]> git.saurik.com Git - redis.git/blame_incremental - src/redis.c
Redis 2.5.11 (2.6 RC5).
[redis.git] / src / redis.c
... / ...
CommitLineData
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
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 */
110struct redisCommand redisCommandTable[] = {
111 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
112 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
113 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
114 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
115 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
116 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
117 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
118 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
119 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
120 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
121 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
122 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
123 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
124 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
125 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
126 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
127 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
128 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
129 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
130 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
131 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
132 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
133 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
134 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
135 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
136 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
137 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
138 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
139 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
140 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
141 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
142 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
143 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
144 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
145 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
146 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
147 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
148 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
149 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
150 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
151 {"srandmember",srandmemberCommand,2,"rR",0,NULL,1,1,1,0,0},
152 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
153 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
154 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
155 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
156 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
157 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
158 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
159 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
160 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
161 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
162 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
163 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
164 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
165 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
166 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
167 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
168 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
169 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
170 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
171 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
172 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
173 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
174 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
175 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
176 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
177 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
178 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
179 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
180 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
182 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
183 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
184 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
185 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
186 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
187 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
188 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
189 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
190 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
191 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
192 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
193 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
194 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
195 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
196 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
197 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
198 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
199 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
200 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
201 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
202 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
203 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
204 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
205 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
206 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
207 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
208 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
209 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
210 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
211 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
212 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
213 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
214 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
215 {"exec",execCommand,1,"s",0,NULL,0,0,0,0,0},
216 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
217 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
218 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
219 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
220 {"sort",sortCommand,-2,"wmS",0,NULL,1,1,1,0,0},
221 {"info",infoCommand,-1,"r",0,NULL,0,0,0,0,0},
222 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
223 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
224 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
225 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
226 {"slaveof",slaveofCommand,3,"as",0,NULL,0,0,0,0,0},
227 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
228 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
229 {"subscribe",subscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
230 {"unsubscribe",unsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
231 {"psubscribe",psubscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
232 {"punsubscribe",punsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
233 {"publish",publishCommand,3,"pf",0,NULL,0,0,0,0,0},
234 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
235 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
236 {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
237 {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
238 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
239 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
240 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
241 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
242 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
243 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
244 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
245 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0},
246 {"bitop",bitopCommand,-4,"wm",0,NULL,2,-1,1,0,0},
247 {"bitcount",bitcountCommand,-2,"r",0,NULL,1,1,1,0,0}
248};
249
250/*============================ Utility functions ============================ */
251
252/* Low level logging. To use only for very big messages, otherwise
253 * redisLog() is to prefer. */
254void redisLogRaw(int level, const char *msg) {
255 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
256 const char *c = ".-*#";
257 FILE *fp;
258 char buf[64];
259 int rawmode = (level & REDIS_LOG_RAW);
260
261 level &= 0xff; /* clear flags */
262 if (level < server.verbosity) return;
263
264 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
265 if (!fp) return;
266
267 if (rawmode) {
268 fprintf(fp,"%s",msg);
269 } else {
270 int off;
271 struct timeval tv;
272
273 gettimeofday(&tv,NULL);
274 off = strftime(buf,sizeof(buf),"%d %b %H:%M:%S.",localtime(&tv.tv_sec));
275 snprintf(buf+off,sizeof(buf)-off,"%03d",(int)tv.tv_usec/1000);
276 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
277 }
278 fflush(fp);
279
280 if (server.logfile) fclose(fp);
281
282 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
283}
284
285/* Like redisLogRaw() but with printf-alike support. This is the funciton that
286 * is used across the code. The raw version is only used in order to dump
287 * the INFO output on crash. */
288void redisLog(int level, const char *fmt, ...) {
289 va_list ap;
290 char msg[REDIS_MAX_LOGMSG_LEN];
291
292 if ((level&0xff) < server.verbosity) return;
293
294 va_start(ap, fmt);
295 vsnprintf(msg, sizeof(msg), fmt, ap);
296 va_end(ap);
297
298 redisLogRaw(level,msg);
299}
300
301/* Log a fixed message without printf-alike capabilities, in a way that is
302 * safe to call from a signal handler.
303 *
304 * We actually use this only for signals that are not fatal from the point
305 * of view of Redis. Signals that are going to kill the server anyway and
306 * where we need printf-alike features are served by redisLog(). */
307void redisLogFromHandler(int level, const char *msg) {
308 int fd;
309 char buf[64];
310
311 if ((level&0xff) < server.verbosity ||
312 (server.logfile == NULL && server.daemonize)) return;
313 fd = server.logfile ?
314 open(server.logfile, O_APPEND|O_CREAT|O_WRONLY, 0644) :
315 STDOUT_FILENO;
316 if (fd == -1) return;
317 ll2string(buf,sizeof(buf),getpid());
318 if (write(fd,"[",1) == -1) goto err;
319 if (write(fd,buf,strlen(buf)) == -1) goto err;
320 if (write(fd," | signal handler] (",20) == -1) goto err;
321 ll2string(buf,sizeof(buf),time(NULL));
322 if (write(fd,buf,strlen(buf)) == -1) goto err;
323 if (write(fd,") ",2) == -1) goto err;
324 if (write(fd,msg,strlen(msg)) == -1) goto err;
325 if (write(fd,"\n",1) == -1) goto err;
326err:
327 if (server.logfile) close(fd);
328}
329
330/* Redis generally does not try to recover from out of memory conditions
331 * when allocating objects or strings, it is not clear if it will be possible
332 * to report this condition to the client since the networking layer itself
333 * is based on heap allocation for send buffers, so we simply abort.
334 * At least the code will be simpler to read... */
335void oom(const char *msg) {
336 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
337 sleep(1);
338 abort();
339}
340
341/* Return the UNIX time in microseconds */
342long long ustime(void) {
343 struct timeval tv;
344 long long ust;
345
346 gettimeofday(&tv, NULL);
347 ust = ((long long)tv.tv_sec)*1000000;
348 ust += tv.tv_usec;
349 return ust;
350}
351
352/* Return the UNIX time in milliseconds */
353long long mstime(void) {
354 return ustime()/1000;
355}
356
357/* After an RDB dump or AOF rewrite we exit from children using _exit() instead of
358 * exit(), because the latter may interact with the same file objects used by
359 * the parent process. However if we are testing the coverage normal exit() is
360 * used in order to obtain the right coverage information. */
361void exitFromChild(int retcode) {
362#ifdef COVERAGE_TEST
363 exit(retcode);
364#else
365 _exit(retcode);
366#endif
367}
368
369/*====================== Hash table type implementation ==================== */
370
371/* This is an hash table type that uses the SDS dynamic strings libary as
372 * keys and radis objects as values (objects can hold SDS strings,
373 * lists, sets). */
374
375void dictVanillaFree(void *privdata, void *val)
376{
377 DICT_NOTUSED(privdata);
378 zfree(val);
379}
380
381void dictListDestructor(void *privdata, void *val)
382{
383 DICT_NOTUSED(privdata);
