]> git.saurik.com Git - redis.git/blob - src/redis.c
Merge pull request #461 from schlenk/unstable
[redis.git] / src / redis.c
1 /*
2 * Copyright (c) 2009-2010, Salvatore Sanfilippo <antirez at gmail dot com>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are met:
7 *
8 * * Redistributions of source code must retain the above copyright notice,
9 * this list of conditions and the following disclaimer.
10 * * Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * * Neither the name of Redis nor the names of its contributors may be used
14 * to endorse or promote products derived from this software without
15 * specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27 * POSSIBILITY OF SUCH DAMAGE.
28 */
29
30 #include "redis.h"
31 #include "slowlog.h"
32 #include "bio.h"
33
34 #include <time.h>
35 #include <signal.h>
36 #include <sys/wait.h>
37 #include <errno.h>
38 #include <assert.h>
39 #include <ctype.h>
40 #include <stdarg.h>
41 #include <arpa/inet.h>
42 #include <sys/stat.h>
43 #include <fcntl.h>
44 #include <sys/time.h>
45 #include <sys/resource.h>
46 #include <sys/uio.h>
47 #include <limits.h>
48 #include <float.h>
49 #include <math.h>
50 #include <sys/resource.h>
51 #include <sys/utsname.h>
52
53 /* Our shared "common" objects */
54
55 struct sharedObjectsStruct shared;
56
57 /* Global vars that are actually used as constants. The following double
58 * values are used for double on-disk serialization, and are initialized
59 * at runtime to avoid strange compiler optimizations. */
60
61 double R_Zero, R_PosInf, R_NegInf, R_Nan;
62
63 /*================================= Globals ================================= */
64
65 /* Alternate stack for SIGSEGV/etc handlers */
66 char altstack[SIGSTKSZ];
67
68 /* Global vars */
69 struct redisServer server; /* server global state */
70 struct redisCommand *commandTable;
71
72 /* Our command table.
73 *
74 * Every entry is composed of the following fields:
75 *
76 * name: a string representing the command name.
77 * function: pointer to the C function implementing the command.
78 * arity: number of arguments, it is possible to use -N to say >= N
79 * sflags: command flags as string. See below for a table of flags.
80 * flags: flags as bitmask. Computed by Redis using the 'sflags' field.
81 * get_keys_proc: an optional function to get key arguments from a command.
82 * This is only used when the following three fields are not
83 * enough to specify what arguments are keys.
84 * first_key_index: first argument that is a key
85 * last_key_index: last argument that is a key
86 * key_step: step to get all the keys from first to last argument. For instance
87 * in MSET the step is two since arguments are key,val,key,val,...
88 * microseconds: microseconds of total execution time for this command.
89 * calls: total number of calls of this command.
90 *
91 * The flags, microseconds and calls fields are computed by Redis and should
92 * always be set to zero.
93 *
94 * Command flags are expressed using strings where every character represents
95 * a flag. Later the populateCommandTable() function will take care of
96 * populating the real 'flags' field using this characters.
97 *
98 * This is the meaning of the flags:
99 *
100 * w: write command (may modify the key space).
101 * r: read command (will never modify the key space).
102 * m: may increase memory usage once called. Don't allow if out of memory.
103 * a: admin command, like SAVE or SHUTDOWN.
104 * p: Pub/Sub related command.
105 * f: force replication of this command, regarless of server.dirty.
106 * s: command not allowed in scripts.
107 * R: random command. Command is not deterministic, that is, the same command
108 * with the same arguments, with the same key space, may have different
109 * results. For instance SPOP and RANDOMKEY are two random commands.
110 * S: Sort command output array if called from script, so that the output
111 * is deterministic.
112 */
113 struct redisCommand redisCommandTable[] = {
114 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
115 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
116 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
117 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
118 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
119 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
120 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
121 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
122 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
123 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
124 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
125 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
126 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
127 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
128 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
129 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
130 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
131 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
132 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
133 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
134 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
135 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
136 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
137 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
138 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
139 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
140 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
141 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
142 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
143 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
144 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
145 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
146 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
147 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
148 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
149 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
150 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
151 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
152 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
153 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
154 {"srandmember",srandmemberCommand,2,"rR",0,NULL,1,1,1,0,0},
155 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
156 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
157 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
158 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
159 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
160 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
161 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
162 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
163 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
164 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
165 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
166 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
167 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
168 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
169 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
170 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
171 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
172 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
173 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
174 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
175 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
176 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
177 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
178 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
179 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
180 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
181 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
182 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
183 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
184 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
185 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
186 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
187 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
188 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
189 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
190 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
191 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
192 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
193 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
194 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
195 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
196 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
197 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
198 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
199 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
200 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
201 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
202 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
203 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
204 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
205 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
206 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
207 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
208 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
209 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
210 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
211 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
212 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
213 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
214 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
215 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
216 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
217 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
218 {"exec",execCommand,1,"s",0,NULL,0,0,0,0,0},
219 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
220 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
221 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
222 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
223 {"sort",sortCommand,-2,"wmS",0,NULL,1,1,1,0,0},
224 {"info",infoCommand,-1,"r",0,NULL,0,0,0,0,0},
225 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
226 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
227 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
228 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
229 {"slaveof",slaveofCommand,3,"as",0,NULL,0,0,0,0,0},
230 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
231 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
232 {"subscribe",subscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
233 {"unsubscribe",unsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
234 {"psubscribe",psubscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
235 {"punsubscribe",punsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
236 {"publish",publishCommand,3,"pf",0,NULL,0,0,0,0,0},
237 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
238 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
239 {"cluster",clusterCommand,-2,"ar",0,NULL,0,0,0,0,0},
240 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
241 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
242 {"asking",askingCommand,1,"r",0,NULL,0,0,0,0,0},
243 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
244 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
245 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
246 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
247 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
248 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
249 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
250 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0}
251 };
252
253 /*============================ Utility functions ============================ */
254
255 /* Low level logging. To use only for very big messages, otherwise
256 * redisLog() is to prefer. */
257 void redisLogRaw(int level, const char *msg) {
258 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
259 const char *c = ".-*#";
260 FILE *fp;
261 char buf[64];
262 int rawmode = (level & REDIS_LOG_RAW);
263
264 level &= 0xff; /* clear flags */
265 if (level < server.verbosity) return;
266
267 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
268 if (!fp) return;
269
270 if (rawmode) {
271 fprintf(fp,"%s",msg);
272 } else {
273 int off;
274 struct timeval tv;
275
276 gettimeofday(&tv,NULL);
277 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
278 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
279 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
280 }
281 fflush(fp);
282
283 if (server.logfile) fclose(fp);
284
285 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
286 }
287
288 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
289 * is used across the code. The raw version is only used in order to dump
290 * the INFO output on crash. */
291 void redisLog(int level, const char *fmt, ...) {
292 va_list ap;
293 char msg[REDIS_MAX_LOGMSG_LEN];
294
295 if ((level&0xff) < server.verbosity) return;
296
297 va_start(ap, fmt);
298 vsnprintf(msg, sizeof(msg), fmt, ap);
299 va_end(ap);
300
301 redisLogRaw(level,msg);
302 }
303
304 /* Log a fixed message without printf-alike capabilities, in a way that is
305 * safe to call from a signal handler.
306 *
307 * We actually use this only for signals that are not fatal from the point
308 * of view of Redis. Signals that are going to kill the server anyway and
309 * where we need printf-alike features are served by redisLog(). */
310 void redisLogFromHandler(int level, const char *msg) {
311 int fd;
312 char buf[64];
313
314 if ((level&0xff) < server.verbosity ||
315 (server.logfile == NULL && server.daemonize)) return;
316 fd = server.logfile ?
317 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
318 STDOUT_FILENO;
319 if (fd == -1) return;
320 ll2string(buf,sizeof(buf),getpid());
321 if (write(fd,"[",1) == -1) goto err;
322 if (write(fd,buf,strlen(buf)) == -1) goto err;
323 if (write(fd," | signal handler] (",20) == -1) goto err;
324 ll2string(buf,sizeof(buf),time(NULL));
325 if (write(fd,buf,strlen(buf)) == -1) goto err;
326 if (write(fd,") ",2) == -1) goto err;
327 if (write(fd,msg,strlen(msg)) == -1) goto err;
328 if (write(fd,"\n",1) == -1) goto err;
329 err:
330 if (server.logfile) close(fd);
331 }
332
333 /* Redis generally does not try to recover from out of memory conditions
334 * when allocating objects or strings, it is not clear if it will be possible
335 * to report this condition to the client since the networking layer itself
336 * is based on heap allocation for send buffers, so we simply abort.
