguments, with the same key space, may have different
106 * results. For instance SPOP and RANDOMKEY are two random commands.
107 * S: Sort command output array if called from script, so that the output
108 * is deterministic.
109 * l: Allow command while loading the database.
110 * t: Allow command while a slave has stale data but is not allowed to
111 * server this data. Normally no command is accepted in this condition
112 * but just a few.
113 * M: Do not automatically propagate the command on MONITOR.
114 */
115struct redisCommand redisCommandTable[] = {
116 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
117 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
118 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
119 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
120 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
121 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
122 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
123 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
124 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
125 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
126 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
127 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
128 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
129 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
130 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
131 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
132 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
133 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
134 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
135 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
136 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
137 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
138 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
139 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
140 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
141 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
142 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
143 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
144 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
145 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
146 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
147 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
148 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
149 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
150 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
151 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
152 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
153 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
154 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
155 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
156 {"srandmember",srandmemberCommand,-2,"rR",0,NULL,1,1,1,0,0},
157 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
158 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
159 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
160 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
161 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
162 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
163 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
164 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
165 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
166 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
167 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
168 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
169 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
170 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
171 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
172 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
173 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
174 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
175 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
176 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
177 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
178 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
179 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
180 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
182 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
183 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
184 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
185 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
186 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
187 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
188 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
189 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
190 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
191 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
192 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
193 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
194 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
195 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
196 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
197 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
198 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
199 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
200 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
201 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
202 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
203 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
204 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
205 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
206 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
207 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
208 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
209 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
210 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
211 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
212 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
213 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
214 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
215 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
216 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
217 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
218 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
219 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
220 {"exec",execCommand,1,"sM",0,NULL,0,0,0,0,0},
221 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
222 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
223 {"replconf",replconfCommand,-1,"ars",0,NULL,0,0,0,0,0},
224 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
225 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
226 {"sort",sortCommand,-2,"wm",0,NULL,1,1,1,0,0},
227 {"info",infoCommand,-1,"rlt",0,NULL,0,0,0,0,0},
228 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
229 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
230 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
231 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
232 {"slaveof",slaveofCommand,3,"ast",0,NULL,0,0,0,0,0},
233 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
234 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
235 {"subscribe",subscribeCommand,-2,"rpslt",0,NULL,0,0,0,0,0},
236 {"unsubscribe",unsubscribeCommand,-1,"rpslt",0,NULL,0,0,0,0,0},
237 {"psubscribe",psubscribeCommand,-2,"rpslt",0,NULL,0,0,0,0,0},
238 {"punsubscribe",punsubscribeCommand,-1,"rpslt",0,NULL,0,0,0,0,0},
239 {"publish",publishCommand,3,"pflt",0,NULL,0,0,0,0,0},
240 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
241 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
242 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
243 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
244 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
245 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
246 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
247 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
248 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
249 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
250 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
251 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0},
252 {"bitop",bitopCommand,-4,"wm",0,NULL,2,-1,1,0,0},
253 {"bitcount",bitcountCommand,-2,"r",0,NULL,1,1,1,0,0}
254};
255
256/*============================ Utility functions ============================ */
257
258/* Low level logging. To use only for very big messages, otherwise
259 * redisLog() is to prefer. */
260void redisLogRaw(int level, const char *msg) {
261 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
262 const char *c = ".-*#";
263 FILE *fp;
264 char buf[64];
265 int rawmode = (level & REDIS_LOG_RAW);
266
267 level &= 0xff; /* clear flags */
268 if (level < server.verbosity) return;
269
270 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
271 if (!fp) return;
272
273 if (rawmode) {
274 fprintf(fp,"%s",msg);
275 } else {
276 int off;
277 struct timeval tv;
278
279 gettimeofday(&tv,NULL);
280 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
281 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
282 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
283 }
284 fflush(fp);
285
286 if (server.logfile) fclose(fp);
287
288 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
289}
290
291/* Like redisLogRaw() but with printf-alike support. This is the funciton that
292 * is used across the code. The raw version is only used in order to dump
293 * the INFO output on crash. */
294void redisLog(int level, const char *fmt, ...) {
295 va_list ap;
296 char msg[REDIS_MAX_LOGMSG_LEN];
297
298 if ((level&0xff) < server.verbosity) return;
299
300 va_start(ap, fmt);
301 vsnprintf(msg, sizeof(msg), fmt, ap);
302 va_end(ap);
303
304 redisLogRaw(level,msg);
305}
306
307/* Log a fixed message without printf-alike capabilities, in a way that is
308 * safe to call from a signal handler.
309 *
310 * We actually use this only for signals that are not fatal from the point
311 * of view of Redis. Signals that are going to kill the server anyway and
312 * where we need printf-alike features are served by redisLog(). */
313void redisLogFromHandler(int level, const char *msg) {
314 int fd;
315 char buf[64];
316
317 if ((level&0xff) < server.verbosity ||
318 (server.logfile == NULL && server.daemonize)) return;
319 fd = server.logfile ?
320 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
321 STDOUT_FILENO;
322 if (fd == -1) return;
323 ll2string(buf,sizeof(buf),getpid());
324 if (write(fd,"[",1) == -1) goto err;
325 if (write(fd,buf,strlen(buf)) == -1) goto err;
326 if (write(fd," | signal handler] (",20) == -1) goto err;
327 ll2string(buf,sizeof(buf),time(NULL));
328 if (write(fd,buf,strlen(buf)) == -1) goto err;
329 if (write(fd,") ",2) == -1) goto err;
330 if (write(fd,msg,strlen(msg)) == -1) goto err;
331 if (write(fd,"\n",1) == -1) goto err;
332err:
333 if (server.logfile) close(fd);
334}
335
336/* Return the UNIX time in microseconds */
337long long ustime(void) {
338 struct timeval tv;
339 long long ust;
340
341 gettimeofday(&tv, NULL);
342 ust = ((long long)tv.tv_sec)*1000000;
343 ust += tv.tv_usec;
344 return ust;
345}
346
347/* Return the UNIX time in milliseconds */
348long long mstime(void) {
349 return ustime()/1000;
350}
351
352/* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
353 * exit(), because the latter may interact with the same file objects used by
354 * the parent process. However if we are testing the coverage normal exit() is
355 * used in order to obtain the right coverage information. */
356void exitFromChild(int retcode) {
357#ifdef COVERAGE_TEST
358 exit(retcode);
359#else
360 _exit(retcode);
361#endif
362}
363
364/*====================== Hash table type implementation ==================== */
365
366/* This is an hash table type that uses the SDS dynamic strings libary as
367 * keys and radis objects as values (objects can hold SDS strings,
368 * lists, sets). */
369
370void dictVanillaFree(void *privdata, void *val)
371{
372 DICT_NOTUSED(privdata);
373 zfree(val);
374}
375
376void dictListDestructor(void *privdata, void *val)
377{
378 DICT_NOTUSED(privdata);
379 listRelease((list*)val);
380}
381
382int dictSdsKeyCompare(void *privdata, const void *key1,
383 const void *key2)
384{
385 int l1,l2;
386 DICT_NOTUSED(privdata);
387
388 l1 = sdslen((sds)key1);
389 l2 = sdslen((sds)key2);
390 if (l1 != l2) return 0;
391 return memcmp(key1, key2, l1) == 0;
392}
393
394/* A case insensitive version used for the command lookup table. */
395int dictSdsKeyCaseCompare(void *privdata, const void *key1,
396 const void *key2)
397{
398 DICT_NOTUSED(privdata);
399
400 return strcasecmp(key1, key2) == 0;
401}
402
403void dictRedisObjectDestructor(void *privdata, void *val)
404{
405 DICT_NOTUSED(privdata);
406
407 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
408 decrRefCount(val);