384 listRelease((list*)val);
385}
386
387int dictSdsKeyCompare(void *privdata, const void *key1,
388 const void *key2)
389{
390 int l1,l2;
391 DICT_NOTUSED(privdata);
392
393 l1 = sdslen((sds)key1);
394 l2 = sdslen((sds)key2);
395 if (l1 != l2) return 0;
396 return memcmp(key1, key2, l1) == 0;
397}
398
399/* A case insensitive version used for the command lookup table. */
400int dictSdsKeyCaseCompare(void *privdata, const void *key1,
401 const void *key2)
402{
403 DICT_NOTUSED(privdata);
404
405 return strcasecmp(key1, key2) == 0;
406}
407
408void dictRedisObjectDestructor(void *privdata, void *val)
409{
410 DICT_NOTUSED(privdata);
411
412 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
413 decrRefCount(val);
414}
415
416void dictSdsDestructor(void *privdata, void *val)
417{
418 DICT_NOTUSED(privdata);
419
420 sdsfree(val);
421}
422
423int dictObjKeyCompare(void *privdata, const void *key1,
424 const void *key2)
425{
426 const robj *o1 = key1, *o2 = key2;
427 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
428}
429
430unsigned int dictObjHash(const void *key) {
431 const robj *o = key;
432 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
433}
434
435unsigned int dictSdsHash(const void *key) {
436 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
437}
438
439unsigned int dictSdsCaseHash(const void *key) {
440 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
441}
442
443int dictEncObjKeyCompare(void *privdata, const void *key1,
444 const void *key2)
445{
446 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
447 int cmp;
448
449 if (o1->encoding == REDIS_ENCODING_INT &&
450 o2->encoding == REDIS_ENCODING_INT)
451 return o1->ptr == o2->ptr;
452
453 o1 = getDecodedObject(o1);
454 o2 = getDecodedObject(o2);
455 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
456 decrRefCount(o1);
457 decrRefCount(o2);
458 return cmp;
459}
460
461unsigned int dictEncObjHash(const void *key) {
462 robj *o = (robj*) key;
463
464 if (o->encoding == REDIS_ENCODING_RAW) {
465 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
466 } else {
467 if (o->encoding == REDIS_ENCODING_INT) {
468 char buf[32];
469 int len;
470
471 len = ll2string(buf,32,(long)o->ptr);
472 return dictGenHashFunction((unsigned char*)buf, len);
473 } else {
474 unsigned int hash;
475
476 o = getDecodedObject(o);
477 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
478 decrRefCount(o);
479 return hash;
480 }
481 }
482}
483
484/* Sets type hash table */
485dictType setDictType = {
486 dictEncObjHash, /* hash function */
487 NULL, /* key dup */
488 NULL, /* val dup */
489 dictEncObjKeyCompare, /* key compare */
490 dictRedisObjectDestructor, /* key destructor */
491 NULL /* val destructor */
492};
493
494/* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
495dictType zsetDictType = {
496 dictEncObjHash, /* hash function */
497 NULL, /* key dup */
498 NULL, /* val dup */
499 dictEncObjKeyCompare, /* key compare */
500 dictRedisObjectDestructor, /* key destructor */
501 NULL /* val destructor */
502};
503
504/* Db->dict, keys are sds strings, vals are Redis objects. */
505dictType dbDictType = {
506 dictSdsHash, /* hash function */
507 NULL, /* key dup */
508 NULL, /* val dup */
509 dictSdsKeyCompare, /* key compare */
510 dictSdsDestructor, /* key destructor */
511 dictRedisObjectDestructor /* val destructor */
512};
513
514/* Db->expires */
515dictType keyptrDictType = {
516 dictSdsHash, /* hash function */
517 NULL, /* key dup */
518 NULL, /* val dup */
519 dictSdsKeyCompare, /* key compare */
520 NULL, /* key destructor */
521 NULL /* val destructor */
522};
523
524/* Command table. sds string -> command struct pointer. */
525dictType commandTableDictType = {
526 dictSdsCaseHash, /* hash function */
527 NULL, /* key dup */
528 NULL, /* val dup */
529 dictSdsKeyCaseCompare, /* key compare */
530 dictSdsDestructor, /* key destructor */
531 NULL /* val destructor */
532};
533
534/* Hash type hash table (note that small hashes are represented with zimpaps) */
535dictType hashDictType = {
536 dictEncObjHash, /* hash function */
537 NULL, /* key dup */
538 NULL, /* val dup */
539 dictEncObjKeyCompare, /* key compare */
540 dictRedisObjectDestructor, /* key destructor */
541 dictRedisObjectDestructor /* val destructor */
542};
543
544/* Keylist hash table type has unencoded redis objects as keys and
545 * lists as values. It's used for blocking operations (BLPOP) and to
546 * map swapped keys to a list of clients waiting for this keys to be loaded. */
547dictType keylistDictType = {
548 dictObjHash, /* hash function */
549 NULL, /* key dup */
550 NULL, /* val dup */
551 dictObjKeyCompare, /* key compare */
552 dictRedisObjectDestructor, /* key destructor */
553 dictListDestructor /* val destructor */
554};
555
556int htNeedsResize(dict *dict) {
557 long long size, used;
558
559 size = dictSlots(dict);
560 used = dictSize(dict);
561 return (size && used && size > DICT_HT_INITIAL_SIZE &&
562 (used*100/size < REDIS_HT_MINFILL));
563}
564
565/* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
566 * we resize the hash table to save memory */
567void tryResizeHashTables(void) {
568 int j;
569
570 for (j = 0; j < server.dbnum; j++) {
571 if (htNeedsResize(server.db[j].dict))
572 dictResize(server.db[j].dict);
573 if (htNeedsResize(server.db[j].expires))
574 dictResize(server.db[j].expires);
575 }
576}
577
578/* Our hash table implementation performs rehashing incrementally while
579 * we write/read from the hash table. Still if the server is idle, the hash
580 * table will use two tables for a long time. So we try to use 1 millisecond
581 * of CPU time at every serverCron() loop in order to rehash some key. */
582void incrementallyRehash(void) {
583 int j;
584
585 for (j = 0; j < server.dbnum; j++) {
586 /* Keys dictionary */
587 if (dictIsRehashing(server.db[j].dict)) {
588 dictRehashMilliseconds(server.db[j].dict,1);
589 break; /* already used our millisecond for this loop... */
590 }
591 /* Expires */
592 if (dictIsRehashing(server.db[j].expires)) {
593 dictRehashMilliseconds(server.db[j].expires,1);
594 break; /* already used our millisecond for this loop... */
595 }
596 }
597}
598
599/* This function is called once a background process of some kind terminates,
600 * as we want to avoid resizing the hash tables when there is a child in order
601 * to play well with copy-on-write (otherwise when a resize happens lots of
602 * memory pages are copied). The goal of this function is to update the ability
603 * for dict.c to resize the hash tables accordingly to the fact we have o not
604 * running childs. */
605void updateDictResizePolicy(void) {
606 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
607 dictEnableResize();
608 else
609 dictDisableResize();
610}
611
612/* ======================= Cron: called every 100 ms ======================== */
613
614/* Try to expire a few timed out keys. The algorithm used is adaptive and
615 * will use few CPU cycles if there are few expiring keys, otherwise
616 * it will get more aggressive to avoid that too much memory is used by
617 * keys that can be removed from the keyspace. */
618void activeExpireCycle(void) {
619 int j, iteration = 0;
620 long long start = ustime(), timelimit;
621
622 /* We can use at max REDIS_EXPIRELOOKUPS_TIME_PERC percentage of CPU time
623 * per iteration. Since this function gets called with a frequency of
624 * REDIS_HZ times per second, the following is the max amount of
625 * microseconds we can spend in this function. */
626 timelimit = 1000000*REDIS_EXPIRELOOKUPS_TIME_PERC/REDIS_HZ/100;
627 if (timelimit <= 0) timelimit = 1;
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 unsigned long num = dictSize(db->expires);
637 unsigned long slots = dictSlots(db->expires);
638 long long now = mstime();
639
640 /* When there are less than 1% filled slots getting random
641 * keys is expensive, so stop here waiting for better times...
642 * The dictionary will be resized asap. */
643 if (num && slots > DICT_HT_INITIAL_SIZE &&
644 (num*100/slots < 1)) break;
645
646 /* The main collection cycle. Sample random keys among keys
647 * with an expire set, checking for expired ones. */
648 expired = 0;
649 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
650 num = REDIS_EXPIRELOOKUPS_PER_CRON;
651 while (num--) {
652 dictEntry *de;
653 long long t;
654
655 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
656 t = dictGetSignedIntegerVal(de);
657 if (now > t) {
658 sds key = dictGetKey(de);
659 robj *keyobj = createStringObject(key,sdslen(key));
660
661 propagateExpire(db,keyobj);
662 dbDelete(db,keyobj);
663 decrRefCount(keyobj);
664 expired++;
665 server.stat_expiredkeys++;
666 }
667 }
668 /* We can't block forever here even if there are many keys to
669 * expire. So after a given amount of milliseconds return to the
670 * caller waiting for the other active expire cycle. */
671 iteration++;
672 if ((iteration & 0xf) == 0 && /* check once every 16 cycles. */
673 (ustime()-start) > timelimit) return;
674 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
675 }
676}
677
678void updateLRUClock(void) {
679 server.lruclock = (server.unixtime/REDIS_LRU_CLOCK_RESOLUTION) &
680 REDIS_LRU_CLOCK_MAX;
681}
682
683
684/* Add a sample to the operations per second array of samples. */
685void trackOperationsPerSecond(void) {
686 long long t = mstime() - server.ops_sec_last_sample_time;
687 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
688 long long ops_sec;
689
690 ops_sec = t > 0 ? (ops*1000/t) : 0;
691
692 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
693 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
694 server.ops_sec_last_sample_time = mstime();
695 server.ops_sec_last_sample_ops = server.stat_numcommands;
696}
697
698/* Return the mean of all the samples. */
699long long getOperationsPerSecond(void) {
700 int j;
701 long long sum = 0;
702
703 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
704 sum += server.ops_sec_samples[j];
705 return sum / REDIS_OPS_SEC_SAMPLES;
706}
707
708/* Check for timeouts. Returns non-zero if the client was terminated */
709int clientsCronHandleTimeout(redisClient *c) {
710 time_t now = server.unixtime;
711
712 if (server.maxidletime &&
713 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
714 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
715 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
716 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
717 listLength(c->pubsub_patterns) == 0 &&
718 (now - c->lastinteraction > server.maxidletime))
719 {
720 redisLog(REDIS_VERBOSE,"Closing idle client");
721 freeClient(c);
722 return 1;
723 } else if (c->flags & REDIS_BLOCKED) {
724 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
725 addReply(c,shared.nullmultibulk);
726 unblockClientWaitingData(c);
727 }
728 }
729 return 0;
730}
731
732/* The client query buffer is an sds.c string that can end with a lot of
733 * free space not used, this function reclaims space if needed.
734 *
735 * The funciton always returns 0 as it never terminates the client. */
736int clientsCronResizeQueryBuffer(redisClient *c) {
737 size_t querybuf_size = sdsAllocSize(c->querybuf);
738 time_t idletime = server.unixtime - c->lastinteraction;
739
740 /* There are two conditions to resize the query buffer:
741 * 1) Query buffer is > BIG_ARG and too big for latest peak.
742 * 2) Client is inactive and the buffer is bigger than 1k. */
743 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
744 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
745 (querybuf_size > 1024 && idletime > 2))
746 {
747 /* Only resize the query buffer if it is actually wasting space. */
748 if (sdsavail(c->querybuf) > 1024) {
749 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
750 }
751 }
752 /* Reset the peak again to capture the peak memory usage in the next
753 * cycle. */
754 c->querybuf_peak = 0;
755 return 0;
756}
757
758void clientsCron(void) {
759 /* Make sure to process at least 1/(REDIS_HZ*10) of clients per call.
760 * Since this function is called REDIS_HZ times per second we are sure that
761 * in the worst case we process all the clients in 10 seconds.
762 * In normal conditions (a reasonable number of clients) we process
763 * all the clients in a shorter time. */
764 int numclients = listLength(server.clients);
765 int iterations = numclients/(REDIS_HZ*10);
766
767 if (iterations < 50)
768 iterations = (numclients < 50) ? numclients : 50;
769 while(listLength(server.clients) && iterations--) {
770 redisClient *c;
771 listNode *head;
772
773 /* Rotate the list, take the current head, process.