337 * At least the code will be simpler to read... */
338 void oom(const char *msg) {
339 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
340 sleep(1);
341 abort();
342 }
343
344 /* Return the UNIX time in microseconds */
345 long long ustime(void) {
346 struct timeval tv;
347 long long ust;
348
349 gettimeofday(&tv, NULL);
350 ust = ((long long)tv.tv_sec)*1000000;
351 ust += tv.tv_usec;
352 return ust;
353 }
354
355 /* Return the UNIX time in milliseconds */
356 long long mstime(void) {
357 return ustime()/1000;
358 }
359
360 /* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
361 * exit(), because the latter may interact with the same file objects used by
362 * the parent process. However if we are testing the coverage normal exit() is
363 * used in order to obtain the right coverage information. */
364 void exitFromChild(int retcode) {
365 #ifdef COVERAGE_TEST
366 exit(retcode);
367 #else
368 _exit(retcode);
369 #endif
370 }
371
372 /*====================== Hash table type implementation ==================== */
373
374 /* This is an hash table type that uses the SDS dynamic strings libary as
375 * keys and radis objects as values (objects can hold SDS strings,
376 * lists, sets). */
377
378 void dictVanillaFree(void *privdata, void *val)
379 {
380 DICT_NOTUSED(privdata);
381 zfree(val);
382 }
383
384 void dictListDestructor(void *privdata, void *val)
385 {
386 DICT_NOTUSED(privdata);
387 listRelease((list*)val);
388 }
389
390 int dictSdsKeyCompare(void *privdata, const void *key1,
391 const void *key2)
392 {
393 int l1,l2;
394 DICT_NOTUSED(privdata);
395
396 l1 = sdslen((sds)key1);
397 l2 = sdslen((sds)key2);
398 if (l1 != l2) return 0;
399 return memcmp(key1, key2, l1) == 0;
400 }
401
402 /* A case insensitive version used for the command lookup table. */
403 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
404 const void *key2)
405 {
406 DICT_NOTUSED(privdata);
407
408 return strcasecmp(key1, key2) == 0;
409 }
410
411 void dictRedisObjectDestructor(void *privdata, void *val)
412 {
413 DICT_NOTUSED(privdata);
414
415 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
416 decrRefCount(val);
417 }
418
419 void dictSdsDestructor(void *privdata, void *val)
420 {
421 DICT_NOTUSED(privdata);
422
423 sdsfree(val);
424 }
425
426 int dictObjKeyCompare(void *privdata, const void *key1,
427 const void *key2)
428 {
429 const robj *o1 = key1, *o2 = key2;
430 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
431 }
432
433 unsigned int dictObjHash(const void *key) {
434 const robj *o = key;
435 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
436 }
437
438 unsigned int dictSdsHash(const void *key) {
439 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
440 }
441
442 unsigned int dictSdsCaseHash(const void *key) {
443 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
444 }
445
446 int dictEncObjKeyCompare(void *privdata, const void *key1,
447 const void *key2)
448 {
449 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
450 int cmp;
451
452 if (o1->encoding == REDIS_ENCODING_INT &&
453 o2->encoding == REDIS_ENCODING_INT)
454 return o1->ptr == o2->ptr;
455
456 o1 = getDecodedObject(o1);
457 o2 = getDecodedObject(o2);
458 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
459 decrRefCount(o1);
460 decrRefCount(o2);
461 return cmp;
462 }
463
464 unsigned int dictEncObjHash(const void *key) {
465 robj *o = (robj*) key;
466
467 if (o->encoding == REDIS_ENCODING_RAW) {
468 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
469 } else {
470 if (o->encoding == REDIS_ENCODING_INT) {
471 char buf[32];
472 int len;
473
474 len = ll2string(buf,32,(long)o->ptr);
475 return dictGenHashFunction((unsigned char*)buf, len);
476 } else {
477 unsigned int hash;
478
479 o = getDecodedObject(o);
480 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
481 decrRefCount(o);
482 return hash;
483 }
484 }
485 }
486
487 /* Sets type hash table */
488 dictType setDictType = {
489 dictEncObjHash, /* hash function */
490 NULL, /* key dup */
491 NULL, /* val dup */
492 dictEncObjKeyCompare, /* key compare */
493 dictRedisObjectDestructor, /* key destructor */
494 NULL /* val destructor */
495 };
496
497 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
498 dictType zsetDictType = {
499 dictEncObjHash, /* hash function */
500 NULL, /* key dup */
501 NULL, /* val dup */
502 dictEncObjKeyCompare, /* key compare */
503 dictRedisObjectDestructor, /* key destructor */
504 NULL /* val destructor */
505 };
506
507 /* Db->dict, keys are sds strings, vals are Redis objects. */
508 dictType dbDictType = {
509 dictSdsHash, /* hash function */
510 NULL, /* key dup */
511 NULL, /* val dup */
512 dictSdsKeyCompare, /* key compare */
513 dictSdsDestructor, /* key destructor */
514 dictRedisObjectDestructor /* val destructor */
515 };
516
517 /* Db->expires */
518 dictType keyptrDictType = {
519 dictSdsHash, /* hash function */
520 NULL, /* key dup */
521 NULL, /* val dup */
522 dictSdsKeyCompare, /* key compare */
523 NULL, /* key destructor */
524 NULL /* val destructor */
525 };
526
527 /* Command table. sds string -> command struct pointer. */
528 dictType commandTableDictType = {
529 dictSdsCaseHash, /* hash function */
530 NULL, /* key dup */
531 NULL, /* val dup */
532 dictSdsKeyCaseCompare, /* key compare */
533 dictSdsDestructor, /* key destructor */
534 NULL /* val destructor */
535 };
536
537 /* Hash type hash table (note that small hashes are represented with zimpaps) */
538 dictType hashDictType = {
539 dictEncObjHash, /* hash function */
540 NULL, /* key dup */
541 NULL, /* val dup */
542 dictEncObjKeyCompare, /* key compare */
543 dictRedisObjectDestructor, /* key destructor */
544 dictRedisObjectDestructor /* val destructor */
545 };
546
547 /* Keylist hash table type has unencoded redis objects as keys and
548 * lists as values. It's used for blocking operations (BLPOP) and to
549 * map swapped keys to a list of clients waiting for this keys to be loaded. */
550 dictType keylistDictType = {
551 dictObjHash, /* hash function */
552 NULL, /* key dup */
553 NULL, /* val dup */
554 dictObjKeyCompare, /* key compare */
555 dictRedisObjectDestructor, /* key destructor */
556 dictListDestructor /* val destructor */
557 };
558
559 /* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
560 * clusterNode structures. */
561 dictType clusterNodesDictType = {
562 dictSdsHash, /* hash function */
563 NULL, /* key dup */
564 NULL, /* val dup */
565 dictSdsKeyCompare, /* key compare */
566 dictSdsDestructor, /* key destructor */
567 NULL /* val destructor */
568 };
569
570 int htNeedsResize(dict *dict) {
571 long long size, used;
572
573 size = dictSlots(dict);
574 used = dictSize(dict);
575 return (size && used && size > DICT_HT_INITIAL_SIZE &&
576 (used*100/size < REDIS_HT_MINFILL));
577 }
578
579 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
580 * we resize the hash table to save memory */
581 void tryResizeHashTables(void) {
582 int j;
583
584 for (j = 0; j < server.dbnum; j++) {
585 if (htNeedsResize(server.db[j].dict))
586 dictResize(server.db[j].dict);
587 if (htNeedsResize(server.db[j].expires))
588 dictResize(server.db[j].expires);
589 }
590 }
591
592 /* Our hash table implementation performs rehashing incrementally while
593 * we write/read from the hash table. Still if the server is idle, the hash
594 * table will use two tables for a long time. So we try to use 1 millisecond
595 * of CPU time at every serverCron() loop in order to rehash some key. */
596 void incrementallyRehash(void) {
597 int j;
598
599 for (j = 0; j < server.dbnum; j++) {
600 if (dictIsRehashing(server.db[j].dict)) {
601 dictRehashMilliseconds(server.db[j].dict,1);
602 break; /* already used our millisecond for this loop... */
603 }
604 }
605 }
606
607 /* This function is called once a background process of some kind terminates,
608 * as we want to avoid resizing the hash tables when there is a child in order
609 * to play well with copy-on-write (otherwise when a resize happens lots of
610 * memory pages are copied). The goal of this function is to update the ability
611 * for dict.c to resize the hash tables accordingly to the fact we have o not
612 * running childs. */
613 void updateDictResizePolicy(void) {
614 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
615 dictEnableResize();
616 else
617 dictDisableResize();
618 }
619
620 /* ======================= Cron: called every 100 ms ======================== */
621
622 /* Try to expire a few timed out keys. The algorithm used is adaptive and
623 * will use few CPU cycles if there are few expiring keys, otherwise
624 * it will get more aggressive to avoid that too much memory is used by
625 * keys that can be removed from the keyspace. */
626 void activeExpireCycle(void) {
627 int j;
628
629 for (j = 0; j < server.dbnum; j++) {
630 int expired;
631 redisDb *db = server.db+j;
632
633 /* Continue to expire if at the end of the cycle more than 25%
634 * of the keys were expired. */
635 do {
636 long num = dictSize(db->expires);
637 long long now = mstime();
638
639 expired = 0;
640 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
641 num = REDIS_EXPIRELOOKUPS_PER_CRON;
642 while (num--) {
643 dictEntry *de;
644 long long t;
645
646 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
647 t = dictGetSignedIntegerVal(de);
648 if (now > t) {
649 sds key = dictGetKey(de);
650 robj *keyobj = createStringObject(key,sdslen(key));
651
652 propagateExpire(db,keyobj);
653 dbDelete(db,keyobj);
654 decrRefCount(keyobj);
655 expired++;
656 server.stat_expiredkeys++;
657 }
658 }
659 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
660 }
661 }
662
663 void updateLRUClock(void) {
664 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
665 REDIS_LRU_CLOCK_MAX;
666 }
667
668
669 /* Add a sample to the operations per second array of samples. */
670 void trackOperationsPerSecond(void) {
671 long long t = mstime() - server.ops_sec_last_sample_time;
672 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
673 long long ops_sec;
674
675 ops_sec = t > 0 ? (ops*1000/t) : 0;
676
677 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
678 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
679 server.ops_sec_last_sample_time = mstime();
680 server.ops_sec_last_sample_ops = server.stat_numcommands;
681 }
682
683 /* Return the mean of all the samples. */
684 long long getOperationsPerSecond(void) {
685 int j;
686 long long sum = 0;
687
688 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
689 sum += server.ops_sec_samples[j];
690 return sum / REDIS_OPS_SEC_SAMPLES;
691 }
692
693 /* Check for timeouts. Returns non-zero if the client was terminated */
694 int clientsCronHandleTimeout(redisClient *c) {
695 time_t now = server.unixtime;
696
697 if (server.maxidletime &&
698 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
699 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
700 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
701 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
702 listLength(c->pubsub_patterns) == 0 &&
703 (now - c->lastinteraction > server.maxidletime))
704 {
705 redisLog(REDIS_VERBOSE,"Closing idle client");
706 freeClient(c);
707 return 1;
708 } else if (c->flags & REDIS_BLOCKED) {
709 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
710 addReply(c,shared.nullmultibulk);
711 unblockClientWaitingData(c);
712 }
713 }
714 return 0;
715 }
716
717 /* The client query buffer is an sds.c string that can end with a lot of
718 * free space not used, this function reclaims space if needed.