409}
410
411void dictSdsDestructor(void *privdata, void *val)
412{
413 DICT_NOTUSED(privdata);
414
415 sdsfree(val);
416}
417
418int dictObjKeyCompare(void *privdata, const void *key1,
419 const void *key2)
420{
421 const robj *o1 = key1, *o2 = key2;
422 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
423}
424
425unsigned int dictObjHash(const void *key) {
426 const robj *o = key;
427 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
428}
429
430unsigned int dictSdsHash(const void *key) {
431 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
432}
433
434unsigned int dictSdsCaseHash(const void *key) {
435 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
436}
437
438int dictEncObjKeyCompare(void *privdata, const void *key1,
439 const void *key2)
440{
441 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
442 int cmp;
443
444 if (o1->encoding == REDIS_ENCODING_INT &&
445 o2->encoding == REDIS_ENCODING_INT)
446 return o1->ptr == o2->ptr;
447
448 o1 = getDecodedObject(o1);
449 o2 = getDecodedObject(o2);
450 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
451 decrRefCount(o1);
452 decrRefCount(o2);
453 return cmp;
454}
455
456unsigned int dictEncObjHash(const void *key) {
457 robj *o = (robj*) key;
458
459 if (o->encoding == REDIS_ENCODING_RAW) {
460 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
461 } else {
462 if (o->encoding == REDIS_ENCODING_INT) {
463 char buf[32];
464 int len;
465
466 len = ll2string(buf,32,(long)o->ptr);
467 return dictGenHashFunction((unsigned char*)buf, len);
468 } else {
469 unsigned int hash;
470
471 o = getDecodedObject(o);
472 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
473 decrRefCount(o);
474 return hash;
475 }
476 }
477}
478
479/* Sets type hash table */
480dictType setDictType = {
481 dictEncObjHash, /* hash function */
482 NULL, /* key dup */
483 NULL, /* val dup */
484 dictEncObjKeyCompare, /* key compare */
485 dictRedisObjectDestructor, /* key destructor */
486 NULL /* val destructor */
487};
488
489/* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
490dictType zsetDictType = {
491 dictEncObjHash, /* hash function */
492 NULL, /* key dup */
493 NULL, /* val dup */
494 dictEncObjKeyCompare, /* key compare */
495 dictRedisObjectDestructor, /* key destructor */
496 NULL /* val destructor */
497};
498
499/* Db->dict, keys are sds strings, vals are Redis objects. */
500dictType dbDictType = {
501 dictSdsHash, /* hash function */
502 NULL, /* key dup */
503 NULL, /* val dup */
504 dictSdsKeyCompare, /* key compare */
505 dictSdsDestructor, /* key destructor */
506 dictRedisObjectDestructor /* val destructor */
507};
508
509/* Db->expires */
510dictType keyptrDictType = {
511 dictSdsHash, /* hash function */
512 NULL, /* key dup */
513 NULL, /* val dup */
514 dictSdsKeyCompare, /* key compare */
515 NULL, /* key destructor */
516 NULL /* val destructor */
517};
518
519/* Command table. sds string -> command struct pointer. */
520dictType commandTableDictType = {
521 dictSdsCaseHash, /* hash function */
522 NULL, /* key dup */
523 NULL, /* val dup */
524 dictSdsKeyCaseCompare, /* key compare */
525 dictSdsDestructor, /* key destructor */
526 NULL /* val destructor */
527};
528
529/* Hash type hash table (note that small hashes are represented with zimpaps) */
530dictType hashDictType = {
531 dictEncObjHash, /* hash function */
532 NULL, /* key dup */
533 NULL, /* val dup */
534 dictEncObjKeyCompare, /* key compare */
535 dictRedisObjectDestructor, /* key destructor */
536 dictRedisObjectDestructor /* val destructor */
537};
538
539/* Keylist hash table type has unencoded redis objects as keys and
540 * lists as values. It's used for blocking operations (BLPOP) and to
541 * map swapped keys to a list of clients waiting for this keys to be loaded. */
542dictType keylistDictType = {
543 dictObjHash, /* hash function */
544 NULL, /* key dup */
545 NULL, /* val dup */
546 dictObjKeyCompare, /* key compare */
547 dictRedisObjectDestructor, /* key destructor */
548 dictListDestructor /* val destructor */
549};
550
551int htNeedsResize(dict *dict) {
552 long long size, used;
553
554 size = dictSlots(dict);
555 used = dictSize(dict);
556 return (size && used && size > DICT_HT_INITIAL_SIZE &&
557 (used*100/size < REDIS_HT_MINFILL));
558}
559
560/* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
561 * we resize the hash table to save memory */
562void tryResizeHashTables(void) {
563 int j;
564
565 for (j = 0; j < server.dbnum; j++) {
566 if (htNeedsResize(server.db[j].dict))
567 dictResize(server.db[j].dict);
568 if (htNeedsResize(server.db[j].expires))
569 dictResize(server.db[j].expires);
570 }
571}
572
573/* Our hash table implementation performs rehashing incrementally while
574 * we write/read from the hash table. Still if the server is idle, the hash
575 * table will use two tables for a long time. So we try to use 1 millisecond
576 * of CPU time at every serverCron() loop in order to rehash some key. */
577void incrementallyRehash(void) {
578 int j;
579
580 for (j = 0; j < server.dbnum; j++) {
581 /* Keys dictionary */
582 if (dictIsRehashing(server.db[j].dict)) {
583 dictRehashMilliseconds(server.db[j].dict,1);
584 break; /* already used our millisecond for this loop... */
585 }
586 /* Expires */
587 if (dictIsRehashing(server.db[j].expires)) {
588 dictRehashMilliseconds(server.db[j].expires,1);
589 break; /* already used our millisecond for this loop... */
590 }
591 }
592}
593
594/* This function is called once a background process of some kind terminates,
595 * as we want to avoid resizing the hash tables when there is a child in order
596 * to play well with copy-on-write (otherwise when a resize happens lots of
597 * memory pages are copied). The goal of this function is to update the ability
598 * for dict.c to resize the hash tables accordingly to the fact we have o not
599 * running childs. */
600void updateDictResizePolicy(void) {
601 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
602 dictEnableResize();
603 else
604 dictDisableResize();
605}
606
607/* ======================= Cron: called every 100 ms ======================== */
608
609/* Try to expire a few timed out keys. The algorithm used is adaptive and
610 * will use few CPU cycles if there are few expiring keys, otherwise
611 * it will get more aggressive to avoid that too much memory is used by
612 * keys that can be removed from the keyspace. */
613void activeExpireCycle(void) {
614 int j, iteration = 0;
615 long long start = ustime(), timelimit;
616
617 /* We can use at max REDIS_EXPIRELOOKUPS_TIME_PERC percentage of CPU time
618 * per iteration. Since this function gets called with a frequency of
619 * REDIS_HZ times per second, the following is the max amount of
620 * microseconds we can spend in this function. */
621 timelimit = 1000000*REDIS_EXPIRELOOKUPS_TIME_PERC/REDIS_HZ/100;
622 if (timelimit <= 0) timelimit = 1;
623
624 for (j = 0; j < server.dbnum; j++) {
625 int expired;
626 redisDb *db = server.db+j;
627
628 /* Continue to expire if at the end of the cycle more than 25%
629 * of the keys were expired. */
630 do {
631 unsigned long num = dictSize(db->expires);
632 unsigned long slots = dictSlots(db->expires);
633 long long now = mstime();
634
635 /* When there are less than 1% filled slots getting random
636 * keys is expensive, so stop here waiting for better times...
637 * The dictionary will be resized asap. */
638 if (num && slots > DICT_HT_INITIAL_SIZE &&
639 (num*100/slots < 1)) break;
640
641 /* The main collection cycle. Sample random keys among keys
642 * with an expire set, checking for expired ones. */
643 expired = 0;
644 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
645 num = REDIS_EXPIRELOOKUPS_PER_CRON;
646 while (num--) {
647 dictEntry *de;
648 long long t;
649
650 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
651 t = dictGetSignedIntegerVal(de);
652 if (now > t) {
653 sds key = dictGetKey(de);
654 robj *keyobj = createStringObject(key,sdslen(key));
655
656 propagateExpire(db,keyobj);
657 dbDelete(db,keyobj);
658 decrRefCount(keyobj);
659 expired++;
660 server.stat_expiredkeys++;
661 }
662 }
663 /* We can't block forever here even if there are many keys to
664 * expire. So after a given amount of milliseconds return to the
665 * caller waiting for the other active expire cycle. */
666 iteration++;
667 if ((iteration & 0xf) == 0 && /* check once every 16 cycles. */
668 (ustime()-start) > timelimit) return;
669 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
670 }
671}
672
673void updateLRUClock(void) {
674 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
675 REDIS_LRU_CLOCK_MAX;
676}
677
678
679/* Add a sample to the operations per second array of samples. */
680void trackOperationsPerSecond(void) {
681 long long t = mstime() - server.ops_sec_last_sample_time;
682 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
683 long long ops_sec;
684
685 ops_sec = t > 0 ? (ops*1000/t) : 0;
686
687 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
688 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
689 server.ops_sec_last_sample_time = mstime();
690 server.ops_sec_last_sample_ops = server.stat_numcommands;
691}
692
693/* Return the mean of all the samples. */
694long long getOperationsPerSecond(void) {
695 int j;
696 long long sum = 0;
697
698 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
699 sum += server.ops_sec_samples[j];
700 return sum / REDIS_OPS_SEC_SAMPLES;
701}
702
703/* Check for timeouts. Returns non-zero if the client was terminated */
704int clientsCronHandleTimeout(redisClient *c) {
705 time_t now = server.unixtime;
706
707 if (server.maxidletime &&
708 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
709 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
710 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
711 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
712 listLength(c->pubsub_patterns) == 0 &&
713 (now - c->lastinteraction > server.maxidletime))
714 {
715 redisLog(REDIS_VERBOSE,"Closing idle client");
716 freeClient(c);
717 return 1;
718 } else if (c->flags & REDIS_BLOCKED) {
719 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
720 addReply(c,shared.nullmultibulk);
721 unblockClientWaitingData(c);
722 }
723 }
724 return 0;
725}
726
727/* The client query buffer is an sds.c string that can end with a lot of
728 * free space not used, this function reclaims space if needed.
729 *
730 * The funciton always returns 0 as it never terminates the client. */
731int clientsCronResizeQueryBuffer(redisClient *c) {
732 size_t querybuf_size = sdsAllocSize(c->querybuf);
733 time_t idletime = server.unixtime - c->lastinteraction;
734
735 /* There are two conditions to resize the query buffer:
736 * 1) Query buffer is > BIG_ARG and too big for latest peak.
737 * 2) Client is inactive and the buffer is bigger than 1k. */
738 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
739 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
740 (querybuf_size > 1024 && idletime > 2))
741 {
742 /* Only resize the query buffer if it is actually wasting space. */
743 if (sdsavail(c->querybuf) > 1024) {
744 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
745 }
746 }
747 /* Reset the peak again to capture the peak memory usage in the next
748 * cycle. */
749 c->querybuf_peak = 0;
750 return 0;
751}
752
753void clientsCron(void) {
754 /* Make sure to process at least 1/(REDIS_HZ*10) of clients per call.
755 * Since this function is called REDIS_HZ times per second we are sure that
756 * in the worst case we process all the clients in 10 seconds.
757 * In normal conditions (a reasonable number of clients) we process
758 * all the clients in a shorter time. */
759 int numclients = listLength(server.clients);
760 int iterations = numclients/(REDIS_HZ*10);
761
762 if (iterations < 50)
763 iterations = (numclients < 50) ? numclients : 50;
764 while(listLength(server.clients) && iterations--) {
765 redisClient *c;
766 listNode *head;
767
768 /* Rotate the list, take the current head, process.
769 * This way if the client must be removed from the list it's the
770 * first element and we don't incur into O(N) computation. */
771 listRotate(server.clients);
772 head = listFirst(server.clients);
773 c = listNodeValue(head);
774 /* The following functions do different service checks on the client.
775 * The protocol is that they return non-zero if the client was
776 * terminated. */
777 if (clientsCronHandleTimeout(c)) continue;
778 if (clientsCronResizeQueryBuffer(c)) continue;
779 }
780}
781
782/* This is our timer interrupt, called REDIS_HZ times per second.
783 * Here is where we do a number of things that need to be done asynchronously.
784 * For instance:
785 *
786 * - Active expired keys collection (it is also performed in a lazy way on
787 * lookup).
788 * - Software watchdong.
789 * - Update some statistic.
790 * - Incremental rehashing of the DBs hash tables.
791 * - Triggering BGSAVE / AOF rewrite, and handling of terminated children.
792 * - Clients timeout of differnet kinds.
793 * - Replication reconnection.
794 * - Many more...