774 * This way if the client must be removed from the list it's the
775 * first element and we don't incur into O(N) computation. */
776 listRotate(server.clients);
777 head = listFirst(server.clients);
778 c = listNodeValue(head);
779 /* The following functions do different service checks on the client.
780 * The protocol is that they return non-zero if the client was
781 * terminated. */
782 if (clientsCronHandleTimeout(c)) continue;
783 if (clientsCronResizeQueryBuffer(c)) continue;
784 }
785}
786
787/* This is our timer interrupt, called REDIS_HZ times per second.
788 * Here is where we do a number of things that need to be done asynchronously.
789 * For instance:
790 *
791 * - Active expired keys collection (it is also performed in a lazy way on
792 * lookup).
793 * - Software watchdong.
794 * - Update some statistic.
795 * - Incremental rehashing of the DBs hash tables.
796 * - Triggering BGSAVE / AOF rewrite, and handling of terminated children.
797 * - Clients timeout of differnet kinds.
798 * - Replication reconnection.
799 * - Many more...
800 *
801 * Everything directly called here will be called REDIS_HZ times per second,
802 * so in order to throttle execution of things we want to do less frequently
803 * a macro is used: run_with_period(milliseconds) { .... }
804 */
805
806/* Using the following macro you can run code inside serverCron() with the
807 * specified period, specified in milliseconds.
808 * The actual resolution depends on REDIS_HZ. */
809#define run_with_period(_ms_) if (!(loops % ((_ms_)/(1000/REDIS_HZ))))
810
811int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
812 int j, loops = server.cronloops;
813 REDIS_NOTUSED(eventLoop);
814 REDIS_NOTUSED(id);
815 REDIS_NOTUSED(clientData);
816
817 /* Software watchdog: deliver the SIGALRM that will reach the signal
818 * handler if we don't return here fast enough. */
819 if (server.watchdog_period) watchdogScheduleSignal(server.watchdog_period);
820
821 /* We take a cached value of the unix time in the global state because
822 * with virtual memory and aging there is to store the current time
823 * in objects at every object access, and accuracy is not needed.
824 * To access a global var is faster than calling time(NULL) */
825 server.unixtime = time(NULL);
826
827 run_with_period(100) trackOperationsPerSecond();
828
829 /* We have just 22 bits per object for LRU information.
830 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
831 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
832 *
833 * Note that even if this will wrap after 1.5 years it's not a problem,
834 * everything will still work but just some object will appear younger
835 * to Redis. But for this to happen a given object should never be touched
836 * for 1.5 years.
837 *
838 * Note that you can change the resolution altering the
839 * REDIS_LRU_CLOCK_RESOLUTION define.
840 */
841 updateLRUClock();
842
843 /* Record the max memory used since the server was started. */
844 if (zmalloc_used_memory() > server.stat_peak_memory)
845 server.stat_peak_memory = zmalloc_used_memory();
846
847 /* We received a SIGTERM, shutting down here in a safe way, as it is
848 * not ok doing so inside the signal handler. */
849 if (server.shutdown_asap) {
850 if (prepareForShutdown(0) == REDIS_OK) exit(0);
851 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
852 }
853
854 /* Show some info about non-empty databases */
855 run_with_period(5000) {
856 for (j = 0; j < server.dbnum; j++) {
857 long long size, used, vkeys;
858
859 size = dictSlots(server.db[j].dict);
860 used = dictSize(server.db[j].dict);
861 vkeys = dictSize(server.db[j].expires);
862 if (used || vkeys) {
863 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
864 /* dictPrintStats(server.dict); */
865 }
866 }
867 }
868
869 /* We don't want to resize the hash tables while a bacground saving
870 * is in progress: the saving child is created using fork() that is
871 * implemented with a copy-on-write semantic in most modern systems, so
872 * if we resize the HT while there is the saving child at work actually
873 * a lot of memory movements in the parent will cause a lot of pages
874 * copied. */
875 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
876 tryResizeHashTables();
877 if (server.activerehashing) incrementallyRehash();
878 }
879
880 /* Show information about connected clients */
881 run_with_period(5000) {
882 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
883 listLength(server.clients)-listLength(server.slaves),
884 listLength(server.slaves),
885 zmalloc_used_memory());
886 }
887
888 /* We need to do a few operations on clients asynchronously. */
889 clientsCron();
890
891 /* Start a scheduled AOF rewrite if this was requested by the user while
892 * a BGSAVE was in progress. */
893 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
894 server.aof_rewrite_scheduled)
895 {
896 rewriteAppendOnlyFileBackground();
897 }
898
899 /* Check if a background saving or AOF rewrite in progress terminated. */
900 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
901 int statloc;
902 pid_t pid;
903
904 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
905 int exitcode = WEXITSTATUS(statloc);
906 int bysignal = 0;
907
908 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
909
910 if (pid == server.rdb_child_pid) {
911 backgroundSaveDoneHandler(exitcode,bysignal);
912 } else {
913 backgroundRewriteDoneHandler(exitcode,bysignal);
914 }
915 updateDictResizePolicy();
916 }
917 } else {
918 /* If there is not a background saving/rewrite in progress check if
919 * we have to save/rewrite now */
920 for (j = 0; j < server.saveparamslen; j++) {
921 struct saveparam *sp = server.saveparams+j;
922
923 if (server.dirty >= sp->changes &&
924 server.unixtime-server.lastsave > sp->seconds) {
925 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
926 sp->changes, sp->seconds);
927 rdbSaveBackground(server.rdb_filename);
928 break;
929 }
930 }
931
932 /* Trigger an AOF rewrite if needed */
933 if (server.rdb_child_pid == -1 &&
934 server.aof_child_pid == -1 &&
935 server.aof_rewrite_perc &&
936 server.aof_current_size > server.aof_rewrite_min_size)
937 {
938 long long base = server.aof_rewrite_base_size ?
939 server.aof_rewrite_base_size : 1;
940 long long growth = (server.aof_current_size*100/base) - 100;
941 if (growth >= server.aof_rewrite_perc) {
942 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
943 rewriteAppendOnlyFileBackground();
944 }
945 }
946 }
947
948
949 /* If we postponed an AOF buffer flush, let's try to do it every time the
950 * cron function is called. */
951 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
952
953 /* Expire a few keys per cycle, only if this is a master.
954 * On slaves we wait for DEL operations synthesized by the master
955 * in order to guarantee a strict consistency. */
956 if (server.masterhost == NULL) activeExpireCycle();
957
958 /* Close clients that need to be closed asynchronous */
959 freeClientsInAsyncFreeQueue();
960
961 /* Replication cron function -- used to reconnect to master and
962 * to detect transfer failures. */
963 run_with_period(1000) replicationCron();
964
965 server.cronloops++;
966 return 1000/REDIS_HZ;
967}
968
969/* This function gets called every time Redis is entering the
970 * main loop of the event driven library, that is, before to sleep
971 * for ready file descriptors. */
972void beforeSleep(struct aeEventLoop *eventLoop) {
973 REDIS_NOTUSED(eventLoop);
974 listNode *ln;
975 redisClient *c;
976
977 /* Try to process pending commands for clients that were just unblocked. */
978 while (listLength(server.unblocked_clients)) {
979 ln = listFirst(server.unblocked_clients);
980 redisAssert(ln != NULL);
981 c = ln->value;
982 listDelNode(server.unblocked_clients,ln);
983 c->flags &= ~REDIS_UNBLOCKED;
984
985 /* Process remaining data in the input buffer. */
986 if (c->querybuf && sdslen(c->querybuf) > 0) {
987 server.current_client = c;
988 processInputBuffer(c);
989 server.current_client = NULL;
990 }
991 }
992
993 /* Write the AOF buffer on disk */
994 flushAppendOnlyFile(0);
995}
996
997/* =========================== Server initialization ======================== */
998
999void createSharedObjects(void) {
1000 int j;
1001
1002 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
1003 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
1004 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
1005 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
1006 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
1007 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
1008 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
1009 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
1010 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
1011 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
1012 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
1013 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
1014 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
1015 "-ERR Operation against a key holding the wrong kind of value\r\n"));
1016 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
1017 "-ERR no such key\r\n"));
1018 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
1019 "-ERR syntax error\r\n"));
1020 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
1021 "-ERR source and destination objects are the same\r\n"));
1022 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
1023 "-ERR index out of range\r\n"));
1024 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
1025 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
1026 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
1027 "-LOADING Redis is loading the dataset in memory\r\n"));
1028 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
1029 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
1030 shared.masterdownerr = createObject(REDIS_STRING,sdsnew(
1031 "-MASTERDOWN Link with MASTER is down and slave-serve-stale-data is set to 'no'.\r\n"));
1032 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
1033 "-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"));
1034 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
1035 "-READONLY You can't write against a read only slave.\r\n"));
1036 shared.oomerr = createObject(REDIS_STRING,sdsnew(
1037 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
1038 shared.space = createObject(REDIS_STRING,sdsnew(" "));
1039 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
1040 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
1041
1042 for (j = 0; j < REDIS_SHARED_SELECT_CMDS; j++) {
1043 shared.select[j] = createObject(REDIS_STRING,
1044 sdscatprintf(sdsempty(),"select %d\r\n", j));
1045 }
1046 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
1047 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
1048 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
1049 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
1050 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
1051 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
1052 shared.del = createStringObject("DEL",3);
1053 shared.rpop = createStringObject("RPOP",4);
1054 shared.lpop = createStringObject("LPOP",4);
1055 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
1056 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
1057 shared.integers[j]->encoding = REDIS_ENCODING_INT;
1058 }
1059 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
1060 shared.mbulkhdr[j] = createObject(REDIS_STRING,
1061 sdscatprintf(sdsempty(),"*%d\r\n",j));
1062 shared.bulkhdr[j] = createObject(REDIS_STRING,
1063 sdscatprintf(sdsempty(),"$%d\r\n",j));
1064 }
1065}
1066
1067void initServerConfig() {
1068 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
1069 server.runid[REDIS_RUN_ID_SIZE] = '\0';
1070 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
1071 server.port = REDIS_SERVERPORT;
1072 server.bindaddr = NULL;
1073 server.unixsocket = NULL;
1074 server.unixsocketperm = 0;
1075 server.ipfd = -1;
1076 server.sofd = -1;
1077 server.dbnum = REDIS_DEFAULT_DBNUM;
1078 server.verbosity = REDIS_NOTICE;
1079 server.maxidletime = REDIS_MAXIDLETIME;
1080 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
1081 server.saveparams = NULL;
1082 server.loading = 0;
1083 server.logfile = NULL; /* NULL = log on standard output */