719 *
720 * The funciton always returns 0 as it never terminates the client. */
721 int clientsCronResizeQueryBuffer(redisClient *c) {
722 size_t querybuf_size = sdsAllocSize(c->querybuf);
723 time_t idletime = server.unixtime - c->lastinteraction;
724
725 /* There are two conditions to resize the query buffer:
726 * 1) Query buffer is > BIG_ARG and too big for latest peak.
727 * 2) Client is inactive and the buffer is bigger than 1k. */
728 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
729 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
730 (querybuf_size > 1024 && idletime > 2))
731 {
732 /* Only resize the query buffer if it is actually wasting space. */
733 if (sdsavail(c->querybuf) > 1024) {
734 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
735 }
736 }
737 /* Reset the peak again to capture the peak memory usage in the next
738 * cycle. */
739 c->querybuf_peak = 0;
740 return 0;
741 }
742
743 void clientsCron(void) {
744 /* Make sure to process at least 1/100 of clients per call.
745 * Since this function is called 10 times per second we are sure that
746 * in the worst case we process all the clients in 10 seconds.
747 * In normal conditions (a reasonable number of clients) we process
748 * all the clients in a shorter time. */
749 int numclients = listLength(server.clients);
750 int iterations = numclients/100;
751
752 if (iterations < 50)
753 iterations = (numclients < 50) ? numclients : 50;
754 while(listLength(server.clients) && iterations--) {
755 redisClient *c;
756 listNode *head;
757
758 /* Rotate the list, take the current head, process.
759 * This way if the client must be removed from the list it's the
760 * first element and we don't incur into O(N) computation. */
761 listRotate(server.clients);
762 head = listFirst(server.clients);
763 c = listNodeValue(head);
764 /* The following functions do different service checks on the client.
765 * The protocol is that they return non-zero if the client was
766 * terminated. */
767 if (clientsCronHandleTimeout(c)) continue;
768 if (clientsCronResizeQueryBuffer(c)) continue;
769 }
770 }
771
772 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
773 int j, loops = server.cronloops;
774 REDIS_NOTUSED(eventLoop);
775 REDIS_NOTUSED(id);
776 REDIS_NOTUSED(clientData);
777
778 /* Software watchdog: deliver the SIGALRM that will reach the signal
779 * handler if we don't return here fast enough. */
780 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
781
782 /* We take a cached value of the unix time in the global state because
783 * with virtual memory and aging there is to store the current time
784 * in objects at every object access, and accuracy is not needed.
785 * To access a global var is faster than calling time(NULL) */
786 server.unixtime = time(NULL);
787
788 trackOperationsPerSecond();
789
790 /* We have just 22 bits per object for LRU information.
791 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
792 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
793 *
794 * Note that even if this will wrap after 1.5 years it's not a problem,
795 * everything will still work but just some object will appear younger
796 * to Redis. But for this to happen a given object should never be touched
797 * for 1.5 years.
798 *
799 * Note that you can change the resolution altering the
800 * REDIS_LRU_CLOCK_RESOLUTION define.
801 */
802 updateLRUClock();
803
804 /* Record the max memory used since the server was started. */
805 if (zmalloc_used_memory() > server.stat_peak_memory)
806 server.stat_peak_memory = zmalloc_used_memory();
807
808 /* We received a SIGTERM, shutting down here in a safe way, as it is
809 * not ok doing so inside the signal handler. */
810 if (server.shutdown_asap) {
811 if (prepareForShutdown(0) == REDIS_OK) exit(0);
812 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
813 }
814
815 /* Show some info about non-empty databases */
816 for (j = 0; j < server.dbnum; j++) {
817 long long size, used, vkeys;
818
819 size = dictSlots(server.db[j].dict);
820 used = dictSize(server.db[j].dict);
821 vkeys = dictSize(server.db[j].expires);
822 if (!(loops % 50) && (used || vkeys)) {
823 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
824 /* dictPrintStats(server.dict); */
825 }
826 }
827
828 /* We don't want to resize the hash tables while a bacground saving
829 * is in progress: the saving child is created using fork() that is
830 * implemented with a copy-on-write semantic in most modern systems, so
831 * if we resize the HT while there is the saving child at work actually
832 * a lot of memory movements in the parent will cause a lot of pages
833 * copied. */
834 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
835 if (!(loops % 10)) tryResizeHashTables();
836 if (server.activerehashing) incrementallyRehash();
837 }
838
839 /* Show information about connected clients */
840 if (!(loops % 50)) {
841 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
842 listLength(server.clients)-listLength(server.slaves),
843 listLength(server.slaves),
844 zmalloc_used_memory());
845 }
846
847 /* We need to do a few operations on clients asynchronously. */
848 clientsCron();
849
850 /* Start a scheduled AOF rewrite if this was requested by the user while
851 * a BGSAVE was in progress. */
852 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
853 server.aof_rewrite_scheduled)
854 {
855 rewriteAppendOnlyFileBackground();
856 }
857
858 /* Check if a background saving or AOF rewrite in progress terminated. */
859 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
860 int statloc;
861 pid_t pid;
862
863 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
864 int exitcode = WEXITSTATUS(statloc);
865 int bysignal = 0;
866
867 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
868
869 if (pid == server.rdb_child_pid) {
870 backgroundSaveDoneHandler(exitcode,bysignal);
871 } else {
872 backgroundRewriteDoneHandler(exitcode,bysignal);
873 }
874 updateDictResizePolicy();
875 }
876 } else {
877 /* If there is not a background saving/rewrite in progress check if
878 * we have to save/rewrite now */
879 for (j = 0; j < server.saveparamslen; j++) {
880 struct saveparam *sp = server.saveparams+j;
881
882 if (server.dirty >= sp->changes &&
883 server.unixtime-server.lastsave > sp->seconds) {
884 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
885 sp->changes, sp->seconds);
886 rdbSaveBackground(server.rdb_filename);
887 break;
888 }
889 }
890
891 /* Trigger an AOF rewrite if needed */
892 if (server.rdb_child_pid == -1 &&
893 server.aof_child_pid == -1 &&
894 server.aof_rewrite_perc &&
895 server.aof_current_size > server.aof_rewrite_min_size)
896 {
897 long long base = server.aof_rewrite_base_size ?
898 server.aof_rewrite_base_size : 1;
899 long long growth = (server.aof_current_size*100/base) - 100;
900 if (growth >= server.aof_rewrite_perc) {
901 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
902 rewriteAppendOnlyFileBackground();
903 }
904 }
905 }
906
907
908 /* If we postponed an AOF buffer flush, let's try to do it every time the
909 * cron function is called. */
910 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
911
912 /* Expire a few keys per cycle, only if this is a master.