795 *
796 * Everything directly called here will be called REDIS_HZ times per second,
797 * so in order to throttle execution of things we want to do less frequently
798 * a macro is used: run_with_period(milliseconds) { .... }
799 */
800
801int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
802 int j;
803 REDIS_NOTUSED(eventLoop);
804 REDIS_NOTUSED(id);
805 REDIS_NOTUSED(clientData);
806
807 /* Software watchdog: deliver the SIGALRM that will reach the signal
808 * handler if we don't return here fast enough. */
809 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
810
811 /* We take a cached value of the unix time in the global state because
812 * with virtual memory and aging there is to store the current time
813 * in objects at every object access, and accuracy is not needed.
814 * To access a global var is faster than calling time(NULL) */
815 server.unixtime = time(NULL);
816
817 run_with_period(100) trackOperationsPerSecond();
818
819 /* We have just 22 bits per object for LRU information.
820 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
821 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
822 *
823 * Note that even if this will wrap after 1.5 years it's not a problem,
824 * everything will still work but just some object will appear younger
825 * to Redis. But for this to happen a given object should never be touched
826 * for 1.5 years.
827 *
828 * Note that you can change the resolution altering the
829 * REDIS_LRU_CLOCK_RESOLUTION define.
830 */
831 updateLRUClock();
832
833 /* Record the max memory used since the server was started. */
834 if (zmalloc_used_memory() > server.stat_peak_memory)
835 server.stat_peak_memory = zmalloc_used_memory();
836
837 /* We received a SIGTERM, shutting down here in a safe way, as it is
838 * not ok doing so inside the signal handler. */
839 if (server.shutdown_asap) {
840 if (prepareForShutdown(0) == REDIS_OK) exit(0);
841 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
842 }
843
844 /* Show some info about non-empty databases */
845 run_with_period(5000) {
846 for (j = 0; j < server.dbnum; j++) {
847 long long size, used, vkeys;
848
849 size = dictSlots(server.db[j].dict);
850 used = dictSize(server.db[j].dict);
851 vkeys = dictSize(server.db[j].expires);
852 if (used || vkeys) {
853 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
854 /* dictPrintStats(server.dict); */
855 }
856 }
857 }
858
859 /* We don't want to resize the hash tables while a bacground saving
860 * is in progress: the saving child is created using fork() that is
861 * implemented with a copy-on-write semantic in most modern systems, so
862 * if we resize the HT while there is the saving child at work actually
863 * a lot of memory movements in the parent will cause a lot of pages
864 * copied. */
865 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
866 tryResizeHashTables();
867 if (server.activerehashing) incrementallyRehash();
868 }
869
870 /* Show information about connected clients */
871 if (!server.sentinel_mode) {
872 run_with_period(5000) {
873 redisLog(REDIS_VERBOSE,
874 "%d clients connected (%d slaves), %zu bytes in use",
875 listLength(server.clients)-listLength(server.slaves),
876 listLength(server.slaves),
877 zmalloc_used_memory());
878 }
879 }
880
881 /* We need to do a few operations on clients asynchronously. */
882 clientsCron();
883
884 /* Start a scheduled AOF rewrite if this was requested by the user while
885 * a BGSAVE was in progress. */
886 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
887 server.aof_rewrite_scheduled)
888 {
889 rewriteAppendOnlyFileBackground();
890 }
891
892 /* Check if a background saving or AOF rewrite in progress terminated. */
893 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
894 int statloc;
895 pid_t pid;
896
897 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
898 int exitcode = WEXITSTATUS(statloc);
899 int bysignal = 0;
900
901 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
902
903 if (pid == server.rdb_child_pid) {
904 backgroundSaveDoneHandler(exitcode,bysignal);
905 } else if (pid == server.aof_child_pid) {
906 backgroundRewriteDoneHandler(exitcode,bysignal);
907 } else {
908 redisLog(REDIS_WARNING,
909 "Warning, detected child with unmatched pid: %ld",
910 (long)pid);
911 }
912 updateDictResizePolicy();
913 }
914 } else {
915 /* If there is not a background saving/rewrite in progress check if
916 * we have to save/rewrite now */
917 for (j = 0; j < server.saveparamslen; j++) {
918 struct saveparam *sp = server.saveparams+j;
919
920 if (server.dirty >= sp->changes &&
921 server.unixtime-server.lastsave > sp->seconds) {
922 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
923 sp->changes, sp->seconds);
924 rdbSaveBackground(server.rdb_filename);
925 break;
926 }
927 }
928
929 /* Trigger an AOF rewrite if needed */
930 if (server.rdb_child_pid == -1 &&
931 server.aof_child_pid == -1 &&
932 server.aof_rewrite_perc &&
933 server.aof_current_size > server.aof_rewrite_min_size)
934 {
935 long long base = server.aof_rewrite_base_size ?
936 server.aof_rewrite_base_size : 1;
937 long long growth = (server.aof_current_size*100/base) - 100;
938 if (growth >= server.aof_rewrite_perc) {
939 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
940 rewriteAppendOnlyFileBackground();
941 }
942 }
943 }
944
945
946 /* If we postponed an AOF buffer flush, let's try to do it every time the
947 * cron function is called. */
948 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
949
950 /* Expire a few keys per cycle, only if this is a master.
951 * On slaves we wait for DEL operations synthesized by the master
952 * in order to guarantee a strict consistency. */
953 if (server.masterhost == NULL) activeExpireCycle();
954
955 /* Close clients that need to be closed asynchronous */
956 freeClientsInAsyncFreeQueue();
957
958 /* Replication cron function -- used to reconnect to master and
959 * to detect transfer failures. */
960 run_with_period(1000) replicationCron();
961
962 /* Run the sentinel timer if we are in sentinel mode. */
963 run_with_period(100) {
964 if (server.sentinel_mode) sentinelTimer();
965 }
966
967 server.cronloops++;
968 return 1000/REDIS_HZ;
969}
970
971/* This function gets called every time Redis is entering the
972 * main loop of the event driven library, that is, before to sleep
973 * for ready file descriptors. */
974void beforeSleep(struct aeEventLoop *eventLoop) {
975 REDIS_NOTUSED(eventLoop);
976 listNode *ln;
977 redisClient *c;
978
979 /* Try to process pending commands for clients that were just unblocked. */
980 while (listLength(server.unblocked_clients)) {
981 ln = listFirst(server.unblocked_clients);
982 redisAssert(ln != NULL);
983 c = ln->value;
984 listDelNode(server.unblocked_clients,ln);
985 c->flags &= ~REDIS_UNBLOCKED;
986
987 /* Process remaining data in the input buffer. */
988 if (c->querybuf && sdslen(c->querybuf) > 0) {
989 server.current_client = c;
990 processInputBuffer(c);
991 server.current_client = NULL;
992 }
993 }
994
995 /* Write the AOF buffer on disk */
996 flushAppendOnlyFile(0);
997}
998
999/* =========================== Server initialization ======================== */
1000
1001void createSharedObjects(void) {
1002 int j;
1003
1004 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
1005 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
1006 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
1007 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
1008 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
1009 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
1010 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
1011 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
1012 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
1013 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
1014 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
1015 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
1016 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
1017 "-ERR Operation against a key holding the wrong kind of value\r\n"));
1018 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
1019 "-ERR no such key\r\n"));
1020 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
1021 "-ERR syntax error\r\n"));
1022 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
1023 "-ERR source and destination objects are the same\r\n"));
1024 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
1025 "-ERR index out of range\r\n"));
1026 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
1027 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
1028 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
1029 "-LOADING Redis is loading the dataset in memory\r\n"));
1030 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
1031 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
1032 shared.masterdownerr = createObject(REDIS_STRING,sdsnew(
1033 "-MASTERDOWN Link with MASTER is down and slave-serve-stale-data is set to 'no'.\r\n"));
1034 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
1035 "-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"));
1036 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
1037 "-READONLY You can't write against a read only slave.\r\n"));
1038 shared.oomerr = createObject(REDIS_STRING,sdsnew(
1039 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
1040 shared.space = createObject(REDIS_STRING,sdsnew(" "));
1041 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1042 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1043
1044 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1045 shared.select[j] = createObject(REDIS_STRING,
1046 sdscatprintf(sdsempty(),"select %d\r\n", j));
1047 }
1048 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1049 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1050 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1051 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1052 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1053 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1054 shared.del = createStringObject("DEL",3);
1055 shared.rpop = createStringObject("RPOP",4);
1056 shared.lpop = createStringObject("LPOP",4);
1057 shared.lpush = createStringObject("LPUSH",5);
1058 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1059 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1060 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1061 }
1062 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1063 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1064 sdscatprintf(sdsempty(),"*%d\r\n",j));
1065 shared.bulkhdr[j] = createObject(REDIS_STRING,
1066 sdscatprintf(sdsempty(),"$%d\r\n",j));
1067 }
1068}
1069
1070void initServerConfig() {
1071 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1072 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1073 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1074 server.port = REDIS_SERVERPORT;
1075 server.bindaddr = NULL;
1076 server.unixsocket = NULL;
1077 server.unixsocketperm = 0;
1078 server.ipfd = -1;
1079 server.sofd = -1;
1080 server.dbnum = REDIS_DEFAULT_DBNUM;
1081 server.verbosity = REDIS_NOTICE;
1082 server.maxidletime = REDIS_MAXIDLETIME;
1083 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1084 server.saveparams = NULL;
1085 server.loading = 0;
1086 server.logfile = NULL; /* NULL = log on standard output */
1087 server.syslog_enabled = 0;
1088 server.syslog_ident = zstrdup("redis");
1089 server.syslog_facility = LOG_LOCAL0;
1090 server.daemonize = 0;
1091 server.aof_state = REDIS_AOF_OFF;
1092 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1093 server.aof_no_fsync_on_rewrite = 0;
1094 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1095 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1096 server.aof_rewrite_base_size = 0;
1097 server.aof_rewrite_scheduled = 0;
1098 server.aof_last_fsync = time(NULL);
1099 server.aof_rewrite_time_last = -1;
1100 server.aof_rewrite_time_start = -1;
1101 server.aof_lastbgrewrite_status = REDIS_OK;
1102 server.aof_delayed_fsync = 0;
1103 server.aof_fd = -1;
1104 server.aof_selected_db = -1; /* Make sure the first time will not match */
1105 server.aof_flush_postponed_start = 0;
1106 server.pidfile = zstrdup("/var/run/redis.pid");
1107 server.rdb_filename = zstrdup("dump.rdb");
1108 server.aof_filename = zstrdup("appendonly.aof");
1109 server.requirepass = NULL;
1110 server.rdb_compression = 1;
1111 server.rdb_checksum = 1;
1112 server.activerehashing = 1;
1113 server.maxclients = REDIS_MAX_CLIENTS;
1114 server.bpop_blocked_clients = 0;
1115 server.maxmemory = 0;
1116 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1117 server.maxmemory_samples = 3;
1118 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1119 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1120 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1121 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1122 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1123 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1124 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1125 server.shutdown_asap = 0;
1126 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1127 server.repl_timeout = REDIS_REPL_TIMEOUT;
1128 server.lua_caller = NULL;
1129 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1130 server.lua_client = NULL;
1131 server.lua_timedout = 0;
1132
1133 updateLRUClock();
1134 resetServerSaveParams();
1135
1136 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1137 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1138 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1139 /* Replication related */
1140 server.masterauth = NULL;
1141 server.masterhost = NULL;
1142 server.masterport = 6379;
1143 server.master = NULL;
1144 server.repl_state = REDIS_REPL_NONE;
1145 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1146 server.repl_serve_stale_data = 1;
1147 server.repl_slave_ro = 1;
1148 server.repl_down_since = time(NULL);
1149 server.slave_priority = REDIS_DEFAULT_SLAVE_PRIORITY;
1150
1151 /* Client output buffer limits */
1152 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1153 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1154 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1155 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1156 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1157 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1158 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1159 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1160 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1161
1162 /* Double constants initialization */
1163 R_Zero = 0.0;
1164 R_PosInf = 1.0/R_Zero;
1165 R_NegInf = -1.0/R_Zero;
1166 R_Nan = R_Zero/R_Zero;
1167
1168 /* Command table -- we intiialize it here as it is part of the
1169 * initial configuration, since command names may be changed via
1170 * redis.conf using the rename-command directive. */
1171 server.commands = dictCreate(&commandTableDictType,NULL);
1172 populateCommandTable();
1173 server.delCommand = lookupCommandByCString("del");
1174 server.multiCommand = lookupCommandByCString("multi");
1175 server.lpushCommand = lookupCommandByCString("lpush");
1176 server.lpopCommand = lookupCommandByCString("lpop");
1177 server.rpopCommand = lookupCommandByCString("rpop");
1178
1179 /* Slow log */
1180 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1181 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1182
1183 /* Debugging */
1184 server.assert_failed = "<no assertion failed>";
1185 server.assert_file = "<no file>";
1186 server.assert_line = 0;
1187 server.bug_report_start = 0;
1188 server.watchdog_period = 0;
1189}
1190
1191/* This function will try to raise the max number of open files accordingly to
1192 * the configured max number of clients. It will also account for 32 additional
1193 * file descriptors as we need a few more for persistence, listening
1194 * sockets, log files and so forth.