1084 server.syslog_enabled = 0;
1085 server.syslog_ident = zstrdup("redis");
1086 server.syslog_facility = LOG_LOCAL0;
1087 server.daemonize = 0;
1088 server.aof_state = REDIS_AOF_OFF;
1089 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1090 server.aof_no_fsync_on_rewrite = 0;
1091 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1092 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1093 server.aof_rewrite_base_size = 0;
1094 server.aof_rewrite_scheduled = 0;
1095 server.aof_last_fsync = time(NULL);
1096 server.aof_rewrite_time_last = -1;
1097 server.aof_rewrite_time_start = -1;
1098 server.aof_delayed_fsync = 0;
1099 server.aof_fd = -1;
1100 server.aof_selected_db = -1; /* Make sure the first time will not match */
1101 server.aof_flush_postponed_start = 0;
1102 server.pidfile = zstrdup("/var/run/redis.pid");
1103 server.rdb_filename = zstrdup("dump.rdb");
1104 server.aof_filename = zstrdup("appendonly.aof");
1105 server.requirepass = NULL;
1106 server.rdb_compression = 1;
1107 server.rdb_checksum = 1;
1108 server.activerehashing = 1;
1109 server.maxclients = REDIS_MAX_CLIENTS;
1110 server.bpop_blocked_clients = 0;
1111 server.maxmemory = 0;
1112 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1113 server.maxmemory_samples = 3;
1114 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1115 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1116 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1117 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1118 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1119 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1120 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1121 server.shutdown_asap = 0;
1122 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1123 server.repl_timeout = REDIS_REPL_TIMEOUT;
1124 server.lua_caller = NULL;
1125 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1126 server.lua_client = NULL;
1127 server.lua_timedout = 0;
1128
1129 updateLRUClock();
1130 resetServerSaveParams();
1131
1132 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1133 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1134 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1135 /* Replication related */
1136 server.masterauth = NULL;
1137 server.masterhost = NULL;
1138 server.masterport = 6379;
1139 server.master = NULL;
1140 server.repl_state = REDIS_REPL_NONE;
1141 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1142 server.repl_serve_stale_data = 1;
1143 server.repl_slave_ro = 1;
1144 server.repl_down_since = time(NULL);
1145
1146 /* Client output buffer limits */
1147 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1148 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1149 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1150 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1151 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1152 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1153 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1154 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1155 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1156
1157 /* Double constants initialization */
1158 R_Zero = 0.0;
1159 R_PosInf = 1.0/R_Zero;
1160 R_NegInf = -1.0/R_Zero;
1161 R_Nan = R_Zero/R_Zero;
1162
1163 /* Command table -- we intiialize it here as it is part of the
1164 * initial configuration, since command names may be changed via
1165 * redis.conf using the rename-command directive. */
1166 server.commands = dictCreate(&commandTableDictType,NULL);
1167 populateCommandTable();
1168 server.delCommand = lookupCommandByCString("del");
1169 server.multiCommand = lookupCommandByCString("multi");
1170 server.lpushCommand = lookupCommandByCString("lpush");
1171
1172 /* Slow log */
1173 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1174 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1175
1176 /* Debugging */
1177 server.assert_failed = "<no assertion failed>";
1178 server.assert_file = "<no file>";
1179 server.assert_line = 0;
1180 server.bug_report_start = 0;
1181 server.watchdog_period = 0;
1182}
1183
1184/* This function will try to raise the max number of open files accordingly to
1185 * the configured max number of clients. It will also account for 32 additional
1186 * file descriptors as we need a few more for persistence, listening
1187 * sockets, log files and so forth.
1188 *
1189 * If it will not be possible to set the limit accordingly to the configured
1190 * max number of clients, the function will do the reverse setting
1191 * server.maxclients to the value that we can actually handle. */
1192void adjustOpenFilesLimit(void) {
1193 rlim_t maxfiles = server.maxclients+32;
1194 struct rlimit limit;
1195
1196 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1197 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1198 strerror(errno));
1199 server.maxclients = 1024-32;
1200 } else {
1201 rlim_t oldlimit = limit.rlim_cur;
1202
1203 /* Set the max number of files if the current limit is not enough
1204 * for our needs. */
1205 if (oldlimit < maxfiles) {
1206 rlim_t f;
1207
1208 f = maxfiles;
1209 while(f > oldlimit) {
1210 limit.rlim_cur = f;
1211 limit.rlim_max = f;
1212 if (setrlimit(RLIMIT_NOFILE,&limit) != -1) break;
1213 f -= 128;
1214 }
1215 if (f < oldlimit) f = oldlimit;
1216 if (f != maxfiles) {
1217 server.maxclients = f-32;
1218 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1219 (int) maxfiles, strerror(errno), (int) server.maxclients);
1220 } else {
1221 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1222 (int) maxfiles);
1223 }
1224 }
1225 }
1226}
1227
1228void initServer() {
1229 int j;
1230
1231 signal(SIGHUP, SIG_IGN);
1232 signal(SIGPIPE, SIG_IGN);
1233 setupSignalHandlers();
1234
1235 if (server.syslog_enabled) {
1236 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1237 server.syslog_facility);
1238 }
1239
1240 server.current_client = NULL;
1241 server.clients = listCreate();
1242 server.clients_to_close = listCreate();
1243 server.slaves = listCreate();
1244 server.monitors = listCreate();
1245 server.unblocked_clients = listCreate();
1246
1247 createSharedObjects();
1248 adjustOpenFilesLimit();
1249 server.el = aeCreateEventLoop(server.maxclients+1024);
1250 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1251
1252 if (server.port != 0) {
1253 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1254 if (server.ipfd == ANET_ERR) {
1255 redisLog(REDIS_WARNING, "Opening port %d: %s",
1256 server.port, server.neterr);
1257 exit(1);
1258 }
1259 }
1260 if (server.unixsocket != NULL) {
1261 unlink(server.unixsocket); /* don't care if this fails */
1262 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1263 if (server.sofd == ANET_ERR) {
1264 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1265 exit(1);
1266 }
1267 }
1268 if (server.ipfd < 0 && server.sofd < 0) {
1269 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1270 exit(1);
1271 }
1272 for (j = 0; j < server.dbnum; j++) {
1273 server.db[j].dict = dictCreate(&dbDictType,NULL);
1274 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1275 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1276 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1277 server.db[j].id = j;
1278 }
1279 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1280 server.pubsub_patterns = listCreate();
1281 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1282 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1283 server.cronloops = 0;
1284 server.rdb_child_pid = -1;
1285 server.aof_child_pid = -1;
1286 aofRewriteBufferReset();
1287 server.aof_buf = sdsempty();
1288 server.lastsave = time(NULL);
1289 server.rdb_save_time_last = -1;
1290 server.rdb_save_time_start = -1;
1291 server.dirty = 0;
1292 server.stat_numcommands = 0;
1293 server.stat_numconnections = 0;
1294 server.stat_expiredkeys = 0;
1295 server.stat_evictedkeys = 0;
1296 server.stat_starttime = time(NULL);
1297 server.stat_keyspace_misses = 0;
1298 server.stat_keyspace_hits = 0;
1299 server.stat_peak_memory = 0;
1300 server.stat_fork_time = 0;
1301 server.stat_rejected_conn = 0;
1302 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1303 server.ops_sec_idx = 0;
1304 server.ops_sec_last_sample_time = mstime();
1305 server.ops_sec_last_sample_ops = 0;
1306 server.unixtime = time(NULL);
1307 server.lastbgsave_status = REDIS_OK;
1308 server.stop_writes_on_bgsave_err = 1;
1309 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1310 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1311 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1312 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1313 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1314
1315 if (server.aof_state == REDIS_AOF_ON) {
1316 server.aof_fd = open(server.aof_filename,
1317 O_WRONLY|O_APPEND|O_CREAT,0644);
1318 if (server.aof_fd == -1) {
1319 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1320 strerror(errno));
1321 exit(1);
1322 }
1323 }
1324
1325 /* 32 bit instances are limited to 4GB of address space, so if there is
1326 * no explicit limit in the user provided configuration we set a limit
1327 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1328 * useless crashes of the Redis instance. */
1329 if (server.arch_bits == 32 && server.maxmemory == 0) {
1330 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1331 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1332 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1333 }
1334
1335 scriptingInit();
1336 slowlogInit();
1337 bioInit();
1338}
1339
1340/* Populates the Redis Command Table starting from the hard coded list
1341 * we have on top of redis.c file. */
1342void populateCommandTable(void) {
1343 int j;
1344 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1345
1346 for (j = 0; j < numcommands; j++) {
1347 struct redisCommand *c = redisCommandTable+j;
1348 char *f = c->sflags;
1349 int retval;
1350
1351 while(*f != '\0') {
1352 switch(*f) {
1353 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1354 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1355 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1356 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1357 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1358 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1359 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1360 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1361 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1362 default: redisPanic("Unsupported command flag"); break;
1363 }
1364 f++;
1365 }
1366
1367 retval = dictAdd(server.commands, sdsnew(c->name), c);
1368 assert(retval == DICT_OK);
1369 }
1370}
1371
1372void resetCommandTableStats(void) {
1373 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1374 int j;
1375
1376 for (j = 0; j < numcommands; j++) {
1377 struct redisCommand *c = redisCommandTable+j;
1378
1379 c->microseconds = 0;
1380 c->calls = 0;
1381 }
1382}
1383
1384/* ========================== Redis OP Array API ============================ */
1385
1386void redisOpArrayInit(redisOpArray *oa) {
1387 oa->ops = NULL;
1388 oa->numops = 0;
1389}
1390
1391int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1392 robj **argv, int argc, int target)
1393{
1394 redisOp *op;
1395
1396 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1397 op = oa->ops+oa->numops;
1398 op->cmd = cmd;
1399 op->dbid = dbid;
1400 op->argv = argv;
1401 op->argc = argc;
1402 op->target = target;
1403 oa->numops++;
1404 return oa->numops;
1405}
1406
1407void redisOpArrayFree(redisOpArray *oa) {
1408 while(oa->numops) {
1409 int j;
1410 redisOp *op;
1411
1412 oa->numops--;
1413 op = oa->ops+oa->numops;
1414 for (j = 0; j < op->argc; j++)
1415 decrRefCount(op->argv[j]);
1416 zfree(op->argv);
1417 }
1418 zfree(oa->ops);
1419}
1420
1421/* ====================== Commands lookup and execution ===================== */
1422
1423struct redisCommand *lookupCommand(sds name) {
1424 return dictFetchValue(server.commands, name);
1425}
1426
1427struct redisCommand *lookupCommandByCString(char *s) {
1428 struct redisCommand *cmd;
1429 sds name = sdsnew(s);
1430
1431 cmd = dictFetchValue(server.commands, name);
1432 sdsfree(name);
1433 return cmd;
1434}
1435
1436/* Propagate the specified command (in the context of the specified database id)
1437 * to AOF, Slaves and Monitors.