913 * On slaves we wait for DEL operations synthesized by the master
914 * in order to guarantee a strict consistency. */
915 if (server.masterhost == NULL) activeExpireCycle();
916
917 /* Close clients that need to be closed asynchronous */
918 freeClientsInAsyncFreeQueue();
919
920 /* Replication cron function -- used to reconnect to master and
921 * to detect transfer failures. */
922 if (!(loops % 10)) replicationCron();
923
924 /* Run other sub-systems specific cron jobs */
925 if (server.cluster_enabled && !(loops % 10)) clusterCron();
926
927 server.cronloops++;
928 return 100;
929 }
930
931 /* This function gets called every time Redis is entering the
932 * main loop of the event driven library, that is, before to sleep
933 * for ready file descriptors. */
934 void beforeSleep(struct aeEventLoop *eventLoop) {
935 REDIS_NOTUSED(eventLoop);
936 listNode *ln;
937 redisClient *c;
938
939 /* Try to process pending commands for clients that were just unblocked. */
940 while (listLength(server.unblocked_clients)) {
941 ln = listFirst(server.unblocked_clients);
942 redisAssert(ln != NULL);
943 c = ln->value;
944 listDelNode(server.unblocked_clients,ln);
945 c->flags &= ~REDIS_UNBLOCKED;
946
947 /* Process remaining data in the input buffer. */
948 if (c->querybuf && sdslen(c->querybuf) > 0) {
949 server.current_client = c;
950 processInputBuffer(c);
951 server.current_client = NULL;
952 }
953 }
954
955 /* Write the AOF buffer on disk */
956 flushAppendOnlyFile(0);
957 }
958
959 /* =========================== Server initialization ======================== */
960
961 void createSharedObjects(void) {
962 int j;
963
964 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
965 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
966 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
967 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
968 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
969 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
970 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
971 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
972 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
973 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
974 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
975 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
976 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
977 "-ERR Operation against a key holding the wrong kind of value\r\n"));
978 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
979 "-ERR no such key\r\n"));
980 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
981 "-ERR syntax error\r\n"));
982 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
983 "-ERR source and destination objects are the same\r\n"));
984 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
985 "-ERR index out of range\r\n"));
986 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
987 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
988 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
989 "-LOADING Redis is loading the dataset in memory\r\n"));
990 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
991 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
992 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
993 "-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"));
994 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
995 "-READONLY You can't write against a read only slave.\r\n"));
996 shared.oomerr = createObject(REDIS_STRING,sdsnew(
997 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
998 shared.space = createObject(REDIS_STRING,sdsnew(" "));
999 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1000 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1001
1002 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1003 shared.select[j] = createObject(REDIS_STRING,
1004 sdscatprintf(sdsempty(),"select %d\r\n", j));
1005 }
1006 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1007 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1008 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1009 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1010 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1011 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1012 shared.del = createStringObject("DEL",3);
1013 shared.rpop = createStringObject("RPOP",4);
1014 shared.lpop = createStringObject("LPOP",4);
1015 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1016 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1017 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1018 }
1019 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1020 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1021 sdscatprintf(sdsempty(),"*%d\r\n",j));
1022 shared.bulkhdr[j] = createObject(REDIS_STRING,
1023 sdscatprintf(sdsempty(),"$%d\r\n",j));
1024 }
1025 }
1026
1027 void initServerConfig() {
1028 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1029 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1030 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1031 server.port = REDIS_SERVERPORT;
1032 server.bindaddr = NULL;
1033 server.unixsocket = NULL;
1034 server.unixsocketperm = 0;
1035 server.ipfd = -1;
1036 server.sofd = -1;
1037 server.dbnum = REDIS_DEFAULT_DBNUM;
1038 server.verbosity = REDIS_NOTICE;
1039 server.maxidletime = REDIS_MAXIDLETIME;
1040 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1041 server.saveparams = NULL;
1042 server.loading = 0;
1043 server.logfile = NULL; /* NULL = log on standard output */
1044 server.syslog_enabled = 0;
1045 server.syslog_ident = zstrdup("redis");
1046 server.syslog_facility = LOG_LOCAL0;
1047 server.daemonize = 0;
1048 server.aof_state = REDIS_AOF_OFF;
1049 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1050 server.aof_no_fsync_on_rewrite = 0;
1051 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1052 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1053 server.aof_rewrite_base_size = 0;
1054 server.aof_rewrite_scheduled = 0;
1055 server.aof_last_fsync = time(NULL);
1056 server.aof_delayed_fsync = 0;
1057 server.aof_fd = -1;
1058 server.aof_selected_db = -1; /* Make sure the first time will not match */
1059 server.aof_flush_postponed_start = 0;
1060 server.pidfile = zstrdup("/var/run/redis.pid");
1061 server.rdb_filename = zstrdup("dump.rdb");
1062 server.aof_filename = zstrdup("appendonly.aof");
1063 server.requirepass = NULL;
1064 server.rdb_compression = 1;
1065 server.rdb_checksum = 1;
1066 server.activerehashing = 1;
1067 server.maxclients = REDIS_MAX_CLIENTS;
1068 server.bpop_blocked_clients = 0;
1069 server.maxmemory = 0;
1070 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1071 server.maxmemory_samples = 3;
1072 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1073 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1074 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1075 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1076 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1077 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1078 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1079 server.shutdown_asap = 0;
1080 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1081 server.repl_timeout = REDIS_REPL_TIMEOUT;
1082 server.cluster_enabled = 0;
1083 server.cluster.configfile = zstrdup("nodes.conf");
1084 server.lua_caller = NULL;
1085 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1086 server.lua_client = NULL;
1087 server.lua_timedout = 0;
1088
1089 updateLRUClock();
1090 resetServerSaveParams();
1091
1092 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1093 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1094 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1095 /* Replication related */
1096 server.masterauth = NULL;
1097 server.masterhost = NULL;
1098 server.masterport = 6379;
1099 server.master = NULL;
1100 server.repl_state = REDIS_REPL_NONE;
1101 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1102 server.repl_serve_stale_data = 1;
1103 server.repl_slave_ro = 1;
1104 server.repl_down_since = time(NULL);
1105
1106 /* Client output buffer limits */
1107 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1108 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1109 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1110 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1111 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1112 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1113 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1114 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1115 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1116
1117 /* Double constants initialization */
1118 R_Zero = 0.0;
1119 R_PosInf = 1.0/R_Zero;
1120 R_NegInf = -1.0/R_Zero;
1121 R_Nan = R_Zero/R_Zero;
1122
1123 /* Command table -- we intiialize it here as it is part of the
1124 * initial configuration, since command names may be changed via
1125 * redis.conf using the rename-command directive. */
1126 server.commands = dictCreate(&commandTableDictType,NULL);
1127 populateCommandTable();
1128 server.delCommand = lookupCommandByCString("del");
1129 server.multiCommand = lookupCommandByCString("multi");
1130 server.lpushCommand = lookupCommandByCString("lpush");
1131
1132 /* Slow log */
1133 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1134 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1135
1136 /* Debugging */
1137 server.assert_failed = "<no assertion failed>";
1138 server.assert_file = "<no file>";
1139 server.assert_line = 0;
1140 server.bug_report_start = 0;
1141 server.watchdog_period = 0;
1142 }
1143
1144 /* This function will try to raise the max number of open files accordingly to
1145 * the configured max number of clients. It will also account for 32 additional
1146 * file descriptors as we need a few more for persistence, listening
1147 * sockets, log files and so forth.
1148 *
1149 * If it will not be possible to set the limit accordingly to the configured
1150 * max number of clients, the function will do the reverse setting
1151 * server.maxclients to the value that we can actually handle. */
1152 void adjustOpenFilesLimit(void) {
1153 rlim_t maxfiles = server.maxclients+32;
1154 struct rlimit limit;
1155
1156 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1157 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1158 strerror(errno));
1159 server.maxclients = 1024-32;
1160 } else {
1161 rlim_t oldlimit = limit.rlim_cur;
1162
1163 /* Set the max number of files if the current limit is not enough
1164 * for our needs. */
1165 if (oldlimit < maxfiles) {
1166 rlim_t f;
1167
1168 f = maxfiles;
1169 while(f > oldlimit) {
1170 limit.rlim_cur = f;
1171 limit.rlim_max = f;
1172 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1173 f -= 128;
1174 }
1175 if (f < oldlimit) f = oldlimit;
1176 if (f != maxfiles) {
1177 server.maxclients = f-32;
1178 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1179 (int) maxfiles, strerror(errno), (int) server.maxclients);
1180 } else {
1181 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1182 (int) maxfiles);
1183 }
1184 }
1185 }
1186 }
1187
1188 void initServer() {
1189 int j;
1190
1191 signal(SIGHUP, SIG_IGN);
1192 signal(SIGPIPE, SIG_IGN);
1193 setupSignalHandlers();
1194
1195 if (server.syslog_enabled) {
1196 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1197 server.syslog_facility);
1198 }
1199
1200 server.current_client = NULL;
1201 server.clients = listCreate();
1202 server.clients_to_close = listCreate();
1203 server.slaves = listCreate();
1204 server.monitors = listCreate();
1205 server.unblocked_clients = listCreate();
1206
1207 createSharedObjects();
1208 adjustOpenFilesLimit();
1209 server.el = aeCreateEventLoop(server.maxclients+1024);
1210 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1211
1212 if (server.port != 0) {
1213 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1214 if (server.ipfd == ANET_ERR) {
1215 redisLog(REDIS_WARNING, "Opening port %d: %s",
1216 server.port, server.neterr);
1217 exit(1);
1218 }
1219 }
1220 if (server.unixsocket != NULL) {
1221 unlink(server.unixsocket); /* don't care if this fails */
1222 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1223 if (server.sofd == ANET_ERR) {
1224 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1225 exit(1);
1226 }
1227 }
1228 if (server.ipfd < 0 && server.sofd < 0) {
1229 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1230 exit(1);
1231 }
1232 for (j = 0; j < server.dbnum; j++) {
1233 server.db[j].dict = dictCreate(&dbDictType,NULL);
1234 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1235 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1236 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1237 server.db[j].id = j;
1238 }
1239 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1240 server.pubsub_patterns = listCreate();
1241 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1242 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1243 server.cronloops = 0;
1244 server.rdb_child_pid = -1;
1245 server.aof_child_pid = -1;
1246 server.aof_rewrite_buf = sdsempty();
1247 server.aof_buf = sdsempty();
1248 server.lastsave = time(NULL);
1249 server.dirty = 0;
1250 server.stat_numcommands = 0;
1251 server.stat_numconnections = 0;
1252 server.stat_expiredkeys = 0;
1253 server.stat_evictedkeys = 0;
1254 server.stat_starttime = time(NULL);
1255 server.stat_keyspace_misses = 0;
1256 server.stat_keyspace_hits = 0;
1257 server.stat_peak_memory = 0;
1258 server.stat_fork_time = 0;
1259 server.stat_rejected_conn = 0;
1260 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1261 server.ops_sec_idx = 0;
1262 server.ops_sec_last_sample_time = mstime();
1263 server.ops_sec_last_sample_ops = 0;
1264 server.unixtime = time(NULL);
1265 server.lastbgsave_status = REDIS_OK;
1266 server.stop_writes_on_bgsave_err = 1;
1267 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1268 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1269 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1270 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1271 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1272
1273 if (server.aof_state == REDIS_AOF_ON) {
1274 server.aof_fd = open(server.aof_filename,
1275 O_WRONLY|O_APPEND|O_CREAT,0644);
1276 if (server.aof_fd == -1) {
1277 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1278 strerror(errno));
1279 exit(1);
1280 }
1281 }
1282
1283 /* 32 bit instances are limited to 4GB of address space, so if there is
1284 * no explicit limit in the user provided configuration we set a limit
1285 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1286 * useless crashes of the Redis instance. */
1287 if (server.arch_bits == 32 && server.maxmemory == 0) {
1288 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1289 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1290 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1291 }
1292
1293 if (server.cluster_enabled) clusterInit();
1294 scriptingInit();
1295 slowlogInit();
1296 bioInit();
1297 }
1298
1299 /* Populates the Redis Command Table starting from the hard coded list
1300 * we have on top of redis.c file. */
1301 void populateCommandTable(void) {
1302 int j;
1303 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1304
1305 for (j = 0; j < numcommands; j++) {
1306 struct redisCommand *c = redisCommandTable+j;
1307 char *f = c->sflags;
1308 int retval;
1309
1310 while(*f != '\0') {
1311 switch(*f) {
1312 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1313 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1314 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1315 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1316 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1317 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1318 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1319 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1320 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1321 default: redisPanic("Unsupported command flag"); break;
1322 }
1323 f++;
1324 }
1325
1326 retval = dictAdd(server.commands, sdsnew(c->name), c);
1327 assert(retval == DICT_OK);
1328 }
1329 }
1330
1331 void resetCommandTableStats(void) {
1332 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1333 int j;
1334
1335 for (j = 0; j < numcommands; j++) {
1336 struct redisCommand *c = redisCommandTable+j;
1337
1338 c->microseconds = 0;
1339 c->calls = 0;
1340 }
1341 }
1342
1343 /* ========================== Redis OP Array API ============================ */
1344
1345 void redisOpArrayInit(redisOpArray *oa) {
1346 oa->ops = NULL;
1347 oa->numops = 0;
1348 }
1349
1350 int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1351 robj **argv, int argc, int target)
1352 {
1353 redisOp *op;
1354
1355 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1356 op = oa->ops+oa->numops;
1357 op->cmd = cmd;
1358 op->dbid = dbid;
1359 op->argv = argv;
1360 op->argc = argc;
1361 op->target = target;
1362 oa->numops++;
1363 return oa->numops;
1364 }
1365
1366 void redisOpArrayFree(redisOpArray *oa) {
1367 while(oa->numops) {
1368 int j;
1369 redisOp *op;
1370
1371 oa->numops--;
1372 op = oa->ops+oa->numops;
1373 for (j = 0; j < op->argc; j++)
1374 decrRefCount(op->argv[j]);
1375 zfree(op->argv);
1376 }
1377 zfree(oa->ops);
1378 }
1379
1380 /* ====================== Commands lookup and execution ===================== */
1381
1382 struct redisCommand *lookupCommand(sds name) {
1383 return dictFetchValue(server.commands, name);
1384 }
1385
1386 struct redisCommand *lookupCommandByCString(char *s) {
1387 struct redisCommand *cmd;
1388 sds name = sdsnew(s);
1389
1390 cmd = dictFetchValue(server.commands, name);
1391 sdsfree(name);
1392 return cmd;
1393 }
1394
1395 /* Propagate the specified command (in the context of the specified database id)
1396 * to AOF, Slaves and Monitors.