1195 *
1196 * If it will not be possible to set the limit accordingly to the configured
1197 * max number of clients, the function will do the reverse setting
1198 * server.maxclients to the value that we can actually handle. */
1199void adjustOpenFilesLimit(void) {
1200 rlim_t maxfiles = server.maxclients+32;
1201 struct rlimit limit;
1202
1203 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1204 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1205 strerror(errno));
1206 server.maxclients = 1024-32;
1207 } else {
1208 rlim_t oldlimit = limit.rlim_cur;
1209
1210 /* Set the max number of files if the current limit is not enough
1211 * for our needs. */
1212 if (oldlimit < maxfiles) {
1213 rlim_t f;
1214
1215 f = maxfiles;
1216 while(f > oldlimit) {
1217 limit.rlim_cur = f;
1218 limit.rlim_max = f;
1219 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1220 f -= 128;
1221 }
1222 if (f < oldlimit) f = oldlimit;
1223 if (f != maxfiles) {
1224 server.maxclients = f-32;
1225 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1226 (int) maxfiles, strerror(errno), (int) server.maxclients);
1227 } else {
1228 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1229 (int) maxfiles);
1230 }
1231 }
1232 }
1233}
1234
1235void initServer() {
1236 int j;
1237
1238 signal(SIGHUP, SIG_IGN);
1239 signal(SIGPIPE, SIG_IGN);
1240 setupSignalHandlers();
1241
1242 if (server.syslog_enabled) {
1243 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1244 server.syslog_facility);
1245 }
1246
1247 server.current_client = NULL;
1248 server.clients = listCreate();
1249 server.clients_to_close = listCreate();
1250 server.slaves = listCreate();
1251 server.monitors = listCreate();
1252 server.unblocked_clients = listCreate();
1253 server.ready_keys = listCreate();
1254
1255 createSharedObjects();
1256 adjustOpenFilesLimit();
1257 server.el = aeCreateEventLoop(server.maxclients+1024);
1258 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1259
1260 if (server.port != 0) {
1261 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1262 if (server.ipfd == ANET_ERR) {
1263 redisLog(REDIS_WARNING, "Opening port %d: %s",
1264 server.port, server.neterr);
1265 exit(1);
1266 }
1267 }
1268 if (server.unixsocket != NULL) {
1269 unlink(server.unixsocket); /* don't care if this fails */
1270 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1271 if (server.sofd == ANET_ERR) {
1272 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1273 exit(1);
1274 }
1275 }
1276 if (server.ipfd < 0 && server.sofd < 0) {
1277 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1278 exit(1);
1279 }
1280 for (j = 0; j < server.dbnum; j++) {
1281 server.db[j].dict = dictCreate(&dbDictType,NULL);
1282 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1283 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1284 server.db[j].ready_keys = dictCreate(&setDictType,NULL);
1285 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1286 server.db[j].id = j;
1287 }
1288 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1289 server.pubsub_patterns = listCreate();
1290 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1291 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1292 server.cronloops = 0;
1293 server.rdb_child_pid = -1;
1294 server.aof_child_pid = -1;
1295 aofRewriteBufferReset();
1296 server.aof_buf = sdsempty();
1297 server.lastsave = time(NULL);
1298 server.rdb_save_time_last = -1;
1299 server.rdb_save_time_start = -1;
1300 server.dirty = 0;
1301 server.stat_numcommands = 0;
1302 server.stat_numconnections = 0;
1303 server.stat_expiredkeys = 0;
1304 server.stat_evictedkeys = 0;
1305 server.stat_starttime = time(NULL);
1306 server.stat_keyspace_misses = 0;
1307 server.stat_keyspace_hits = 0;
1308 server.stat_peak_memory = 0;
1309 server.stat_fork_time = 0;
1310 server.stat_rejected_conn = 0;
1311 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1312 server.ops_sec_idx = 0;
1313 server.ops_sec_last_sample_time = mstime();
1314 server.ops_sec_last_sample_ops = 0;
1315 server.unixtime = time(NULL);
1316 server.lastbgsave_status = REDIS_OK;
1317 server.stop_writes_on_bgsave_err = 1;
1318 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1319 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1320 acceptTcpHandler,NULL) == AE_ERR) redisPanic("Unrecoverable error creating server.ipfd file event.");
1321 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1322 acceptUnixHandler,NULL) == AE_ERR) redisPanic("Unrecoverable error creating server.sofd file event.");
1323
1324 if (server.aof_state == REDIS_AOF_ON) {
1325 server.aof_fd = open(server.aof_filename,
1326 O_WRONLY|O_APPEND|O_CREAT,0644);
1327 if (server.aof_fd == -1) {
1328 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1329 strerror(errno));
1330 exit(1);
1331 }
1332 }
1333
1334 /* 32 bit instances are limited to 4GB of address space, so if there is
1335 * no explicit limit in the user provided configuration we set a limit
1336 * at 3 GB using maxmemory with 'noeviction' policy'. This avoids
1337 * useless crashes of the Redis instance for out of memory. */
1338 if (server.arch_bits == 32 && server.maxmemory == 0) {
1339 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3 GB maxmemory limit with 'noeviction' policy now.");
1340 server.maxmemory = 3072LL*(1024*1024); /* 3 GB */
1341 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1342 }
1343
1344 scriptingInit();
1345 slowlogInit();
1346 bioInit();
1347}
1348
1349/* Populates the Redis Command Table starting from the hard coded list
1350 * we have on top of redis.c file. */
1351void populateCommandTable(void) {
1352 int j;
1353 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1354
1355 for (j = 0; j < numcommands; j++) {
1356 struct redisCommand *c = redisCommandTable+j;
1357 char *f = c->sflags;
1358 int retval;
1359
1360 while(*f != '\0') {
1361 switch(*f) {
1362 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1363 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1364 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1365 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1366 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1367 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1368 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1369 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1370 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1371 case 'l': c->flags |= REDIS_CMD_LOADING; break;
1372 case 't': c->flags |= REDIS_CMD_STALE; break;
1373 case 'M': c->flags |= REDIS_CMD_SKIP_MONITOR; break;
1374 default: redisPanic("Unsupported command flag"); break;
1375 }
1376 f++;
1377 }
1378
1379 retval = dictAdd(server.commands, sdsnew(c->name), c);
1380 assert(retval == DICT_OK);
1381 }
1382}
1383
1384void resetCommandTableStats(void) {
1385 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1386 int j;
1387
1388 for (j = 0; j < numcommands; j++) {
1389 struct redisCommand *c = redisCommandTable+j;
1390
1391 c->microseconds = 0;
1392 c->calls = 0;
1393 }
1394}
1395
1396/* ========================== Redis OP Array API ============================ */
1397
1398void redisOpArrayInit(redisOpArray *oa) {
1399 oa->ops = NULL;
1400 oa->numops = 0;
1401}
1402
1403int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1404 robj **argv, int argc, int target)
1405{
1406 redisOp *op;
1407
1408 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1409 op = oa->ops+oa->numops;
1410 op->cmd = cmd;
1411 op->dbid = dbid;
1412 op->argv = argv;
1413 op->argc = argc;
1414 op->target = target;
1415 oa->numops++;
1416 return oa->numops;
1417}
1418
1419void redisOpArrayFree(redisOpArray *oa) {
1420 while(oa->numops) {
1421 int j;
1422 redisOp *op;
1423
1424 oa->numops--;
1425 op = oa->ops+oa->numops;
1426 for (j = 0; j < op->argc; j++)
1427 decrRefCount(op->argv[j]);
1428 zfree(op->argv);
1429 }
1430 zfree(oa->ops);
1431}
1432
1433/* ====================== Commands lookup and execution ===================== */
1434
1435struct redisCommand *lookupCommand(sds name) {
1436 return dictFetchValue(server.commands, name);
1437}
1438
1439struct redisCommand *lookupCommandByCString(char *s) {
1440 struct redisCommand *cmd;
1441 sds name = sdsnew(s);
1442
1443 cmd = dictFetchValue(server.commands, name);
1444 sdsfree(name);
1445 return cmd;
1446}
1447
1448/* Propagate the specified command (in the context of the specified database id)
1449 * to AOF and Slaves.