1438 *
1439 * flags are an xor between:
1440 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1441 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1442 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1443 */
1444void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1445 int flags)
1446{
1447 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1448 feedAppendOnlyFile(cmd,dbid,argv,argc);
1449 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1450 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1451}
1452
1453/* Used inside commands to schedule the propagation of additional commands
1454 * after the current command is propagated to AOF / Replication. */
1455void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1456 int target)
1457{
1458 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1459}
1460
1461/* Call() is the core of Redis execution of a command */
1462void call(redisClient *c, int flags) {
1463 long long dirty, start = ustime(), duration;
1464
1465 /* Sent the command to clients in MONITOR mode, only if the commands are
1466 * not geneated from reading an AOF. */
1467 if (listLength(server.monitors) && !server.loading)
1468 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1469
1470 /* Call the command. */
1471 redisOpArrayInit(&server.also_propagate);
1472 dirty = server.dirty;
1473 c->cmd->proc(c);
1474 dirty = server.dirty-dirty;
1475 duration = ustime()-start;
1476
1477 /* When EVAL is called loading the AOF we don't want commands called
1478 * from Lua to go into the slowlog or to populate statistics. */
1479 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1480 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1481
1482 /* Log the command into the Slow log if needed, and populate the
1483 * per-command statistics that we show in INFO commandstats. */
1484 if (flags & REDIS_CALL_SLOWLOG)
1485 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1486 if (flags & REDIS_CALL_STATS) {
1487 c->cmd->microseconds += duration;
1488 c->cmd->calls++;
1489 }
1490
1491 /* Propagate the command into the AOF and replication link */
1492 if (flags & REDIS_CALL_PROPAGATE) {
1493 int flags = REDIS_PROPAGATE_NONE;
1494
1495 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1496 flags |= REDIS_PROPAGATE_REPL;
1497 if (dirty)
1498 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1499 if (flags != REDIS_PROPAGATE_NONE)
1500 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1501 }
1502 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1503 * PUSH command. */
1504 if (server.also_propagate.numops) {
1505 int j;
1506 redisOp *rop;
1507
1508 for (j = 0; j < server.also_propagate.numops; j++) {
1509 rop = &server.also_propagate.ops[j];
1510 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1511 }
1512 redisOpArrayFree(&server.also_propagate);
1513 }
1514 server.stat_numcommands++;
1515}
1516
1517/* If this function gets called we already read a whole
1518 * command, argments are in the client argv/argc fields.
1519 * processCommand() execute the command or prepare the
1520 * server for a bulk read from the client.
1521 *
1522 * If 1 is returned the client is still alive and valid and
1523 * and other operations can be performed by the caller. Otherwise
1524 * if 0 is returned the client was destroied (i.e. after QUIT). */
1525int processCommand(redisClient *c) {
1526 /* The QUIT command is handled separately. Normal command procs will
1527 * go through checking for replication and QUIT will cause trouble
1528 * when FORCE_REPLICATION is enabled and would be implemented in
1529 * a regular command proc. */
1530 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1531 addReply(c,shared.ok);
1532 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1533 return REDIS_ERR;
1534 }
1535
1536 /* Now lookup the command and check ASAP about trivial error conditions
1537 * such as wrong arity, bad command name and so forth. */
1538 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1539 if (!c->cmd) {
1540 addReplyErrorFormat(c,"unknown command '%s'",
1541 (char*)c->argv[0]->ptr);
1542 return REDIS_OK;
1543 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1544 (c->argc < -c->cmd->arity)) {
1545 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1546 c->cmd->name);
1547 return REDIS_OK;
1548 }
1549
1550 /* Check if the user is authenticated */
1551 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1552 {
1553 addReplyError(c,"operation not permitted");
1554 return REDIS_OK;
1555 }
1556
1557 /* Handle the maxmemory directive.
1558 *
1559 * First we try to free some memory if possible (if there are volatile
1560 * keys in the dataset). If there are not the only thing we can do
1561 * is returning an error. */
1562 if (server.maxmemory) {
1563 int retval = freeMemoryIfNeeded();
1564 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1565 addReply(c, shared.oomerr);
1566 return REDIS_OK;
1567 }
1568 }
1569
1570 /* Don't accept write commands if there are problems persisting on disk. */
1571 if (server.stop_writes_on_bgsave_err &&
1572 server.saveparamslen > 0
1573 && server.lastbgsave_status == REDIS_ERR &&
1574 c->cmd->flags & REDIS_CMD_WRITE)
1575 {
1576 addReply(c, shared.bgsaveerr);
1577 return REDIS_OK;
1578 }
1579
1580 /* Don't accept wirte commands if this is a read only slave. But
1581 * accept write commands if this is our master. */
1582 if (server.masterhost && server.repl_slave_ro &&
1583 !(c->flags & REDIS_MASTER) &&
1584 c->cmd->flags & REDIS_CMD_WRITE)
1585 {
1586 addReply(c, shared.roslaveerr);
1587 return REDIS_OK;
1588 }
1589
1590 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1591 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1592 &&
1593 c->cmd->proc != subscribeCommand &&
1594 c->cmd->proc != unsubscribeCommand &&
1595 c->cmd->proc != psubscribeCommand &&
1596 c->cmd->proc != punsubscribeCommand) {
1597 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1598 return REDIS_OK;
1599 }
1600
1601 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1602 * we are a slave with a broken link with master. */
1603 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1604 server.repl_serve_stale_data == 0 &&
1605 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1606 {
1607 addReply(c, shared.masterdownerr);
1608 return REDIS_OK;
1609 }
1610
1611 /* Loading DB? Return an error if the command is not INFO */
1612 if (server.loading && c->cmd->proc != infoCommand) {
1613 addReply(c, shared.loadingerr);
1614 return REDIS_OK;
1615 }
1616
1617 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1618 if (server.lua_timedout &&
1619 !(c->cmd->proc == shutdownCommand &&
1620 c->argc == 2 &&
1621 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1622 !(c->cmd->proc == scriptCommand &&
1623 c->argc == 2 &&
1624 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1625 {
1626 addReply(c, shared.slowscripterr);
1627 return REDIS_OK;
1628 }
1629
1630 /* Exec the command */
1631 if (c->flags & REDIS_MULTI &&
1632 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1633 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1634 {
1635 queueMultiCommand(c);
1636 addReply(c,shared.queued);
1637 } else {
1638 call(c,REDIS_CALL_FULL);
1639 }
1640 return REDIS_OK;
1641}
1642
1643/*================================== Shutdown =============================== */
1644
1645int prepareForShutdown(int flags) {
1646 int save = flags & REDIS_SHUTDOWN_SAVE;
1647 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1648
1649 redisLog(REDIS_WARNING,"User requested shutdown...");
1650 /* Kill the saving child if there is a background saving in progress.