1397 *
1398 * flags are an xor between:
1399 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1400 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1401 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1402 */
1403 void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1404 int flags)
1405 {
1406 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1407 feedAppendOnlyFile(cmd,dbid,argv,argc);
1408 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1409 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1410 }
1411
1412 /* Used inside commands to schedule the propagation of additional commands
1413 * after the current command is propagated to AOF / Replication. */
1414 void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1415 int target)
1416 {
1417 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1418 }
1419
1420 /* Call() is the core of Redis execution of a command */
1421 void call(redisClient *c, int flags) {
1422 long long dirty, start = ustime(), duration;
1423
1424 /* Sent the command to clients in MONITOR mode, only if the commands are
1425 * not geneated from reading an AOF. */
1426 if (listLength(server.monitors) && !server.loading)
1427 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1428
1429 /* Call the command. */
1430 redisOpArrayInit(&server.also_propagate);
1431 dirty = server.dirty;
1432 c->cmd->proc(c);
1433 dirty = server.dirty-dirty;
1434 duration = ustime()-start;
1435
1436 /* When EVAL is called loading the AOF we don't want commands called
1437 * from Lua to go into the slowlog or to populate statistics. */
1438 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1439 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1440
1441 /* Log the command into the Slow log if needed, and populate the
1442 * per-command statistics that we show in INFO commandstats. */
1443 if (flags & REDIS_CALL_SLOWLOG)
1444 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1445 if (flags & REDIS_CALL_STATS) {
1446 c->cmd->microseconds += duration;
1447 c->cmd->calls++;
1448 }
1449
1450 /* Propagate the command into the AOF and replication link */
1451 if (flags & REDIS_CALL_PROPAGATE) {
1452 int flags = REDIS_PROPAGATE_NONE;
1453
1454 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1455 flags |= REDIS_PROPAGATE_REPL;
1456 if (dirty)
1457 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1458 if (flags != REDIS_PROPAGATE_NONE)
1459 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1460 }
1461 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1462 * PUSH command. */
1463 if (server.also_propagate.numops) {
1464 int j;
1465 redisOp *rop;
1466
1467 for (j = 0; j < server.also_propagate.numops; j++) {
1468 rop = &server.also_propagate.ops[j];
1469 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1470 }
1471 redisOpArrayFree(&server.also_propagate);
1472 }
1473 server.stat_numcommands++;
1474 }
1475
1476 /* If this function gets called we already read a whole
1477 * command, argments are in the client argv/argc fields.
1478 * processCommand() execute the command or prepare the
1479 * server for a bulk read from the client.
1480 *
1481 * If 1 is returned the client is still alive and valid and
1482 * and other operations can be performed by the caller. Otherwise
1483 * if 0 is returned the client was destroied (i.e. after QUIT). */
1484 int processCommand(redisClient *c) {
1485 /* The QUIT command is handled separately. Normal command procs will
1486 * go through checking for replication and QUIT will cause trouble
1487 * when FORCE_REPLICATION is enabled and would be implemented in
1488 * a regular command proc. */
1489 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1490 addReply(c,shared.ok);
1491 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1492 return REDIS_ERR;
1493 }
1494
1495 /* Now lookup the command and check ASAP about trivial error conditions
1496 * such as wrong arity, bad command name and so forth. */
1497 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1498 if (!c->cmd) {
1499 addReplyErrorFormat(c,"unknown command '%s'",
1500 (char*)c->argv[0]->ptr);
1501 return REDIS_OK;
1502 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1503 (c->argc < -c->cmd->arity)) {
1504 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1505 c->cmd->name);
1506 return REDIS_OK;
1507 }
1508
1509 /* Check if the user is authenticated */
1510 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1511 {
1512 addReplyError(c,"operation not permitted");
1513 return REDIS_OK;
1514 }
1515
1516 /* If cluster is enabled, redirect here */
1517 if (server.cluster_enabled &&
1518 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1519 int hashslot;
1520
1521 if (server.cluster.state != REDIS_CLUSTER_OK) {
1522 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1523 return REDIS_OK;
1524 } else {
1525 int ask;
1526 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1527 if (n == NULL) {
1528 addReplyError(c,"Multi keys request invalid in cluster");
1529 return REDIS_OK;
1530 } else if (n != server.cluster.myself) {
1531 addReplySds(c,sdscatprintf(sdsempty(),
1532 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1533 hashslot,n->ip,n->port));
1534 return REDIS_OK;
1535 }
1536 }
1537 }
1538
1539 /* Handle the maxmemory directive.
1540 *
1541 * First we try to free some memory if possible (if there are volatile
1542 * keys in the dataset). If there are not the only thing we can do
1543 * is returning an error. */
1544 if (server.maxmemory) {
1545 int retval = freeMemoryIfNeeded();
1546 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1547 addReply(c, shared.oomerr);
1548 return REDIS_OK;
1549 }
1550 }
1551
1552 /* Don't accept write commands if there are problems persisting on disk. */
1553 if (server.stop_writes_on_bgsave_err &&
1554 server.saveparamslen > 0
1555 && server.lastbgsave_status == REDIS_ERR &&
1556 c->cmd->flags & REDIS_CMD_WRITE)
1557 {
1558 addReply(c, shared.bgsaveerr);
1559 return REDIS_OK;
1560 }
1561
1562 /* Don't accept wirte commands if this is a read only slave. But
1563 * accept write commands if this is our master. */
1564 if (server.masterhost && server.repl_slave_ro &&
1565 !(c->flags & REDIS_MASTER) &&
1566 c->cmd->flags & REDIS_CMD_WRITE)
1567 {
1568 addReply(c, shared.roslaveerr);
1569 return REDIS_OK;
1570 }
1571
1572 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1573 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1574 &&
1575 c->cmd->proc != subscribeCommand &&
1576 c->cmd->proc != unsubscribeCommand &&
1577 c->cmd->proc != psubscribeCommand &&
1578 c->cmd->proc != punsubscribeCommand) {
1579 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1580 return REDIS_OK;
1581 }
1582
1583 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1584 * we are a slave with a broken link with master. */
1585 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1586 server.repl_serve_stale_data == 0 &&
1587 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1588 {
1589 addReplyError(c,
1590 "link with MASTER is down and slave-serve-stale-data is set to no");
1591 return REDIS_OK;
1592 }
1593
1594 /* Loading DB? Return an error if the command is not INFO */
1595 if (server.loading && c->cmd->proc != infoCommand) {
1596 addReply(c, shared.loadingerr);
1597 return REDIS_OK;
1598 }
1599
1600 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1601 if (server.lua_timedout &&
1602 !(c->cmd->proc == shutdownCommand &&
1603 c->argc == 2 &&
1604 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1605 !(c->cmd->proc == scriptCommand &&
1606 c->argc == 2 &&
1607 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1608 {
1609 addReply(c, shared.slowscripterr);
1610 return REDIS_OK;
1611 }
1612
1613 /* Exec the command */
1614 if (c->flags & REDIS_MULTI &&
1615 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1616 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1617 {
1618 queueMultiCommand(c);
1619 addReply(c,shared.queued);
1620 } else {
1621 call(c,REDIS_CALL_FULL);
1622 }
1623 return REDIS_OK;
1624 }
1625
1626 /*================================== Shutdown =============================== */
1627
1628 int prepareForShutdown(int flags) {
1629 int save = flags & REDIS_SHUTDOWN_SAVE;
1630 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1631
1632 redisLog(REDIS_WARNING,"User requested shutdown...");
1633 /* Kill the saving child if there is a background saving in progress.