1450 *
1451 * flags are an xor between:
1452 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1453 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1454 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1455 */
1456void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1457 int flags)
1458{
1459 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1460 feedAppendOnlyFile(cmd,dbid,argv,argc);
1461 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1462 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1463}
1464
1465/* Used inside commands to schedule the propagation of additional commands
1466 * after the current command is propagated to AOF / Replication. */
1467void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1468 int target)
1469{
1470 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1471}
1472
1473/* Call() is the core of Redis execution of a command */
1474void call(redisClient *c, int flags) {
1475 long long dirty, start = ustime(), duration;
1476
1477 /* Sent the command to clients in MONITOR mode, only if the commands are
1478 * not geneated from reading an AOF. */
1479 if (listLength(server.monitors) &&
1480 !server.loading &&
1481 !(c->cmd->flags & REDIS_CMD_SKIP_MONITOR))
1482 {
1483 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1484 }
1485
1486 /* Call the command. */
1487 redisOpArrayInit(&server.also_propagate);
1488 dirty = server.dirty;
1489 c->cmd->proc(c);
1490 dirty = server.dirty-dirty;
1491 duration = ustime()-start;
1492
1493 /* When EVAL is called loading the AOF we don't want commands called
1494 * from Lua to go into the slowlog or to populate statistics. */
1495 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1496 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1497
1498 /* Log the command into the Slow log if needed, and populate the
1499 * per-command statistics that we show in INFO commandstats. */
1500 if (flags & REDIS_CALL_SLOWLOG)
1501 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1502 if (flags & REDIS_CALL_STATS) {
1503 c->cmd->microseconds += duration;
1504 c->cmd->calls++;
1505 }
1506
1507 /* Propagate the command into the AOF and replication link */
1508 if (flags & REDIS_CALL_PROPAGATE) {
1509 int flags = REDIS_PROPAGATE_NONE;
1510
1511 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1512 flags |= REDIS_PROPAGATE_REPL;
1513 if (dirty)
1514 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1515 if (flags != REDIS_PROPAGATE_NONE)
1516 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1517 }
1518 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1519 * PUSH command. */
1520 if (server.also_propagate.numops) {
1521 int j;
1522 redisOp *rop;
1523
1524 for (j = 0; j < server.also_propagate.numops; j++) {
1525 rop = &server.also_propagate.ops[j];
1526 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1527 }
1528 redisOpArrayFree(&server.also_propagate);
1529 }
1530 server.stat_numcommands++;
1531}
1532
1533/* If this function gets called we already read a whole
1534 * command, arguments are in the client argv/argc fields.
1535 * processCommand() execute the command or prepare the
1536 * server for a bulk read from the client.
1537 *
1538 * If 1 is returned the client is still alive and valid and
1539 * and other operations can be performed by the caller. Otherwise
1540 * if 0 is returned the client was destroied (i.e. after QUIT). */
1541int processCommand(redisClient *c) {
1542 /* The QUIT command is handled separately. Normal command procs will
1543 * go through checking for replication and QUIT will cause trouble
1544 * when FORCE_REPLICATION is enabled and would be implemented in
1545 * a regular command proc. */
1546 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1547 addReply(c,shared.ok);
1548 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1549 return REDIS_ERR;
1550 }
1551
1552 /* Now lookup the command and check ASAP about trivial error conditions
1553 * such as wrong arity, bad command name and so forth. */
1554 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1555 if (!c->cmd) {
1556 addReplyErrorFormat(c,"unknown command '%s'",
1557 (char*)c->argv[0]->ptr);
1558 return REDIS_OK;
1559 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1560 (c->argc < -c->cmd->arity)) {
1561 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1562 c->cmd->name);
1563 return REDIS_OK;
1564 }
1565
1566 /* Check if the user is authenticated */
1567 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1568 {
1569 addReplyError(c,"operation not permitted");
1570 return REDIS_OK;
1571 }
1572
1573 /* Handle the maxmemory directive.
1574 *
1575 * First we try to free some memory if possible (if there are volatile
1576 * keys in the dataset). If there are not the only thing we can do
1577 * is returning an error. */
1578 if (server.maxmemory) {
1579 int retval = freeMemoryIfNeeded();
1580 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1581 addReply(c, shared.oomerr);
1582 return REDIS_OK;
1583 }
1584 }
1585
1586 /* Don't accept write commands if there are problems persisting on disk. */
1587 if (server.stop_writes_on_bgsave_err &&
1588 server.saveparamslen > 0
1589 && server.lastbgsave_status == REDIS_ERR &&
1590 c->cmd->flags & REDIS_CMD_WRITE)
1591 {
1592 addReply(c, shared.bgsaveerr);
1593 return REDIS_OK;
1594 }
1595
1596 /* Don't accept write commands if this is a read only slave. But
1597 * accept write commands if this is our master. */
1598 if (server.masterhost && server.repl_slave_ro &&
1599 !(c->flags & REDIS_MASTER) &&
1600 c->cmd->flags & REDIS_CMD_WRITE)
1601 {
1602 addReply(c, shared.roslaveerr);
1603 return REDIS_OK;
1604 }
1605
1606 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1607 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1608 &&
1609 c->cmd->proc != subscribeCommand &&
1610 c->cmd->proc != unsubscribeCommand &&
1611 c->cmd->proc != psubscribeCommand &&
1612 c->cmd->proc != punsubscribeCommand) {
1613 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1614 return REDIS_OK;
1615 }
1616
1617 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1618 * we are a slave with a broken link with master. */
1619 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1620 server.repl_serve_stale_data == 0 &&
1621 !(c->cmd->flags & REDIS_CMD_STALE))
1622 {
1623 addReply(c, shared.masterdownerr);
1624 return REDIS_OK;
1625 }
1626
1627 /* Loading DB? Return an error if the command has not the
1628 * REDIS_CMD_LOADING flag. */
1629 if (server.loading && !(c->cmd->flags & REDIS_CMD_LOADING)) {
1630 addReply(c, shared.loadingerr);
1631 return REDIS_OK;
1632 }
1633
1634 /* Lua script too slow? Only allow commands with REDIS_CMD_STALE flag. */
1635 if (server.lua_timedout &&
1636 c->cmd->proc != authCommand &&
1637 !(c->cmd->proc == shutdownCommand &&
1638 c->argc == 2 &&
1639 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1640 !(c->cmd->proc == scriptCommand &&
1641 c->argc == 2 &&
1642 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1643 {
1644 addReply(c, shared.slowscripterr);
1645 return REDIS_OK;
1646 }
1647
1648 /* Exec the command */
1649 if (c->flags & REDIS_MULTI &&
1650 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1651 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1652 {
1653 queueMultiCommand(c);
1654 addReply(c,shared.queued);
1655 } else {
1656 call(c,REDIS_CALL_FULL);
1657 if (listLength(server.ready_keys))
1658 handleClientsBlockedOnLists();
1659 }
1660 return REDIS_OK;
1661}
1662
1663/*================================== Shutdown =============================== */
1664
1665int prepareForShutdown(int flags) {
1666 int save = flags & REDIS_SHUTDOWN_SAVE;
1667 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1668
1669 redisLog(REDIS_WARNING,"User requested shutdown...");
1670 /* Kill the saving child if there is a background saving in progress.
1671 We want to avoid race conditions, for instance our saving child may
1672 overwrite the synchronous saving did by SHUTDOWN. */
1673 if (server.rdb_child_pid != -1) {
1674 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1675 kill(server.rdb_child_pid,SIGKILL);
1676 rdbRemoveTempFile(server.rdb_child_pid);
1677 }
1678 if (server.aof_state != REDIS_AOF_OFF) {
1679 /* Kill the AOF saving child as the AOF we already have may be longer
1680 * but contains the full dataset anyway. */
1681 if (server.aof_child_pid != -1) {
1682 redisLog(REDIS_WARNING,
1683 "There is a child rewriting the AOF. Killing it!");
1684 kill(server.aof_child_pid,SIGKILL);
1685 }
1686 /* Append only file: fsync() the AOF and exit */
1687 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1688 aof_fsync(server.aof_fd);
1689 }
1690 if ((server.saveparamslen > 0 && !nosave) || save) {
1691 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1692 /* Snapshotting. Perform a SYNC SAVE and exit */
1693 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1694 /* Ooops.. error saving! The best we can do is to continue
1695 * operating. Note that if there was a background saving process,
1696 * in the next cron() Redis will be notified that the background
1697 * saving aborted, handling special stuff like slaves pending for
1698 * synchronization... */
1699 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1700 return REDIS_ERR;
1701 }
1702 }
1703 if (server.daemonize) {
1704 redisLog(REDIS_NOTICE,"Removing the pid file.");
1705 unlink(server.pidfile);
1706 }
1707 /* Close the listening sockets. Apparently this allows faster restarts. */
1708 if (server.ipfd != -1) close(server.ipfd);
1709 if (server.sofd != -1) close(server.sofd);
1710 if (server.unixsocket) {
1711 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1712 unlink(server.unixsocket); /* don't care if this fails */
1713 }
1714
1715 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1716 return REDIS_OK;
1717}
1718
1719/*================================== Commands =============================== */
1720
1721/* Return zero if strings are the same, non-zero if they are not.
1722 * The comparison is performed in a way that prevents an attacker to obtain
1723 * information about the nature of the strings just monitoring the execution
1724 * time of the function.
1725 *
1726 * Note that limiting the comparison length to strings up to 512 bytes we
1727 * can avoid leaking any information about the password length and any
1728 * possible branch misprediction related leak.
1729 */
1730int time_independent_strcmp(char *a, char *b) {
1731 char bufa[REDIS_AUTHPASS_MAX_LEN], bufb[REDIS_AUTHPASS_MAX_LEN];
1732 /* The above two strlen perform len(a) + len(b) operations where either
1733 * a or b are fixed (our password) length, and the difference is only
1734 * relative to the length of the user provided string, so no information
1735 * leak is possible in the following two lines of code. */
1736 int alen = strlen(a);
1737 int blen = strlen(b);
1738 int j;
1739 int diff = 0;
1740
1741 /* We can't compare strings longer than our static buffers.