1651 We want to avoid race conditions, for instance our saving child may
1652 overwrite the synchronous saving did by SHUTDOWN. */
1653 if (server.rdb_child_pid != -1) {
1654 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1655 kill(server.rdb_child_pid,SIGKILL);
1656 rdbRemoveTempFile(server.rdb_child_pid);
1657 }
1658 if (server.aof_state != REDIS_AOF_OFF) {
1659 /* Kill the AOF saving child as the AOF we already have may be longer
1660 * but contains the full dataset anyway. */
1661 if (server.aof_child_pid != -1) {
1662 redisLog(REDIS_WARNING,
1663 "There is a child rewriting the AOF. Killing it!");
1664 kill(server.aof_child_pid,SIGKILL);
1665 }
1666 /* Append only file: fsync() the AOF and exit */
1667 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1668 aof_fsync(server.aof_fd);
1669 }
1670 if ((server.saveparamslen > 0 && !nosave) || save) {
1671 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1672 /* Snapshotting. Perform a SYNC SAVE and exit */
1673 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1674 /* Ooops.. error saving! The best we can do is to continue
1675 * operating. Note that if there was a background saving process,
1676 * in the next cron() Redis will be notified that the background
1677 * saving aborted, handling special stuff like slaves pending for
1678 * synchronization... */
1679 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1680 return REDIS_ERR;
1681 }
1682 }
1683 if (server.daemonize) {
1684 redisLog(REDIS_NOTICE,"Removing the pid file.");
1685 unlink(server.pidfile);
1686 }
1687 /* Close the listening sockets. Apparently this allows faster restarts. */
1688 if (server.ipfd != -1) close(server.ipfd);
1689 if (server.sofd != -1) close(server.sofd);
1690 if (server.unixsocket) {
1691 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1692 unlink(server.unixsocket); /* don't care if this fails */
1693 }
1694
1695 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1696 return REDIS_OK;
1697}
1698
1699/*================================== Commands =============================== */
1700
1701void authCommand(redisClient *c) {
1702 if (!server.requirepass) {
1703 addReplyError(c,"Client sent AUTH, but no password is set");
1704 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1705 c->authenticated = 1;
1706 addReply(c,shared.ok);
1707 } else {
1708 c->authenticated = 0;
1709 addReplyError(c,"invalid password");
1710 }
1711}
1712
1713void pingCommand(redisClient *c) {
1714 addReply(c,shared.pong);
1715}
1716
1717void echoCommand(redisClient *c) {
1718 addReplyBulk(c,c->argv[1]);
1719}
1720
1721void timeCommand(redisClient *c) {
1722 struct timeval tv;
1723
1724 /* gettimeofday() can only fail if &tv is a bad addresss so we
1725 * don't check for errors. */
1726 gettimeofday(&tv,NULL);
1727 addReplyMultiBulkLen(c,2);
1728 addReplyBulkLongLong(c,tv.tv_sec);
1729 addReplyBulkLongLong(c,tv.tv_usec);
1730}
1731
1732/* Convert an amount of bytes into a human readable string in the form
1733 * of 100B, 2G, 100M, 4K, and so forth. */
1734void bytesToHuman(char *s, unsigned long long n) {
1735 double d;
1736
1737 if (n < 1024) {
1738 /* Bytes */
1739 sprintf(s,"%lluB",n);
1740 return;
1741 } else if (n < (1024*1024)) {
1742 d = (double)n/(1024);
1743 sprintf(s,"%.2fK",d);
1744 } else if (n < (1024LL*1024*1024)) {
1745 d = (double)n/(1024*1024);
1746 sprintf(s,"%.2fM",d);
1747 } else if (n < (1024LL*1024*1024*1024)) {
1748 d = (double)n/(1024LL*1024*1024);
1749 sprintf(s,"%.2fG",d);
1750 }
1751}
1752
1753/* Create the string returned by the INFO command. This is decoupled
1754 * by the INFO command itself as we need to report the same information
1755 * on memory corruption problems. */
1756sds genRedisInfoString(char *section) {
1757 sds info = sdsempty();
1758 time_t uptime = server.unixtime-server.stat_starttime;
1759 int j, numcommands;
1760 struct rusage self_ru, c_ru;
1761 unsigned long lol, bib;
1762 int allsections = 0, defsections = 0;
1763 int sections = 0;
1764
1765 if (section) {
1766 allsections = strcasecmp(section,"all") == 0;
1767 defsections = strcasecmp(section,"default") == 0;
1768 }
1769
1770 getrusage(RUSAGE_SELF, &self_ru);
1771 getrusage(RUSAGE_CHILDREN, &c_ru);
1772 getClientsMaxBuffers(&lol,&bib);
1773
1774 /* Server */
1775 if (allsections || defsections || !strcasecmp(section,"server")) {
1776 struct utsname name;
1777
1778 if (sections++) info = sdscat(info,"\r\n");
1779 uname(&name);
1780 info = sdscatprintf(info,
1781 "# Server\r\n"
1782 "redis_version:%s\r\n"
1783 "redis_git_sha1:%s\r\n"
1784 "redis_git_dirty:%d\r\n"
1785 "os:%s %s %s\r\n"
1786 "arch_bits:%d\r\n"
1787 "multiplexing_api:%s\r\n"
1788 "gcc_version:%d.%d.%d\r\n"
1789 "process_id:%ld\r\n"
1790 "run_id:%s\r\n"
1791 "tcp_port:%d\r\n"
1792 "uptime_in_seconds:%ld\r\n"
1793 "uptime_in_days:%ld\r\n"
1794 "lru_clock:%ld\r\n",
1795 REDIS_VERSION,
1796 redisGitSHA1(),
1797 strtol(redisGitDirty(),NULL,10) > 0,
1798 name.sysname, name.release, name.machine,
1799 server.arch_bits,
1800 aeGetApiName(),
1801#ifdef __GNUC__
1802 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1803#else
1804 0,0,0,
1805#endif
1806 (long) getpid(),
1807 server.runid,
1808 server.port,
1809 uptime,
1810 uptime/(3600*24),
1811 (unsigned long) server.lruclock);
1812 }
1813
1814 /* Clients */
1815 if (allsections || defsections || !strcasecmp(section,"clients")) {
1816 if (sections++) info = sdscat(info,"\r\n");
1817 info = sdscatprintf(info,
1818 "# Clients\r\n"
1819 "connected_clients:%lu\r\n"
1820 "client_longest_output_list:%lu\r\n"
1821 "client_biggest_input_buf:%lu\r\n"
1822 "blocked_clients:%d\r\n",
1823 listLength(server.clients)-listLength(server.slaves),
1824 lol, bib,
1825 server.bpop_blocked_clients);
1826 }
1827
1828 /* Memory */
1829 if (allsections || defsections || !strcasecmp(section,"memory")) {
1830 char hmem[64];
1831 char peak_hmem[64];
1832
1833 bytesToHuman(hmem,zmalloc_used_memory());
1834 bytesToHuman(peak_hmem,server.stat_peak_memory);
1835 if (sections++) info = sdscat(info,"\r\n");
1836 info = sdscatprintf(info,
1837 "# Memory\r\n"
1838 "used_memory:%zu\r\n"
1839 "used_memory_human:%s\r\n"
1840 "used_memory_rss:%zu\r\n"
1841 "used_memory_peak:%zu\r\n"
1842 "used_memory_peak_human:%s\r\n"
1843 "used_memory_lua:%lld\r\n"
1844 "mem_fragmentation_ratio:%.2f\r\n"
1845 "mem_allocator:%s\r\n",
1846 zmalloc_used_memory(),
1847 hmem,
1848 zmalloc_get_rss(),
1849 server.stat_peak_memory,
1850 peak_hmem,
1851 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1852 zmalloc_get_fragmentation_ratio(),
1853 ZMALLOC_LIB
1854 );
1855 }
1856
1857 /* Persistence */
1858 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1859 if (sections++) info = sdscat(info,"\r\n");
1860 info = sdscatprintf(info,
1861 "# Persistence\r\n"
1862 "loading:%d\r\n"
1863 "rdb_changes_since_last_save:%lld\r\n"
1864 "rdb_bgsave_in_progress:%d\r\n"
1865 "rdb_last_save_time:%ld\r\n"
1866 "rdb_last_bgsave_status:%s\r\n"
1867 "rdb_last_bgsave_time_sec:%ld\r\n"
1868 "rdb_current_bgsave_time_sec:%ld\r\n"
1869 "aof_enabled:%d\r\n"
1870 "aof_rewrite_in_progress:%d\r\n"
1871 "aof_rewrite_scheduled:%d\r\n"
1872 "aof_last_rewrite_time_sec:%ld\r\n"
1873 "aof_current_rewrite_time_sec:%ld\r\n",
1874 server.loading,
1875 server.dirty,
1876 server.rdb_child_pid != -1,
1877 server.lastsave,
1878 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1879 server.rdb_save_time_last,
1880 (server.rdb_child_pid == -1) ?
1881 -1 : time(NULL)-server.rdb_save_time_start,
1882 server.aof_state != REDIS_AOF_OFF,
1883 server.aof_child_pid != -1,
1884 server.aof_rewrite_scheduled,
1885 server.aof_rewrite_time_last,
1886 (server.aof_child_pid == -1) ?