1634 We want to avoid race conditions, for instance our saving child may
1635 overwrite the synchronous saving did by SHUTDOWN. */
1636 if (server.rdb_child_pid != -1) {
1637 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1638 kill(server.rdb_child_pid,SIGKILL);
1639 rdbRemoveTempFile(server.rdb_child_pid);
1640 }
1641 if (server.aof_state != REDIS_AOF_OFF) {
1642 /* Kill the AOF saving child as the AOF we already have may be longer
1643 * but contains the full dataset anyway. */
1644 if (server.aof_child_pid != -1) {
1645 redisLog(REDIS_WARNING,
1646 "There is a child rewriting the AOF. Killing it!");
1647 kill(server.aof_child_pid,SIGKILL);
1648 }
1649 /* Append only file: fsync() the AOF and exit */
1650 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1651 aof_fsync(server.aof_fd);
1652 }
1653 if ((server.saveparamslen > 0 && !nosave) || save) {
1654 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1655 /* Snapshotting. Perform a SYNC SAVE and exit */
1656 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1657 /* Ooops.. error saving! The best we can do is to continue
1658 * operating. Note that if there was a background saving process,
1659 * in the next cron() Redis will be notified that the background
1660 * saving aborted, handling special stuff like slaves pending for
1661 * synchronization... */
1662 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1663 return REDIS_ERR;
1664 }
1665 }
1666 if (server.daemonize) {
1667 redisLog(REDIS_NOTICE,"Removing the pid file.");
1668 unlink(server.pidfile);
1669 }
1670 /* Close the listening sockets. Apparently this allows faster restarts. */
1671 if (server.ipfd != -1) close(server.ipfd);
1672 if (server.sofd != -1) close(server.sofd);
1673 if (server.unixsocket) {
1674 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1675 unlink(server.unixsocket); /* don't care if this fails */
1676 }
1677
1678 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1679 return REDIS_OK;
1680 }
1681
1682 /*================================== Commands =============================== */
1683
1684 void authCommand(redisClient *c) {
1685 if (!server.requirepass) {
1686 addReplyError(c,"Client sent AUTH, but no password is set");
1687 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1688 c->authenticated = 1;
1689 addReply(c,shared.ok);
1690 } else {
1691 c->authenticated = 0;
1692 addReplyError(c,"invalid password");
1693 }
1694 }
1695
1696 void pingCommand(redisClient *c) {
1697 addReply(c,shared.pong);
1698 }
1699
1700 void echoCommand(redisClient *c) {
1701 addReplyBulk(c,c->argv[1]);
1702 }
1703
1704 void timeCommand(redisClient *c) {
1705 struct timeval tv;
1706
1707 /* gettimeofday() can only fail if &tv is a bad addresss so we
1708 * don't check for errors. */
1709 gettimeofday(&tv,NULL);
1710 addReplyMultiBulkLen(c,2);
1711 addReplyBulkLongLong(c,tv.tv_sec);
1712 addReplyBulkLongLong(c,tv.tv_usec);
1713 }
1714
1715 /* Convert an amount of bytes into a human readable string in the form
1716 * of 100B, 2G, 100M, 4K, and so forth. */
1717 void bytesToHuman(char *s, unsigned long long n) {
1718 double d;
1719
1720 if (n < 1024) {
1721 /* Bytes */
1722 sprintf(s,"%lluB",n);
1723 return;
1724 } else if (n < (1024*1024)) {
1725 d = (double)n/(1024);
1726 sprintf(s,"%.2fK",d);
1727 } else if (n < (1024LL*1024*1024)) {
1728 d = (double)n/(1024*1024);
1729 sprintf(s,"%.2fM",d);
1730 } else if (n < (1024LL*1024*1024*1024)) {
1731 d = (double)n/(1024LL*1024*1024);
1732 sprintf(s,"%.2fG",d);
1733 }
1734 }
1735
1736 /* Create the string returned by the INFO command. This is decoupled
1737 * by the INFO command itself as we need to report the same information
1738 * on memory corruption problems. */
1739 sds genRedisInfoString(char *section) {
1740 sds info = sdsempty();
1741 time_t uptime = server.unixtime-server.stat_starttime;
1742 int j, numcommands;
1743 struct rusage self_ru, c_ru;
1744 unsigned long lol, bib;
1745 int allsections = 0, defsections = 0;
1746 int sections = 0;
1747
1748 if (section) {
1749 allsections = strcasecmp(section,"all") == 0;
1750 defsections = strcasecmp(section,"default") == 0;
1751 }
1752
1753 getrusage(RUSAGE_SELF, &self_ru);
1754 getrusage(RUSAGE_CHILDREN, &c_ru);
1755 getClientsMaxBuffers(&lol,&bib);
1756
1757 /* Server */
1758 if (allsections || defsections || !strcasecmp(section,"server")) {
1759 struct utsname name;
1760
1761 if (sections++) info = sdscat(info,"\r\n");
1762 uname(&name);
1763 info = sdscatprintf(info,
1764 "# Server\r\n"
1765 "redis_version:%s\r\n"
1766 "redis_git_sha1:%s\r\n"
1767 "redis_git_dirty:%d\r\n"
1768 "os:%s %s %s\r\n"
1769 "arch_bits:%d\r\n"
1770 "multiplexing_api:%s\r\n"
1771 "gcc_version:%d.%d.%d\r\n"
1772 "process_id:%ld\r\n"
1773 "run_id:%s\r\n"
1774 "tcp_port:%d\r\n"
1775 "uptime_in_seconds:%ld\r\n"
1776 "uptime_in_days:%ld\r\n"
1777 "lru_clock:%ld\r\n",
1778 REDIS_VERSION,
1779 redisGitSHA1(),
1780 strtol(redisGitDirty(),NULL,10) > 0,
1781 name.sysname, name.release, name.machine,
1782 server.arch_bits,
1783 aeGetApiName(),
1784 #ifdef __GNUC__
1785 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1786 #else
1787 0,0,0,
1788 #endif
1789 (long) getpid(),
1790 server.runid,
1791 server.port,
1792 uptime,
1793 uptime/(3600*24),
1794 (unsigned long) server.lruclock);
1795 }
1796
1797 /* Clients */
1798 if (allsections || defsections || !strcasecmp(section,"clients")) {
1799 if (sections++) info = sdscat(info,"\r\n");
1800 info = sdscatprintf(info,
1801 "# Clients\r\n"
1802 "connected_clients:%lu\r\n"
1803 "client_longest_output_list:%lu\r\n"
1804 "client_biggest_input_buf:%lu\r\n"
1805 "blocked_clients:%d\r\n",
1806 listLength(server.clients)-listLength(server.slaves),
1807 lol, bib,
1808 server.bpop_blocked_clients);
1809 }
1810
1811 /* Memory */
1812 if (allsections || defsections || !strcasecmp(section,"memory")) {
1813 char hmem[64];
1814 char peak_hmem[64];
1815
1816 bytesToHuman(hmem,zmalloc_used_memory());
1817 bytesToHuman(peak_hmem,server.stat_peak_memory);
1818 if (sections++) info = sdscat(info,"\r\n");
1819 info = sdscatprintf(info,
1820 "# Memory\r\n"
1821 "used_memory:%zu\r\n"
1822 "used_memory_human:%s\r\n"
1823 "used_memory_rss:%zu\r\n"
1824 "used_memory_peak:%zu\r\n"
1825 "used_memory_peak_human:%s\r\n"
1826 "used_memory_lua:%lld\r\n"
1827 "mem_fragmentation_ratio:%.2f\r\n"
1828 "mem_allocator:%s\r\n",
1829 zmalloc_used_memory(),
1830 hmem,
1831 zmalloc_get_rss(),
1832 server.stat_peak_memory,
1833 peak_hmem,
1834 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1835 zmalloc_get_fragmentation_ratio(),
1836 ZMALLOC_LIB
1837 );
1838 }
1839
1840 /* Persistence */
1841 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1842 if (sections++) info = sdscat(info,"\r\n");
1843 info = sdscatprintf(info,
1844 "# Persistence\r\n"
1845 "loading:%d\r\n"
1846 "aof_enabled:%d\r\n"
1847 "changes_since_last_save:%lld\r\n"
1848 "bgsave_in_progress:%d\r\n"
1849 "last_save_time:%ld\r\n"
1850 "last_bgsave_status:%s\r\n"
1851 "bgrewriteaof_in_progress:%d\r\n"
1852 "bgrewriteaof_scheduled:%d\r\n",
1853 server.loading,
1854 server.aof_state != REDIS_AOF_OFF,
1855 server.dirty,
1856 server.rdb_child_pid != -1,
1857 server.lastsave,
1858 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1859 server.aof_child_pid != -1,
1860 server.aof_rewrite_scheduled);
1861
1862 if (server.aof_state != REDIS_AOF_OFF) {
1863 info = sdscatprintf(info,
1864 "aof_current_size:%lld\r\n"
1865 "aof_base_size:%lld\r\n"
1866 "aof_pending_rewrite:%d\r\n"
1867 "aof_buffer_length:%zu\r\n"
1868 "aof_pending_bio_fsync:%llu\r\n"
1869 "aof_delayed_fsync:%lu\r\n",
1870 (long long) server.aof_current_size,
1871 (long long) server.aof_rewrite_base_size,
1872 server.aof_rewrite_scheduled,
1873 sdslen(server.aof_buf),
1874 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1875 server.aof_delayed_fsync);
1876 }
1877
1878 if (server.loading) {
1879 double perc;
1880 time_t eta, elapsed;
1881 off_t remaining_bytes = server.loading_total_bytes-
1882 server.loading_loaded_bytes;
1883
1884 perc = ((double)server.loading_loaded_bytes /
1885 server.loading_total_bytes) * 100;
1886
1887 elapsed = server.unixtime-server.loading_start_time;
1888 if (elapsed == 0) {
1889 eta = 1; /* A fake 1 second figure if we don't have
1890 enough info */
1891 } else {
1892 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1893 }
1894
1895 info = sdscatprintf(info,
1896 "loading_start_time:%ld\r\n"
1897 "loading_total_bytes:%llu\r\n"
1898 "loading_loaded_bytes:%llu\r\n"
1899 "loading_loaded_perc:%.2f\r\n"
1900 "loading_eta_seconds:%ld\r\n"
1901 ,(unsigned long) server.loading_start_time,
1902 (unsigned long long) server.loading_total_bytes,
1903 (unsigned long long) server.loading_loaded_bytes,
1904 perc,
1905 eta
1906 );
1907 }
1908 }
1909
1910 /* Stats */
1911 if (allsections || defsections || !strcasecmp(section,"stats")) {
1912 if (sections++) info = sdscat(info,"\r\n");
1913 info = sdscatprintf(info,
1914 "# Stats\r\n"
1915 "total_connections_received:%lld\r\n"
1916 "total_commands_processed:%lld\r\n"
1917 "instantaneous_ops_per_sec:%lld\r\n"
1918 "rejected_connections:%lld\r\n"
1919 "expired_keys:%lld\r\n"
1920 "evicted_keys:%lld\r\n"
1921 "keyspace_hits:%lld\r\n"
1922 "keyspace_misses:%lld\r\n"
1923 "pubsub_channels:%ld\r\n"
1924 "pubsub_patterns:%lu\r\n"
1925 "latest_fork_usec:%lld\r\n",
1926 server.stat_numconnections,
1927 server.stat_numcommands,
1928 getOperationsPerSecond(),
1929 server.stat_rejected_conn,
1930 server.stat_expiredkeys,
1931 server.stat_evictedkeys,
1932 server.stat_keyspace_hits,
1933 server.stat_keyspace_misses,
1934 dictSize(server.pubsub_channels),
1935 listLength(server.pubsub_patterns),
1936 server.stat_fork_time);
1937 }
1938
1939 /* Replication */
1940 if (allsections || defsections || !strcasecmp(section,"replication")) {
1941 if (sections++) info = sdscat(info,"\r\n");
1942 info = sdscatprintf(info,
1943 "# Replication\r\n"
1944 "role:%s\r\n",
1945 server.masterhost == NULL ? "master" : "slave");
1946 if (server.masterhost) {
1947 info = sdscatprintf(info,
1948 "master_host:%s\r\n"
1949 "master_port:%d\r\n"
1950 "master_link_status:%s\r\n"
1951 "master_last_io_seconds_ago:%d\r\n"
1952 "master_sync_in_progress:%d\r\n"
1953 ,server.masterhost,
1954 server.masterport,
1955 (server.repl_state == REDIS_REPL_CONNECTED) ?