1742 * Note that this will never pass the first test in practical circumstances
1743 * so there is no info leak. */
1744 if (alen > sizeof(bufa) || blen > sizeof(bufb)) return 1;
1745
1746 memset(bufa,0,sizeof(bufa)); /* Constant time. */
1747 memset(bufb,0,sizeof(bufb)); /* Constant time. */
1748 /* Again the time of the following two copies is proportional to
1749 * len(a) + len(b) so no info is leaked. */
1750 memcpy(bufa,a,alen);
1751 memcpy(bufb,b,blen);
1752
1753 /* Always compare all the chars in the two buffers without
1754 * conditional expressions. */
1755 for (j = 0; j < sizeof(bufa); j++) {
1756 diff |= (bufa[j] ^ bufb[j]);
1757 }
1758 /* Length must be equal as well. */
1759 diff |= alen ^ blen;
1760 return diff; /* If zero strings are the same. */
1761}
1762
1763void authCommand(redisClient *c) {
1764 if (!server.requirepass) {
1765 addReplyError(c,"Client sent AUTH, but no password is set");
1766 } else if (!time_independent_strcmp(c->argv[1]->ptr, server.requirepass)) {
1767 c->authenticated = 1;
1768 addReply(c,shared.ok);
1769 } else {
1770 c->authenticated = 0;
1771 addReplyError(c,"invalid password");
1772 }
1773}
1774
1775void pingCommand(redisClient *c) {
1776 addReply(c,shared.pong);
1777}
1778
1779void echoCommand(redisClient *c) {
1780 addReplyBulk(c,c->argv[1]);
1781}
1782
1783void timeCommand(redisClient *c) {
1784 struct timeval tv;
1785
1786 /* gettimeofday() can only fail if &tv is a bad addresss so we
1787 * don't check for errors. */
1788 gettimeofday(&tv,NULL);
1789 addReplyMultiBulkLen(c,2);
1790 addReplyBulkLongLong(c,tv.tv_sec);
1791 addReplyBulkLongLong(c,tv.tv_usec);
1792}
1793
1794/* Convert an amount of bytes into a human readable string in the form
1795 * of 100B, 2G, 100M, 4K, and so forth. */
1796void bytesToHuman(char *s, unsigned long long n) {
1797 double d;
1798
1799 if (n < 1024) {
1800 /* Bytes */
1801 sprintf(s,"%lluB",n);
1802 return;
1803 } else if (n < (1024*1024)) {
1804 d = (double)n/(1024);
1805 sprintf(s,"%.2fK",d);
1806 } else if (n < (1024LL*1024*1024)) {
1807 d = (double)n/(1024*1024);
1808 sprintf(s,"%.2fM",d);
1809 } else if (n < (1024LL*1024*1024*1024)) {
1810 d = (double)n/(1024LL*1024*1024);
1811 sprintf(s,"%.2fG",d);
1812 }
1813}
1814
1815/* Create the string returned by the INFO command. This is decoupled
1816 * by the INFO command itself as we need to report the same information
1817 * on memory corruption problems. */
1818sds genRedisInfoString(char *section) {
1819 sds info = sdsempty();
1820 time_t uptime = server.unixtime-server.stat_starttime;
1821 int j, numcommands;
1822 struct rusage self_ru, c_ru;
1823 unsigned long lol, bib;
1824 int allsections = 0, defsections = 0;
1825 int sections = 0;
1826
1827 if (section) {
1828 allsections = strcasecmp(section,"all") == 0;
1829 defsections = strcasecmp(section,"default") == 0;
1830 }
1831
1832 getrusage(RUSAGE_SELF, &self_ru);
1833 getrusage(RUSAGE_CHILDREN, &c_ru);
1834 getClientsMaxBuffers(&lol,&bib);
1835
1836 /* Server */
1837 if (allsections || defsections || !strcasecmp(section,"server")) {
1838 struct utsname name;
1839 char *mode;
1840
1841 if (server.sentinel_mode) mode = "sentinel";
1842 else mode = "standalone";
1843
1844 if (sections++) info = sdscat(info,"\r\n");
1845 uname(&name);
1846 info = sdscatprintf(info,
1847 "# Server\r\n"
1848 "redis_version:%s\r\n"
1849 "redis_git_sha1:%s\r\n"
1850 "redis_git_dirty:%d\r\n"
1851 "redis_mode:%s\r\n"
1852 "os:%s %s %s\r\n"
1853 "arch_bits:%d\r\n"
1854 "multiplexing_api:%s\r\n"
1855 "gcc_version:%d.%d.%d\r\n"
1856 "process_id:%ld\r\n"
1857 "run_id:%s\r\n"
1858 "tcp_port:%d\r\n"
1859 "uptime_in_seconds:%ld\r\n"
1860 "uptime_in_days:%ld\r\n"
1861 "lru_clock:%ld\r\n",
1862 REDIS_VERSION,
1863 redisGitSHA1(),
1864 strtol(redisGitDirty(),NULL,10) > 0,
1865 mode,
1866 name.sysname, name.release, name.machine,
1867 server.arch_bits,
1868 aeGetApiName(),
1869#ifdef __GNUC__
1870 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1871#else
1872 0,0,0,
1873#endif
1874 (long) getpid(),
1875 server.runid,
1876 server.port,
1877 uptime,
1878 uptime/(3600*24),
1879 (unsigned long) server.lruclock);
1880 }
1881
1882 /* Clients */
1883 if (allsections || defsections || !strcasecmp(section,"clients")) {
1884 if (sections++) info = sdscat(info,"\r\n");
1885 info = sdscatprintf(info,
1886 "# Clients\r\n"
1887 "connected_clients:%lu\r\n"
1888 "client_longest_output_list:%lu\r\n"
1889 "client_biggest_input_buf:%lu\r\n"
1890 "blocked_clients:%d\r\n",
1891 listLength(server.clients)-listLength(server.slaves),
1892 lol, bib,
1893 server.bpop_blocked_clients);
1894 }
1895
1896 /* Memory */
1897 if (allsections || defsections || !strcasecmp(section,"memory")) {
1898 char hmem[64];
1899 char peak_hmem[64];
1900
1901 bytesToHuman(hmem,zmalloc_used_memory());
1902 bytesToHuman(peak_hmem,server.stat_peak_memory);
1903 if (sections++) info = sdscat(info,"\r\n");
1904 info = sdscatprintf(info,
1905 "# Memory\r\n"
1906 "used_memory:%zu\r\n"
1907 "used_memory_human:%s\r\n"
1908 "used_memory_rss:%zu\r\n"
1909 "used_memory_peak:%zu\r\n"
1910 "used_memory_peak_human:%s\r\n"
1911 "used_memory_lua:%lld\r\n"
1912 "mem_fragmentation_ratio:%.2f\r\n"
1913 "mem_allocator:%s\r\n",
1914 zmalloc_used_memory(),
1915 hmem,
1916 zmalloc_get_rss(),
1917 server.stat_peak_memory,
1918 peak_hmem,
1919 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1920 zmalloc_get_fragmentation_ratio(),
1921 ZMALLOC_LIB
1922 );
1923 }
1924
1925 /* Persistence */
1926 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1927 if (sections++) info = sdscat(info,"\r\n");
1928 info = sdscatprintf(info,
1929 "# Persistence\r\n"
1930 "loading:%d\r\n"
1931 "rdb_changes_since_last_save:%lld\r\n"
1932 "rdb_bgsave_in_progress:%d\r\n"
1933 "rdb_last_save_time:%ld\r\n"
1934 "rdb_last_bgsave_status:%s\r\n"
1935 "rdb_last_bgsave_time_sec:%ld\r\n"
1936 "rdb_current_bgsave_time_sec:%ld\r\n"
1937 "aof_enabled:%d\r\n"
1938 "aof_rewrite_in_progress:%d\r\n"
1939 "aof_rewrite_scheduled:%d\r\n"
1940 "aof_last_rewrite_time_sec:%ld\r\n"
1941 "aof_current_rewrite_time_sec:%ld\r\n"
1942 "aof_last_bgrewrite_status:%s\r\n",
1943 server.loading,
1944 server.dirty,
1945 server.rdb_child_pid != -1,
1946 server.lastsave,
1947 (server.lastbgsave_status == REDIS_OK) ? "ok" : "err",
1948 server.rdb_save_time_last,
1949 (server.rdb_child_pid == -1) ?
1950 -1 : time(NULL)-server.rdb_save_time_start,
1951 server.aof_state != REDIS_AOF_OFF,
1952 server.aof_child_pid != -1,
1953 server.aof_rewrite_scheduled,
1954 server.aof_rewrite_time_last,
1955 (server.aof_child_pid == -1) ?