1887 -1 : time(NULL)-server.aof_rewrite_time_start);
1888
1889 if (server.aof_state != REDIS_AOF_OFF) {
1890 info = sdscatprintf(info,
1891 "aof_current_size:%lld\r\n"
1892 "aof_base_size:%lld\r\n"
1893 "aof_pending_rewrite:%d\r\n"
1894 "aof_buffer_length:%zu\r\n"
1895 "aof_rewrite_buffer_length:%zu\r\n"
1896 "aof_pending_bio_fsync:%llu\r\n"
1897 "aof_delayed_fsync:%lu\r\n",
1898 (long long) server.aof_current_size,
1899 (long long) server.aof_rewrite_base_size,
1900 server.aof_rewrite_scheduled,
1901 sdslen(server.aof_buf),
1902 aofRewriteBufferSize(),
1903 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1904 server.aof_delayed_fsync);
1905 }
1906
1907 if (server.loading) {
1908 double perc;
1909 time_t eta, elapsed;
1910 off_t remaining_bytes = server.loading_total_bytes-
1911 server.loading_loaded_bytes;
1912
1913 perc = ((double)server.loading_loaded_bytes /
1914 server.loading_total_bytes) * 100;
1915
1916 elapsed = server.unixtime-server.loading_start_time;
1917 if (elapsed == 0) {
1918 eta = 1; /* A fake 1 second figure if we don't have
1919 enough info */
1920 } else {
1921 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1922 }
1923
1924 info = sdscatprintf(info,
1925 "loading_start_time:%ld\r\n"
1926 "loading_total_bytes:%llu\r\n"
1927 "loading_loaded_bytes:%llu\r\n"
1928 "loading_loaded_perc:%.2f\r\n"
1929 "loading_eta_seconds:%ld\r\n"
1930 ,(unsigned long) server.loading_start_time,
1931 (unsigned long long) server.loading_total_bytes,
1932 (unsigned long long) server.loading_loaded_bytes,
1933 perc,
1934 eta
1935 );
1936 }
1937 }
1938
1939 /* Stats */
1940 if (allsections || defsections || !strcasecmp(section,"stats")) {
1941 if (sections++) info = sdscat(info,"\r\n");
1942 info = sdscatprintf(info,
1943 "# Stats\r\n"
1944 "total_connections_received:%lld\r\n"
1945 "total_commands_processed:%lld\r\n"
1946 "instantaneous_ops_per_sec:%lld\r\n"
1947 "rejected_connections:%lld\r\n"
1948 "expired_keys:%lld\r\n"
1949 "evicted_keys:%lld\r\n"
1950 "keyspace_hits:%lld\r\n"
1951 "keyspace_misses:%lld\r\n"
1952 "pubsub_channels:%ld\r\n"
1953 "pubsub_patterns:%lu\r\n"
1954 "latest_fork_usec:%lld\r\n",
1955 server.stat_numconnections,
1956 server.stat_numcommands,
1957 getOperationsPerSecond(),
1958 server.stat_rejected_conn,
1959 server.stat_expiredkeys,
1960 server.stat_evictedkeys,
1961 server.stat_keyspace_hits,
1962 server.stat_keyspace_misses,
1963 dictSize(server.pubsub_channels),
1964 listLength(server.pubsub_patterns),
1965 server.stat_fork_time);
1966 }
1967
1968 /* Replication */
1969 if (allsections || defsections || !strcasecmp(section,"replication")) {
1970 if (sections++) info = sdscat(info,"\r\n");
1971 info = sdscatprintf(info,
1972 "# Replication\r\n"
1973 "role:%s\r\n",
1974 server.masterhost == NULL ? "master" : "slave");
1975 if (server.masterhost) {
1976 info = sdscatprintf(info,
1977 "master_host:%s\r\n"
1978 "master_port:%d\r\n"
1979 "master_link_status:%s\r\n"
1980 "master_last_io_seconds_ago:%d\r\n"
1981 "master_sync_in_progress:%d\r\n"
1982 ,server.masterhost,
1983 server.masterport,
1984 (server.repl_state == REDIS_REPL_CONNECTED) ?
1985 "up" : "down",
1986 server.master ?
1987 ((int)(server.unixtime-server.master->lastinteraction)) : -1,
1988 server.repl_state == REDIS_REPL_TRANSFER
1989 );
1990
1991 if (server.repl_state == REDIS_REPL_TRANSFER) {
1992 info = sdscatprintf(info,
1993 "master_sync_left_bytes:%ld\r\n"
1994 "master_sync_last_io_seconds_ago:%d\r\n"
1995 ,(long)server.repl_transfer_left,
1996 (int)(server.unixtime-server.repl_transfer_lastio)
1997 );
1998 }
1999
2000 if (server.repl_state != REDIS_REPL_CONNECTED) {
2001 info = sdscatprintf(info,
2002 "master_link_down_since_seconds:%ld\r\n",
2003 (long)server.unixtime-server.repl_down_since);
2004 }
2005 }
2006 info = sdscatprintf(info,
2007 "connected_slaves:%lu\r\n",
2008 listLength(server.slaves));
2009 if (listLength(server.slaves)) {
2010 int slaveid = 0;
2011 listNode *ln;
2012 listIter li;
2013
2014 listRewind(server.slaves,&li);
2015 while((ln = listNext(&li))) {
2016 redisClient *slave = listNodeValue(ln);
2017 char *state = NULL;
2018 char ip[32];
2019 int port;
2020
2021 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
2022 switch(slave->replstate) {
2023 case REDIS_REPL_WAIT_BGSAVE_START:
2024 case REDIS_REPL_WAIT_BGSAVE_END:
2025 state = "wait_bgsave";
2026 break;
2027 case REDIS_REPL_SEND_BULK:
2028 state = "send_bulk";
2029 break;
2030 case REDIS_REPL_ONLINE:
2031 state = "online";
2032 break;
2033 }
2034 if (state == NULL) continue;
2035 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
2036 slaveid,ip,port,state);
2037 slaveid++;
2038 }
2039 }
2040 }
2041
2042 /* CPU */
2043 if (allsections || defsections || !strcasecmp(section,"cpu")) {
2044 if (sections++) info = sdscat(info,"\r\n");
2045 info = sdscatprintf(info,
2046 "# CPU\r\n"
2047 "used_cpu_sys:%.2f\r\n"
2048 "used_cpu_user:%.2f\r\n"
2049 "used_cpu_sys_children:%.2f\r\n"
2050 "used_cpu_user_children:%.2f\r\n",
2051 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
2052 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
2053 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
2054 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
2055 }
2056
2057 /* cmdtime */
2058 if (allsections || !strcasecmp(section,"commandstats")) {
2059 if (sections++) info = sdscat(info,"\r\n");
2060 info = sdscatprintf(info, "# Commandstats\r\n");
2061 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
2062 for (j = 0; j < numcommands; j++) {
2063 struct redisCommand *c = redisCommandTable+j;
2064
2065 if (!c->calls) continue;
2066 info = sdscatprintf(info,
2067 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
2068 c->name, c->calls, c->microseconds,
2069 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
2070 }
2071 }
2072
2073 /* Key space */
2074 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
2075 if (sections++) info = sdscat(info,"\r\n");
2076 info = sdscatprintf(info, "# Keyspace\r\n");
2077 for (j = 0; j < server.dbnum; j++) {
2078 long long keys, vkeys;
2079
2080 keys = dictSize(server.db[j].dict);
2081 vkeys = dictSize(server.db[j].expires);
2082 if (keys || vkeys) {
2083 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2084 j, keys, vkeys);
2085 }
2086 }
2087 }
2088 return info;
2089}
2090
2091void infoCommand(redisClient *c) {
2092 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2093
2094 if (c->argc > 2) {
2095 addReply(c,shared.syntaxerr);
2096 return;
2097 }
2098 sds info = genRedisInfoString(section);
2099 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2100 (unsigned long)sdslen(info)));
2101 addReplySds(c,info);
2102 addReply(c,shared.crlf);
2103}
2104
2105void monitorCommand(redisClient *c) {
2106 /* ignore MONITOR if aleady slave or in monitor mode */
2107 if (c->flags & REDIS_SLAVE) return;
2108
2109 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2110 c->slaveseldb = 0;
2111 listAddNodeTail(server.monitors,c);
2112 addReply(c,shared.ok);
2113}
2114
2115/* ============================ Maxmemory directive ======================== */
2116
2117/* This function gets called when 'maxmemory' is set on the config file to limit
2118 * the max memory used by the server, before processing a command.
2119 *
2120 * The goal of the function is to free enough memory to keep Redis under the
2121 * configured memory limit.
2122 *
2123 * The function starts calculating how many bytes should be freed to keep
2124 * Redis under the limit, and enters a loop selecting the best keys to
2125 * evict accordingly to the configured policy.
2126 *
2127 * If all the bytes needed to return back under the limit were freed the
2128 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2129 * should block the execution of commands that will result in more memory
2130 * used by the server.