1956 "up" : "down",
1957 server.master ?
1958 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
1959 server.repl_state == REDIS_REPL_TRANSFER
1960 );
1961
1962 if (server.repl_state == REDIS_REPL_TRANSFER) {
1963 info = sdscatprintf(info,
1964 "master_sync_left_bytes:%ld\r\n"
1965 "master_sync_last_io_seconds_ago:%d\r\n"
1966 ,(long)server.repl_transfer_left,
1967 (int)(server.unixtime-server.repl_transfer_lastio)
1968 );
1969 }
1970
1971 if (server.repl_state != REDIS_REPL_CONNECTED) {
1972 info = sdscatprintf(info,
1973 "master_link_down_since_seconds:%ld\r\n",
1974 (long)server.unixtime-server.repl_down_since);
1975 }
1976 }
1977 info = sdscatprintf(info,
1978 "connected_slaves:%lu\r\n",
1979 listLength(server.slaves));
1980 if (listLength(server.slaves)) {
1981 int slaveid = 0;
1982 listNode *ln;
1983 listIter li;
1984
1985 listRewind(server.slaves,&li);
1986 while((ln = listNext(&li))) {
1987 redisClient *slave = listNodeValue(ln);
1988 char *state = NULL;
1989 char ip[32];
1990 int port;
1991
1992 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
1993 switch(slave->replstate) {
1994 case REDIS_REPL_WAIT_BGSAVE_START:
1995 case REDIS_REPL_WAIT_BGSAVE_END:
1996 state = "wait_bgsave";
1997 break;
1998 case REDIS_REPL_SEND_BULK:
1999 state = "send_bulk";
2000 break;
2001 case REDIS_REPL_ONLINE:
2002 state = "online";
2003 break;
2004 }
2005 if (state == NULL) continue;
2006 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2007 slaveid,ip,port,state);
2008 slaveid++;
2009 }
2010 }
2011 }
2012
2013 /* CPU */
2014 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2015 if (sections++) info = sdscat(info,"\r\n");
2016 info = sdscatprintf(info,
2017 "# CPU\r\n"
2018 "used_cpu_sys:%.2f\r\n"
2019 "used_cpu_user:%.2f\r\n"
2020 "used_cpu_sys_children:%.2f\r\n"
2021 "used_cpu_user_children:%.2f\r\n",
2022 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2023 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2024 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2025 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2026 }
2027
2028 /* cmdtime */
2029 if (allsections || !strcasecmp(section,"commandstats")) {
2030 if (sections++) info = sdscat(info,"\r\n");
2031 info = sdscatprintf(info, "# Commandstats\r\n");
2032 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2033 for (j = 0; j < numcommands; j++) {
2034 struct redisCommand *c = redisCommandTable+j;
2035
2036 if (!c->calls) continue;
2037 info = sdscatprintf(info,
2038 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2039 c->name, c->calls, c->microseconds,
2040 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2041 }
2042 }
2043
2044 /* Cluster */
2045 if (allsections || defsections || !strcasecmp(section,"cluster")) {
2046 if (sections++) info = sdscat(info,"\r\n");
2047 info = sdscatprintf(info,
2048 "# Cluster\r\n"
2049 "cluster_enabled:%d\r\n",
2050 server.cluster_enabled);
2051 }
2052
2053 /* Key space */
2054 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2055 if (sections++) info = sdscat(info,"\r\n");
2056 info = sdscatprintf(info, "# Keyspace\r\n");
2057 for (j = 0; j < server.dbnum; j++) {
2058 long long keys, vkeys;
2059
2060 keys = dictSize(server.db[j].dict);
2061 vkeys = dictSize(server.db[j].expires);
2062 if (keys || vkeys) {
2063 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2064 j, keys, vkeys);
2065 }
2066 }
2067 }
2068 return info;
2069 }
2070
2071 void infoCommand(redisClient *c) {
2072 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2073
2074 if (c->argc > 2) {
2075 addReply(c,shared.syntaxerr);
2076 return;
2077 }
2078 sds info = genRedisInfoString(section);
2079 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2080 (unsigned long)sdslen(info)));
2081 addReplySds(c,info);
2082 addReply(c,shared.crlf);
2083 }
2084
2085 void monitorCommand(redisClient *c) {
2086 /* ignore MONITOR if aleady slave or in monitor mode */
2087 if (c->flags & REDIS_SLAVE) return;
2088
2089 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2090 c->slaveseldb = 0;
2091 listAddNodeTail(server.monitors,c);
2092 addReply(c,shared.ok);
2093 }
2094
2095 /* ============================ Maxmemory directive ======================== */
2096
2097 /* This function gets called when 'maxmemory' is set on the config file to limit
2098 * the max memory used by the server, before processing a command.
2099 *
2100 * The goal of the function is to free enough memory to keep Redis under the
2101 * configured memory limit.
2102 *
2103 * The function starts calculating how many bytes should be freed to keep
2104 * Redis under the limit, and enters a loop selecting the best keys to
2105 * evict accordingly to the configured policy.
2106 *
2107 * If all the bytes needed to return back under the limit were freed the
2108 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2109 * should block the execution of commands that will result in more memory
2110 * used by the server.
2111 */
2112 int freeMemoryIfNeeded(void) {
2113 size_t mem_used, mem_tofree, mem_freed;
2114 int slaves = listLength(server.slaves);
2115
2116 /* Remove the size of slaves output buffers and AOF buffer from the
2117 * count of used memory. */
2118 mem_used = zmalloc_used_memory();
2119 if (slaves) {
2120 listIter li;
2121 listNode *ln;
2122
2123 listRewind(server.slaves,&li);
2124 while((ln = listNext(&li))) {
2125 redisClient *slave = listNodeValue(ln);
2126 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2127 if (obuf_bytes > mem_used)
2128 mem_used = 0;
2129 else
2130 mem_used -= obuf_bytes;
2131 }
2132 }
2133 if (server.aof_state != REDIS_AOF_OFF) {
2134 mem_used -= sdslen(server.aof_buf);
2135 mem_used -= sdslen(server.aof_rewrite_buf);
2136 }
2137
2138 /* Check if we are over the memory limit. */
2139 if (mem_used <= server.maxmemory) return REDIS_OK;
2140
2141 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2142 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2143
2144 /* Compute how much memory we need to free. */
2145 mem_tofree = mem_used - server.maxmemory;
2146 mem_freed = 0;
2147 while (mem_freed < mem_tofree) {
2148 int j, k, keys_freed = 0;
2149
2150 for (j = 0; j < server.dbnum; j++) {
2151 long bestval = 0; /* just to prevent warning */
2152 sds bestkey = NULL;
2153 struct dictEntry *de;
2154 redisDb *db = server.db+j;
2155 dict *dict;
2156
2157 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2158 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2159 {
2160 dict = server.db[j].dict;
2161 } else {
2162 dict = server.db[j].expires;
2163 }
2164 if (dictSize(dict) == 0) continue;
2165
2166 /* volatile-random and allkeys-random policy */
2167 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2168 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2169 {
2170 de = dictGetRandomKey(dict);
2171 bestkey = dictGetKey(de);
2172 }
2173
2174 /* volatile-lru and allkeys-lru policy */
2175 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2176 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2177 {
2178 for (k = 0; k < server.maxmemory_samples; k++) {
2179 sds thiskey;
2180 long thisval;
2181 robj *o;
2182
2183 de = dictGetRandomKey(dict);
2184 thiskey = dictGetKey(de);
2185 /* When policy is volatile-lru we need an additonal lookup
2186 * to locate the real key, as dict is set to db->expires. */
2187 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2188 de = dictFind(db->dict, thiskey);
2189 o = dictGetVal(de);
2190 thisval = estimateObjectIdleTime(o);
2191
2192 /* Higher idle time is better candidate for deletion */
2193 if (bestkey == NULL || thisval > bestval) {
2194 bestkey = thiskey;
2195 bestval = thisval;
2196 }
2197 }
2198 }
2199
2200 /* volatile-ttl */
2201 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2202 for (k = 0; k < server.maxmemory_samples; k++) {
2203 sds thiskey;
2204 long thisval;
2205
2206 de = dictGetRandomKey(dict);
2207 thiskey = dictGetKey(de);
2208 thisval = (long) dictGetVal(de);
2209
2210 /* Expire sooner (minor expire unix timestamp) is better
2211 * candidate for deletion */
2212 if (bestkey == NULL || thisval < bestval) {
2213 bestkey = thiskey;
2214 bestval = thisval;
2215 }
2216 }
2217 }
2218
2219 /* Finally remove the selected key. */
2220 if (bestkey) {
2221 long long delta;
2222
2223 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2224 propagateExpire(db,keyobj);
2225 /* We compute the amount of memory freed by dbDelete() alone.