1956 -1 : time(NULL)-server.aof_rewrite_time_start,
1957 (server.aof_lastbgrewrite_status == REDIS_OK) ? "ok" : "err");
1958
1959 if (server.aof_state != REDIS_AOF_OFF) {
1960 info = sdscatprintf(info,
1961 "aof_current_size:%lld\r\n"
1962 "aof_base_size:%lld\r\n"
1963 "aof_pending_rewrite:%d\r\n"
1964 "aof_buffer_length:%zu\r\n"
1965 "aof_rewrite_buffer_length:%lu\r\n"
1966 "aof_pending_bio_fsync:%llu\r\n"
1967 "aof_delayed_fsync:%lu\r\n",
1968 (long long) server.aof_current_size,
1969 (long long) server.aof_rewrite_base_size,
1970 server.aof_rewrite_scheduled,
1971 sdslen(server.aof_buf),
1972 aofRewriteBufferSize(),
1973 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1974 server.aof_delayed_fsync);
1975 }
1976
1977 if (server.loading) {
1978 double perc;
1979 time_t eta, elapsed;
1980 off_t remaining_bytes = server.loading_total_bytes-
1981 server.loading_loaded_bytes;
1982
1983 perc = ((double)server.loading_loaded_bytes /
1984 server.loading_total_bytes) * 100;
1985
1986 elapsed = server.unixtime-server.loading_start_time;
1987 if (elapsed == 0) {
1988 eta = 1; /* A fake 1 second figure if we don't have
1989 enough info */
1990 } else {
1991 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1992 }
1993
1994 info = sdscatprintf(info,
1995 "loading_start_time:%ld\r\n"
1996 "loading_total_bytes:%llu\r\n"
1997 "loading_loaded_bytes:%llu\r\n"
1998 "loading_loaded_perc:%.2f\r\n"
1999 "loading_eta_seconds:%ld\r\n"
2000 ,(unsigned long) server.loading_start_time,
2001 (unsigned long long) server.loading_total_bytes,
2002 (unsigned long long) server.loading_loaded_bytes,
2003 perc,
2004 eta
2005 );
2006 }
2007 }
2008
2009 /* Stats */
2010 if (allsections || defsections || !strcasecmp(section,"stats")) {
2011 if (sections++) info = sdscat(info,"\r\n");
2012 info = sdscatprintf(info,
2013 "# Stats\r\n"
2014 "total_connections_received:%lld\r\n"
2015 "total_commands_processed:%lld\r\n"
2016 "instantaneous_ops_per_sec:%lld\r\n"
2017 "rejected_connections:%lld\r\n"
2018 "expired_keys:%lld\r\n"
2019 "evicted_keys:%lld\r\n"
2020 "keyspace_hits:%lld\r\n"
2021 "keyspace_misses:%lld\r\n"
2022 "pubsub_channels:%ld\r\n"
2023 "pubsub_patterns:%lu\r\n"
2024 "latest_fork_usec:%lld\r\n",
2025 server.stat_numconnections,
2026 server.stat_numcommands,
2027 getOperationsPerSecond(),
2028 server.stat_rejected_conn,
2029 server.stat_expiredkeys,
2030 server.stat_evictedkeys,
2031 server.stat_keyspace_hits,
2032 server.stat_keyspace_misses,
2033 dictSize(server.pubsub_channels),
2034 listLength(server.pubsub_patterns),
2035 server.stat_fork_time);
2036 }
2037
2038 /* Replication */
2039 if (allsections || defsections || !strcasecmp(section,"replication")) {
2040 if (sections++) info = sdscat(info,"\r\n");
2041 info = sdscatprintf(info,
2042 "# Replication\r\n"
2043 "role:%s\r\n",
2044 server.masterhost == NULL ? "master" : "slave");
2045 if (server.masterhost) {
2046 info = sdscatprintf(info,
2047 "master_host:%s\r\n"
2048 "master_port:%d\r\n"
2049 "master_link_status:%s\r\n"
2050 "master_last_io_seconds_ago:%d\r\n"
2051 "master_sync_in_progress:%d\r\n"
2052 ,server.masterhost,
2053 server.masterport,
2054 (server.repl_state == REDIS_REPL_CONNECTED) ?
2055 "up" : "down",
2056 server.master ?
2057 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
2058 server.repl_state == REDIS_REPL_TRANSFER
2059 );
2060
2061 if (server.repl_state == REDIS_REPL_TRANSFER) {
2062 info = sdscatprintf(info,
2063 "master_sync_left_bytes:%lld\r\n"
2064 "master_sync_last_io_seconds_ago:%d\r\n"
2065 , (long long)
2066 (server.repl_transfer_size - server.repl_transfer_read),
2067 (int)(server.unixtime-server.repl_transfer_lastio)
2068 );
2069 }
2070
2071 if (server.repl_state != REDIS_REPL_CONNECTED) {
2072 info = sdscatprintf(info,
2073 "master_link_down_since_seconds:%ld\r\n",
2074 (long)server.unixtime-server.repl_down_since);
2075 }
2076 info = sdscatprintf(info,
2077 "slave_priority:%d\r\n"
2078 "slave_read_only:%d\r\n",
2079 server.slave_priority,
2080 server.repl_slave_ro);
2081 }
2082 info = sdscatprintf(info,
2083 "connected_slaves:%lu\r\n",
2084 listLength(server.slaves));
2085 if (listLength(server.slaves)) {
2086 int slaveid = 0;
2087 listNode *ln;
2088 listIter li;
2089
2090 listRewind(server.slaves,&li);
2091 while((ln = listNext(&li))) {
2092 redisClient *slave = listNodeValue(ln);
2093 char *state = NULL;
2094 char ip[32];
2095 int port;
2096
2097 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
2098 switch(slave->replstate) {
2099 case REDIS_REPL_WAIT_BGSAVE_START:
2100 case REDIS_REPL_WAIT_BGSAVE_END:
2101 state = "wait_bgsave";
2102 break;
2103 case REDIS_REPL_SEND_BULK:
2104 state = "send_bulk";
2105 break;
2106 case REDIS_REPL_ONLINE:
2107 state = "online";
2108 break;
2109 }
2110 if (state == NULL) continue;
2111 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2112 slaveid,ip,slave->slave_listening_port,state);
2113 slaveid++;
2114 }
2115 }
2116 }
2117
2118 /* CPU */
2119 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2120 if (sections++) info = sdscat(info,"\r\n");
2121 info = sdscatprintf(info,
2122 "# CPU\r\n"
2123 "used_cpu_sys:%.2f\r\n"
2124 "used_cpu_user:%.2f\r\n"
2125 "used_cpu_sys_children:%.2f\r\n"
2126 "used_cpu_user_children:%.2f\r\n",
2127 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2128 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2129 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2130 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2131 }
2132
2133 /* cmdtime */
2134 if (allsections || !strcasecmp(section,"commandstats")) {
2135 if (sections++) info = sdscat(info,"\r\n");
2136 info = sdscatprintf(info, "# Commandstats\r\n");
2137 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2138 for (j = 0; j < numcommands; j++) {
2139 struct redisCommand *c = redisCommandTable+j;
2140
2141 if (!c->calls) continue;
2142 info = sdscatprintf(info,
2143 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2144 c->name, c->calls, c->microseconds,
2145 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2146 }
2147 }
2148
2149 /* Key space */
2150 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2151 if (sections++) info = sdscat(info,"\r\n");
2152 info = sdscatprintf(info, "# Keyspace\r\n");
2153 for (j = 0; j < server.dbnum; j++) {
2154 long long keys, vkeys;
2155
2156 keys = dictSize(server.db[j].dict);
2157 vkeys = dictSize(server.db[j].expires);
2158 if (keys || vkeys) {
2159 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2160 j, keys, vkeys);
2161 }
2162 }
2163 }
2164 return info;
2165}
2166
2167void infoCommand(redisClient *c) {
2168 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2169
2170 if (c->argc > 2) {
2171 addReply(c,shared.syntaxerr);
2172 return;
2173 }
2174 sds info = genRedisInfoString(section);
2175 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2176 (unsigned long)sdslen(info)));
2177 addReplySds(c,info);
2178 addReply(c,shared.crlf);
2179}
2180
2181void monitorCommand(redisClient *c) {
2182 /* ignore MONITOR if already slave or in monitor mode */
2183 if (c->flags & REDIS_SLAVE) return;
2184
2185 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2186 c->slaveseldb = 0;
2187 listAddNodeTail(server.monitors,c);
2188 addReply(c,shared.ok);
2189}
2190
2191/* ============================ Maxmemory directive ======================== */
2192
2193/* This function gets called when 'maxmemory' is set on the config file to limit
2194 * the max memory used by the server, before processing a command.
2195 *
2196 * The goal of the function is to free enough memory to keep Redis under the
2197 * configured memory limit.
2198 *
2199 * The function starts calculating how many bytes should be freed to keep
2200 * Redis under the limit, and enters a loop selecting the best keys to
2201 * evict accordingly to the configured policy.
2202 *
2203 * If all the bytes needed to return back under the limit were freed the
2204 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2205 * should block the execution of commands that will result in more memory
2206 * used by the server.
2207 */
2208int freeMemoryIfNeeded(void) {
2209 size_t mem_used, mem_tofree, mem_freed;
2210 int slaves = listLength(server.slaves);
2211
2212 /* Remove the size of slaves output buffers and AOF buffer from the
2213 * count of used memory. */
2214 mem_used = zmalloc_used_memory();
2215 if (slaves) {
2216 listIter li;
2217 listNode *ln;
2218
2219 listRewind(server.slaves,&li);
2220 while((ln = listNext(&li))) {
2221 redisClient *slave = listNodeValue(ln);
2222 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2223 if (obuf_bytes > mem_used)
2224 mem_used = 0;
2225 else
2226 mem_used -= obuf_bytes;
2227 }
2228 }
2229 if (server.aof_state != REDIS_AOF_OFF) {
2230 mem_used -= sdslen(server.aof_buf);
2231 mem_used -= aofRewriteBufferSize();
2232 }
2233
2234 /* Check if we are over the memory limit. */
2235 if (mem_used <= server.maxmemory) return REDIS_OK;
2236
2237 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2238 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2239
2240 /* Compute how much memory we need to free. */
2241 mem_tofree = mem_used - server.maxmemory;
2242 mem_freed = 0;
2243 while (mem_freed < mem_tofree) {
2244 int j, k, keys_freed = 0;
2245
2246 for (j = 0; j < server.dbnum; j++) {
2247 long bestval = 0; /* just to prevent warning */
2248 sds bestkey = NULL;
2249 struct dictEntry *de;
2250 redisDb *db = server.db+j;
2251 dict *dict;
2252
2253 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2254 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2255 {
2256 dict = server.db[j].dict;
2257 } else {
2258 dict = server.db[j].expires;
2259 }
2260 if (dictSize(dict) == 0) continue;
2261
2262 /* volatile-random and allkeys-random policy */
2263 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2264 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2265 {
2266 de = dictGetRandomKey(dict);
2267 bestkey = dictGetKey(de);
2268 }
2269
2270 /* volatile-lru and allkeys-lru policy */
2271 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2272 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2273 {
2274 for (k = 0; k < server.maxmemory_samples; k++) {
2275 sds thiskey;
2276 long thisval;
2277 robj *o;
2278
2279 de = dictGetRandomKey(dict);
2280 thiskey = dictGetKey(de);
2281 /* When policy is volatile-lru we need an additional lookup
2282 * to locate the real key, as dict is set to db->expires. */
2283 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2284 de = dictFind(db->dict, thiskey);
2285 o = dictGetVal(de);
2286 thisval = estimateObjectIdleTime(o);
2287
2288 /* Higher idle time is better candidate for deletion */
2289 if (bestkey == NULL || thisval > bestval) {
2290 bestkey = thiskey;
2291 bestval = thisval;
2292 }
2293 }
2294 }
2295
2296 /* volatile-ttl */
2297 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2298 for (k = 0; k < server.maxmemory_samples; k++) {
2299 sds thiskey;
2300 long thisval;
2301
2302 de = dictGetRandomKey(dict);
2303 thiskey = dictGetKey(de);
2304 thisval = (long) dictGetVal(de);
2305
2306 /* Expire sooner (minor expire unix timestamp) is better
2307 * candidate for deletion */
2308 if (bestkey == NULL || thisval < bestval) {
2309 bestkey = thiskey;
2310 bestval = thisval;
2311 }
2312 }
2313 }
2314
2315 /* Finally remove the selected key. */
2316 if (bestkey) {
2317 long long delta;
2318
2319 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2320 propagateExpire(db,keyobj);
2321 /* We compute the amount of memory freed by dbDelete() alone.