2131 */
2132int freeMemoryIfNeeded(void) {
2133 size_t mem_used, mem_tofree, mem_freed;
2134 int slaves = listLength(server.slaves);
2135
2136 /* Remove the size of slaves output buffers and AOF buffer from the
2137 * count of used memory. */
2138 mem_used = zmalloc_used_memory();
2139 if (slaves) {
2140 listIter li;
2141 listNode *ln;
2142
2143 listRewind(server.slaves,&li);
2144 while((ln = listNext(&li))) {
2145 redisClient *slave = listNodeValue(ln);
2146 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2147 if (obuf_bytes > mem_used)
2148 mem_used = 0;
2149 else
2150 mem_used -= obuf_bytes;
2151 }
2152 }
2153 if (server.aof_state != REDIS_AOF_OFF) {
2154 mem_used -= sdslen(server.aof_buf);
2155 mem_used -= aofRewriteBufferSize();
2156 }
2157
2158 /* Check if we are over the memory limit. */
2159 if (mem_used <= server.maxmemory) return REDIS_OK;
2160
2161 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2162 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2163
2164 /* Compute how much memory we need to free. */
2165 mem_tofree = mem_used - server.maxmemory;
2166 mem_freed = 0;
2167 while (mem_freed < mem_tofree) {
2168 int j, k, keys_freed = 0;
2169
2170 for (j = 0; j < server.dbnum; j++) {
2171 long bestval = 0; /* just to prevent warning */
2172 sds bestkey = NULL;
2173 struct dictEntry *de;
2174 redisDb *db = server.db+j;
2175 dict *dict;
2176
2177 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2178 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2179 {
2180 dict = server.db[j].dict;
2181 } else {
2182 dict = server.db[j].expires;
2183 }
2184 if (dictSize(dict) == 0) continue;
2185
2186 /* volatile-random and allkeys-random policy */
2187 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2188 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2189 {
2190 de = dictGetRandomKey(dict);
2191 bestkey = dictGetKey(de);
2192 }
2193
2194 /* volatile-lru and allkeys-lru policy */
2195 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2196 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2197 {
2198 for (k = 0; k < server.maxmemory_samples; k++) {
2199 sds thiskey;
2200 long thisval;
2201 robj *o;
2202
2203 de = dictGetRandomKey(dict);
2204 thiskey = dictGetKey(de);
2205 /* When policy is volatile-lru we need an additonal lookup
2206 * to locate the real key, as dict is set to db->expires. */
2207 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2208 de = dictFind(db->dict, thiskey);
2209 o = dictGetVal(de);
2210 thisval = estimateObjectIdleTime(o);
2211
2212 /* Higher idle time is better candidate for deletion */
2213 if (bestkey == NULL || thisval > bestval) {
2214 bestkey = thiskey;
2215 bestval = thisval;
2216 }
2217 }
2218 }
2219
2220 /* volatile-ttl */
2221 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2222 for (k = 0; k < server.maxmemory_samples; k++) {
2223 sds thiskey;
2224 long thisval;
2225
2226 de = dictGetRandomKey(dict);
2227 thiskey = dictGetKey(de);
2228 thisval = (long) dictGetVal(de);
2229
2230 /* Expire sooner (minor expire unix timestamp) is better
2231 * candidate for deletion */
2232 if (bestkey == NULL || thisval < bestval) {
2233 bestkey = thiskey;
2234 bestval = thisval;
2235 }
2236 }
2237 }
2238
2239 /* Finally remove the selected key. */
2240 if (bestkey) {
2241 long long delta;
2242
2243 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2244 propagateExpire(db,keyobj);
2245 /* We compute the amount of memory freed by dbDelete() alone.
2246 * It is possible that actually the memory needed to propagate
2247 * the DEL in AOF and replication link is greater than the one
2248 * we are freeing removing the key, but we can't account for
2249 * that otherwise we would never exit the loop.
2250 *
2251 * AOF and Output buffer memory will be freed eventually so
2252 * we only care about memory used by the key space. */
2253 delta = (long long) zmalloc_used_memory();
2254 dbDelete(db,keyobj);
2255 delta -= (long long) zmalloc_used_memory();
2256 mem_freed += delta;
2257 server.stat_evictedkeys++;
2258 decrRefCount(keyobj);
2259 keys_freed++;
2260
2261 /* When the memory to free starts to be big enough, we may
2262 * start spending so much time here that is impossible to
2263 * deliver data to the slaves fast enough, so we force the
2264 * transmission here inside the loop. */
2265 if (slaves) flushSlavesOutputBuffers();
2266 }
2267 }
2268 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2269 }
2270 return REDIS_OK;
2271}
2272
2273/* =================================== Main! ================================ */
2274
2275#ifdef __linux__
2276int linuxOvercommitMemoryValue(void) {
2277 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2278 char buf[64];
2279
2280 if (!fp) return -1;
2281 if (fgets(buf,64,fp) == NULL) {
2282 fclose(fp);
2283 return -1;
2284 }
2285 fclose(fp);
2286
2287 return atoi(buf);
2288}
2289
2290void linuxOvercommitMemoryWarning(void) {
2291 if (linuxOvercommitMemoryValue() == 0) {
2292 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.");
2293 }
2294}
2295#endif /* __linux__ */
2296
2297void createPidFile(void) {
2298 /* Try to write the pid file in a best-effort way. */
2299 FILE *fp = fopen(server.pidfile,"w");
2300 if (fp) {
2301 fprintf(fp,"%d\n",(int)getpid());
2302 fclose(fp);
2303 }
2304}
2305
2306void daemonize(void) {
2307 int fd;
2308
2309 if (fork() != 0) exit(0); /* parent exits */
2310 setsid(); /* create a new session */
2311
2312 /* Every output goes to /dev/null. If Redis is daemonized but
2313 * the 'logfile' is set to 'stdout' in the configuration file
2314 * it will not log at all. */
2315 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2316 dup2(fd, STDIN_FILENO);
2317 dup2(fd, STDOUT_FILENO);
2318 dup2(fd, STDERR_FILENO);
2319 if (fd > STDERR_FILENO) close(fd);
2320 }
2321}
2322
2323void version() {
2324 printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d\n",
2325 REDIS_VERSION,
2326 redisGitSHA1(),
2327 atoi(redisGitDirty()) > 0,
2328 ZMALLOC_LIB,
2329 sizeof(long) == 4 ? 32 : 64);
2330 exit(0);
2331}
2332
2333void usage() {
2334 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2335 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2336 fprintf(stderr," ./redis-server -v or --version\n");
2337 fprintf(stderr," ./redis-server -h or --help\n");
2338 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2339 fprintf(stderr,"Examples:\n");
2340 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2341 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2342 fprintf(stderr," ./redis-server --port 7777\n");
2343 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2344 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2345 exit(1);
2346}
2347
2348void redisAsciiArt(void) {
2349#include "asciilogo.h"
2350 char *buf = zmalloc(1024*16);
2351
2352 snprintf(buf,1024*16,ascii_logo,
2353 REDIS_VERSION,
2354 redisGitSHA1(),
2355 strtol(redisGitDirty(),NULL,10) > 0,
2356 (sizeof(long) == 8) ? "64" : "32",
2357 "stand alone",
2358 server.port,
2359 (long) getpid()
2360 );
2361 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2362 zfree(buf);
2363}
2364
2365static void sigtermHandler(int sig) {
2366 REDIS_NOTUSED(sig);
2367
2368 redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2369 server.shutdown_asap = 1;
2370}
2371
2372void setupSignalHandlers(void) {
2373 struct sigaction act;
2374
2375 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2376 * Otherwise, sa_handler is used. */
2377 sigemptyset(&act.sa_mask);
2378 act.sa_flags = 0;
2379 act.sa_handler = sigtermHandler;
2380 sigaction(SIGTERM, &act, NULL);
2381
2382#ifdef HAVE_BACKTRACE
2383 sigemptyset(&act.sa_mask);
2384 act.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
2385 act.sa_sigaction = sigsegvHandler;
2386 sigaction(SIGSEGV, &act, NULL);
2387 sigaction(SIGBUS, &act, NULL);
2388 sigaction(SIGFPE, &act, NULL);
2389 sigaction(SIGILL, &act, NULL);
2390#endif
2391 return;
2392}
2393
2394void memtest(size_t megabytes, int passes);
2395
2396int main(int argc, char **argv) {
2397 long long start;
2398 struct timeval tv;
2399
2400 /* We need to initialize our libraries, and the server configuration. */
2401 zmalloc_enable_thread_safeness();
2402 srand(time(NULL)^getpid());
2403 gettimeofday(&tv,NULL);
2404 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2405 initServerConfig();
2406
2407 if (argc >= 2) {
2408 int j = 1; /* First option to parse in argv[] */
2409 sds options = sdsempty();
2410 char *configfile = NULL;
2411
2412 /* Handle special options --help and --version */
2413 if (strcmp(argv[1], "-v") == 0 ||
2414 strcmp(argv[1], "--version") == 0) version();
2415 if (strcmp(argv[1], "--help") == 0 ||
2416 strcmp(argv[1], "-h") == 0) usage();
2417 if (strcmp(argv[1], "--test-memory") == 0) {
2418 if (argc == 3) {
2419 memtest(atoi(argv[2]),50);
2420 exit(0);
2421 } else {
2422 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2423 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2424 exit(1);
2425 }
2426 }
2427
2428 /* First argument is the config file name? */
2429 if (argv[j][0] != '-' || argv[j][1] != '-')
2430 configfile = argv[j++];
2431 /* All the other options are parsed and conceptually appended to the
2432 * configuration file. For instance --port 6380 will generate the
2433 * string "port 6380\n" to be parsed after the actual file name
2434 * is parsed, if any. */
2435 while(j != argc) {
2436 if (argv[j][0] == '-' && argv[j][1] == '-') {
2437 /* Option name */
2438 if (sdslen(options)) options = sdscat(options,"\n");
2439 options = sdscat(options,argv[j]+2);
2440 options = sdscat(options," ");
2441 } else {
2442 /* Option argument */
2443 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2444 options = sdscat(options," ");
2445 }
2446 j++;
2447 }
2448 resetServerSaveParams();
2449 loadServerConfig(configfile,options);
2450 sdsfree(options);
2451 } else {
2452 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'");
2453 }
2454 if (server.daemonize) daemonize();
2455 initServer();
2456 if (server.daemonize) createPidFile();
2457 redisAsciiArt();
2458 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2459#ifdef __linux__
2460 linuxOvercommitMemoryWarning();
2461#endif
2462 start = ustime();
2463 if (server.aof_state == REDIS_AOF_ON) {
2464 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2465 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2466 } else {
2467 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2468 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2469 (float)(ustime()-start)/1000000);
2470 } else if (errno != ENOENT) {
2471 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2472 exit(1);
2473 }
2474 }
2475 if (server.ipfd > 0)
2476 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2477 if (server.sofd > 0)
2478 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2479 aeSetBeforeSleepProc(server.el,beforeSleep);
2480 aeMain(server.el);
2481 aeDeleteEventLoop(server.el);
2482 return 0;
2483}
2484
2485/* The End */