2226 * It is possible that actually the memory needed to propagate
2227 * the DEL in AOF and replication link is greater than the one
2228 * we are freeing removing the key, but we can't account for
2229 * that otherwise we would never exit the loop.
2230 *
2231 * AOF and Output buffer memory will be freed eventually so
2232 * we only care about memory used by the key space. */
2233 delta = (long long) zmalloc_used_memory();
2234 dbDelete(db,keyobj);
2235 delta -= (long long) zmalloc_used_memory();
2236 mem_freed += delta;
2237 server.stat_evictedkeys++;
2238 decrRefCount(keyobj);
2239 keys_freed++;
2240
2241 /* When the memory to free starts to be big enough, we may
2242 * start spending so much time here that is impossible to
2243 * deliver data to the slaves fast enough, so we force the
2244 * transmission here inside the loop. */
2245 if (slaves) flushSlavesOutputBuffers();
2246 }
2247 }
2248 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2249 }
2250 return REDIS_OK;
2251 }
2252
2253 /* =================================== Main! ================================ */
2254
2255 #ifdef __linux__
2256 int linuxOvercommitMemoryValue(void) {
2257 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2258 char buf[64];
2259
2260 if (!fp) return -1;
2261 if (fgets(buf,64,fp) == NULL) {
2262 fclose(fp);
2263 return -1;
2264 }
2265 fclose(fp);
2266
2267 return atoi(buf);
2268 }
2269
2270 void linuxOvercommitMemoryWarning(void) {
2271 if (linuxOvercommitMemoryValue() == 0) {
2272 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.");
2273 }
2274 }
2275 #endif /* __linux__ */
2276
2277 void createPidFile(void) {
2278 /* Try to write the pid file in a best-effort way. */
2279 FILE *fp = fopen(server.pidfile,"w");
2280 if (fp) {
2281 fprintf(fp,"%d\n",(int)getpid());
2282 fclose(fp);
2283 }
2284 }
2285
2286 void daemonize(void) {
2287 int fd;
2288
2289 if (fork() != 0) exit(0); /* parent exits */
2290 setsid(); /* create a new session */
2291
2292 /* Every output goes to /dev/null. If Redis is daemonized but
2293 * the 'logfile' is set to 'stdout' in the configuration file
2294 * it will not log at all. */
2295 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2296 dup2(fd, STDIN_FILENO);
2297 dup2(fd, STDOUT_FILENO);
2298 dup2(fd, STDERR_FILENO);
2299 if (fd > STDERR_FILENO) close(fd);
2300 }
2301 }
2302
2303 void version() {
2304 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2305 REDIS_VERSION,
2306 redisGitSHA1(),
2307 atoi(redisGitDirty()) > 0,
2308 ZMALLOC_LIB,
2309 sizeof(long) == 4 ? 32 : 64);
2310 exit(0);
2311 }
2312
2313 void usage() {
2314 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2315 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2316 fprintf(stderr," ./redis-server -v or --version\n");
2317 fprintf(stderr," ./redis-server -h or --help\n");
2318 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2319 fprintf(stderr,"Examples:\n");
2320 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2321 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2322 fprintf(stderr," ./redis-server --port 7777\n");
2323 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2324 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2325 exit(1);
2326 }
2327
2328 void redisAsciiArt(void) {
2329 #include "asciilogo.h"
2330 char *buf = zmalloc(1024*16);
2331
2332 snprintf(buf,1024*16,ascii_logo,
2333 REDIS_VERSION,
2334 redisGitSHA1(),
2335 strtol(redisGitDirty(),NULL,10) > 0,
2336 (sizeof(long) == 8) ? "64" : "32",
2337 server.cluster_enabled ? "cluster" : "stand alone",
2338 server.port,
2339 (long) getpid()
2340 );
2341 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2342 zfree(buf);
2343 }
2344
2345 static void sigtermHandler(int sig) {
2346 REDIS_NOTUSED(sig);
2347
2348 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2349 server.shutdown_asap = 1;
2350 }
2351
2352 void setupSignalHandlers(void) {
2353 struct sigaction act;
2354 stack_t stack;
2355
2356 stack.ss_sp = altstack;
2357 stack.ss_flags = 0;
2358 stack.ss_size = SIGSTKSZ;
2359
2360 sigaltstack(&stack, NULL);
2361
2362 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2363 * Otherwise, sa_handler is used. */
2364 sigemptyset(&act.sa_mask);
2365 act.sa_flags = 0;
2366 act.sa_handler = sigtermHandler;
2367 sigaction(SIGTERM, &act, NULL);
2368
2369 #ifdef HAVE_BACKTRACE
2370 /* Use alternate stack so we don't clobber stack in case of segv, or when we run out of stack ..
2371 * also resethand & nodefer so we can get interrupted (and killed) if we cause SEGV during SEGV handler */
2372 sigemptyset(&act.sa_mask);
2373 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
2374 act.sa_sigaction = sigsegvHandler;
2375 sigaction(SIGSEGV, &act, NULL);
2376 sigaction(SIGBUS, &act, NULL);
2377 sigaction(SIGFPE, &act, NULL);
2378 sigaction(SIGILL, &act, NULL);
2379 #endif
2380 return;
2381 }
2382
2383 void memtest(size_t megabytes, int passes);
2384
2385 int main(int argc, char **argv) {
2386 long long start;
2387 struct timeval tv;
2388
2389 /* We need to initialize our libraries, and the server configuration. */
2390 zmalloc_enable_thread_safeness();
2391 srand(time(NULL)^getpid());
2392 gettimeofday(&tv,NULL);
2393 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2394 initServerConfig();
2395
2396 if (argc >= 2) {
2397 int j = 1; /* First option to parse in argv[] */
2398 sds options = sdsempty();
2399 char *configfile = NULL;
2400
2401 /* Handle special options --help and --version */
2402 if (strcmp(argv[1], "-v") == 0 ||
2403 strcmp(argv[1], "--version") == 0) version();
2404 if (strcmp(argv[1], "--help") == 0 ||
2405 strcmp(argv[1], "-h") == 0) usage();
2406 if (strcmp(argv[1], "--test-memory") == 0) {
2407 if (argc == 3) {
2408 memtest(atoi(argv[2]),50);
2409 exit(0);
2410 } else {
2411 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2412 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2413 exit(1);
2414 }
2415 }
2416
2417 /* First argument is the config file name? */
2418 if (argv[j][0] != '-' || argv[j][1] != '-')
2419 configfile = argv[j++];
2420 /* All the other options are parsed and conceptually appended to the
2421 * configuration file. For instance --port 6380 will generate the
2422 * string "port 6380\n" to be parsed after the actual file name
2423 * is parsed, if any. */
2424 while(j != argc) {
2425 if (argv[j][0] == '-' && argv[j][1] == '-') {
2426 /* Option name */
2427 if (sdslen(options)) options = sdscat(options,"\n");
2428 options = sdscat(options,argv[j]+2);
2429 options = sdscat(options," ");
2430 } else {
2431 /* Option argument */
2432 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2433 options = sdscat(options," ");
2434 }
2435 j++;
2436 }
2437 resetServerSaveParams();
2438 loadServerConfig(configfile,options);
2439 sdsfree(options);
2440 } else {
2441 redisLog(REDIS_WARNING,"Warning: no config file specified, using the default config. In order to specify a config file use 'redis-server /path/to/redis.conf'");
2442 }
2443 if (server.daemonize) daemonize();
2444 initServer();
2445 if (server.daemonize) createPidFile();
2446 redisAsciiArt();
2447 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2448 #ifdef __linux__
2449 linuxOvercommitMemoryWarning();
2450 #endif
2451 start = ustime();
2452 if (server.aof_state == REDIS_AOF_ON) {
2453 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2454 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2455 } else {
2456 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2457 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2458 (float)(ustime()-start)/1000000);
2459 } else if (errno != ENOENT) {
2460 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2461 exit(1);
2462 }
2463 }
2464 if (server.ipfd > 0)
2465 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2466 if (server.sofd > 0)
2467 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2468 aeSetBeforeSleepProc(server.el,beforeSleep);
2469 aeMain(server.el);
2470 aeDeleteEventLoop(server.el);
2471 return 0;
2472 }
2473
2474 /* The End */