2322 * It is possible that actually the memory needed to propagate
2323 * the DEL in AOF and replication link is greater than the one
2324 * we are freeing removing the key, but we can't account for
2325 * that otherwise we would never exit the loop.
2326 *
2327 * AOF and Output buffer memory will be freed eventually so
2328 * we only care about memory used by the key space. */
2329 delta = (long long) zmalloc_used_memory();
2330 dbDelete(db,keyobj);
2331 delta -= (long long) zmalloc_used_memory();
2332 mem_freed += delta;
2333 server.stat_evictedkeys++;
2334 decrRefCount(keyobj);
2335 keys_freed++;
2336
2337 /* When the memory to free starts to be big enough, we may
2338 * start spending so much time here that is impossible to
2339 * deliver data to the slaves fast enough, so we force the
2340 * transmission here inside the loop. */
2341 if (slaves) flushSlavesOutputBuffers();
2342 }
2343 }
2344 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2345 }
2346 return REDIS_OK;
2347}
2348
2349/* =================================== Main! ================================ */
2350
2351#ifdef __linux__
2352int linuxOvercommitMemoryValue(void) {
2353 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2354 char buf[64];
2355
2356 if (!fp) return -1;
2357 if (fgets(buf,64,fp) == NULL) {
2358 fclose(fp);
2359 return -1;
2360 }
2361 fclose(fp);
2362
2363 return atoi(buf);
2364}
2365
2366void linuxOvercommitMemoryWarning(void) {
2367 if (linuxOvercommitMemoryValue() == 0) {
2368 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.");
2369 }
2370}
2371#endif /* __linux__ */
2372
2373void createPidFile(void) {
2374 /* Try to write the pid file in a best-effort way. */
2375 FILE *fp = fopen(server.pidfile,"w");
2376 if (fp) {
2377 fprintf(fp,"%d\n",(int)getpid());
2378 fclose(fp);
2379 }
2380}
2381
2382void daemonize(void) {
2383 int fd;
2384
2385 if (fork() != 0) exit(0); /* parent exits */
2386 setsid(); /* create a new session */
2387
2388 /* Every output goes to /dev/null. If Redis is daemonized but
2389 * the 'logfile' is set to 'stdout' in the configuration file
2390 * it will not log at all. */
2391 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2392 dup2(fd, STDIN_FILENO);
2393 dup2(fd, STDOUT_FILENO);
2394 dup2(fd, STDERR_FILENO);
2395 if (fd > STDERR_FILENO) close(fd);
2396 }
2397}
2398
2399void version() {
2400 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2401 REDIS_VERSION,
2402 redisGitSHA1(),
2403 atoi(redisGitDirty()) > 0,
2404 ZMALLOC_LIB,
2405 sizeof(long) == 4 ? 32 : 64);
2406 exit(0);
2407}
2408
2409void usage() {
2410 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2411 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2412 fprintf(stderr," ./redis-server -v or --version\n");
2413 fprintf(stderr," ./redis-server -h or --help\n");
2414 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2415 fprintf(stderr,"Examples:\n");
2416 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2417 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2418 fprintf(stderr," ./redis-server --port 7777\n");
2419 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2420 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n\n");
2421 fprintf(stderr,"Sentinel mode:\n");
2422 fprintf(stderr," ./redis-server /etc/sentinel.conf --sentinel\n");
2423 exit(1);
2424}
2425
2426void redisAsciiArt(void) {
2427#include "asciilogo.h"
2428 char *buf = zmalloc(1024*16);
2429 char *mode = "stand alone";
2430
2431 if (server.sentinel_mode) mode = "sentinel";
2432
2433 snprintf(buf,1024*16,ascii_logo,
2434 REDIS_VERSION,
2435 redisGitSHA1(),
2436 strtol(redisGitDirty(),NULL,10) > 0,
2437 (sizeof(long) == 8) ? "64" : "32",
2438 mode, server.port,
2439 (long) getpid()
2440 );
2441 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2442 zfree(buf);
2443}
2444
2445static void sigtermHandler(int sig) {
2446 REDIS_NOTUSED(sig);
2447
2448 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2449 server.shutdown_asap = 1;
2450}
2451
2452void setupSignalHandlers(void) {
2453 struct sigaction act;
2454
2455 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2456 * Otherwise, sa_handler is used. */
2457 sigemptyset(&act.sa_mask);
2458 act.sa_flags = 0;
2459 act.sa_handler = sigtermHandler;
2460 sigaction(SIGTERM, &act, NULL);
2461
2462#ifdef HAVE_BACKTRACE
2463 sigemptyset(&act.sa_mask);
2464 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2465 act.sa_sigaction = sigsegvHandler;
2466 sigaction(SIGSEGV, &act, NULL);
2467 sigaction(SIGBUS, &act, NULL);
2468 sigaction(SIGFPE, &act, NULL);
2469 sigaction(SIGILL, &act, NULL);
2470#endif
2471 return;
2472}
2473
2474void memtest(size_t megabytes, int passes);
2475
2476/* Returns 1 if there is --sentinel among the arguments or if
2477 * argv[0] is exactly "redis-sentinel". */
2478int checkForSentinelMode(int argc, char **argv) {
2479 int j;
2480
2481 if (strstr(argv[0],"redis-sentinel") != NULL) return 1;
2482 for (j = 1; j < argc; j++)
2483 if (!strcmp(argv[j],"--sentinel")) return 1;
2484 return 0;
2485}
2486
2487/* Function called at startup to load RDB or AOF file in memory. */
2488void loadDataFromDisk(void) {
2489 long long start = ustime();
2490 if (server.aof_state == REDIS_AOF_ON) {
2491 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2492 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2493 } else {
2494 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2495 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2496 (float)(ustime()-start)/1000000);
2497 } else if (errno != ENOENT) {
2498 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2499 exit(1);
2500 }
2501 }
2502}
2503
2504void redisOutOfMemoryHandler(size_t allocation_size) {
2505 redisLog(REDIS_WARNING,"Out Of Memory allocating %zu bytes!",
2506 allocation_size);
2507 redisPanic("OOM");
2508}
2509
2510int main(int argc, char **argv) {
2511 struct timeval tv;
2512
2513 /* We need to initialize our libraries, and the server configuration. */
2514 zmalloc_enable_thread_safeness();
2515 zmalloc_set_oom_handler(redisOutOfMemoryHandler);
2516 srand(time(NULL)^getpid());
2517 gettimeofday(&tv,NULL);
2518 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2519 server.sentinel_mode = checkForSentinelMode(argc,argv);
2520 initServerConfig();
2521
2522 /* We need to init sentinel right now as parsing the configuration file
2523 * in sentinel mode will have the effect of populating the sentinel
2524 * data structures with master nodes to monitor. */
2525 if (server.sentinel_mode) {
2526 initSentinelConfig();
2527 initSentinel();
2528 }
2529
2530 if (argc >= 2) {
2531 int j = 1; /* First option to parse in argv[] */
2532 sds options = sdsempty();
2533 char *configfile = NULL;
2534
2535 /* Handle special options --help and --version */
2536 if (strcmp(argv[1], "-v") == 0 ||
2537 strcmp(argv[1], "--version") == 0) version();
2538 if (strcmp(argv[1], "--help") == 0 ||
2539 strcmp(argv[1], "-h") == 0) usage();
2540 if (strcmp(argv[1], "--test-memory") == 0) {
2541 if (argc == 3) {
2542 memtest(atoi(argv[2]),50);
2543 exit(0);
2544 } else {
2545 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2546 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2547 exit(1);
2548 }
2549 }
2550
2551 /* First argument is the config file name? */
2552 if (argv[j][0] != '-' || argv[j][1] != '-')
2553 configfile = argv[j++];
2554 /* All the other options are parsed and conceptually appended to the
2555 * configuration file. For instance --port 6380 will generate the
2556 * string "port 6380\n" to be parsed after the actual file name
2557 * is parsed, if any. */
2558 while(j != argc) {
2559 if (argv[j][0] == '-' && argv[j][1] == '-') {
2560 /* Option name */
2561 if (sdslen(options)) options = sdscat(options,"\n");
2562 options = sdscat(options,argv[j]+2);
2563 options = sdscat(options," ");
2564 } else {
2565 /* Option argument */
2566 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2567 options = sdscat(options," ");
2568 }
2569 j++;
2570 }
2571 resetServerSaveParams();
2572 loadServerConfig(configfile,options);
2573 sdsfree(options);
2574 } else {
2575 redisLog(REDIS_WARNING, "Warning: no config file specified, using the default config. In order to specify a config file use %s /path/to/%s.conf", argv[0], server.sentinel_mode ? "sentinel" : "redis");
2576 }
2577 if (server.daemonize) daemonize();
2578 initServer();
2579 if (server.daemonize) createPidFile();
2580 redisAsciiArt();
2581
2582 if (!server.sentinel_mode) {
2583 /* Things only needed when not running in Sentinel mode. */
2584 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2585 #ifdef __linux__
2586 linuxOvercommitMemoryWarning();
2587 #endif
2588 loadDataFromDisk();
2589 if (server.ipfd > 0)
2590 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2591 if (server.sofd > 0)
2592 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2593 }
2594
2595 /* Warning the user about suspicious maxmemory setting. */
2596 if (server.maxmemory > 0 && server.maxmemory < 1024*1024) {
2597 redisLog(REDIS_WARNING,"WARNING: You specified a maxmemory value that is less than 1MB (current value is %llu bytes). Are you sure this is what you really want?", server.maxmemory);
2598 }
2599
2600 aeSetBeforeSleepProc(server.el,beforeSleep);
2601 aeMain(server.el);
2602 aeDeleteEventLoop(server.el);
2603 return 0;
2604}
2605
2606/* The End */