]> git.saurik.com Git - redis.git/blame_incremental - src/redis.c
SIGSEGV handler refactored so that we can reuse stack trace and current client loggin...
[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
52/* Our shared "common" objects */
53
54struct sharedObjectsStruct shared;
55
56/* Global vars that are actually used as constants. The following double
57 * values are used for double on-disk serialization, and are initialized
58 * at runtime to avoid strange compiler optimizations. */
59
60double R_Zero, R_PosInf, R_NegInf, R_Nan;
61
62/*================================= Globals ================================= */
63
64/* Global vars */
65struct redisServer server; /* server global state */
66struct redisCommand *commandTable;
67
68/* Our command table.
69 *
70 * Every entry is composed of the following fields:
71 *
72 * name: a string representing the command name.
73 * function: pointer to the C function implementing the command.
74 * arity: number of arguments, it is possible to use -N to say >= N
75 * sflags: command flags as string. See below for a table of flags.
76 * flags: flags as bitmask. Computed by Redis using the 'sflags' field.
77 * get_keys_proc: an optional function to get key arguments from a command.
78 * This is only used when the following three fields are not
79 * enough to specify what arguments are keys.
80 * first_key_index: first argument that is a key
81 * last_key_index: last argument that is a key
82 * key_step: step to get all the keys from first to last argument. For instance
83 * in MSET the step is two since arguments are key,val,key,val,...
84 * microseconds: microseconds of total execution time for this command.
85 * calls: total number of calls of this command.
86 *
87 * The flags, microseconds and calls fields are computed by Redis and should
88 * always be set to zero.
89 *
90 * Command flags are expressed using strings where every character represents
91 * a flag. Later the populateCommandTable() function will take care of
92 * populating the real 'flags' field using this characters.
93 *
94 * This is the meaning of the flags:
95 *
96 * w: write command (may modify the key space).
97 * r: read command (will never modify the key space).
98 * m: may increase memory usage once called. Don't allow if out of memory.
99 * a: admin command, like SAVE or SHUTDOWN.
100 * p: Pub/Sub related command.
101 * f: force replication of this command, regarless of server.dirty.
102 * s: command not allowed in scripts.
103 * R: random command. Command is not deterministic, that is, the same command
104 * with the same arguments, with the same key space, may have different
105 * results. For instance SPOP and RANDOMKEY are two random commands.
106 * S: Sort command output array if called from script, so that the output
107 * is deterministic.
108 */
109struct redisCommand redisCommandTable[] = {
110 {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
111 {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
112 {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
113 {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
114 {"psetex",psetexCommand,4,"wm",0,noPreloadGetKeys,1,1,1,0,0},
115 {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
116 {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
117 {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
118 {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
119 {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
120 {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
121 {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
122 {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
123 {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
124 {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
125 {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
126 {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
127 {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
128 {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
129 {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
130 {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
131 {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
132 {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
133 {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
134 {"brpop",brpopCommand,-3,"ws",0,NULL,1,1,1,0,0},
135 {"brpoplpush",brpoplpushCommand,4,"wms",0,NULL,1,2,1,0,0},
136 {"blpop",blpopCommand,-3,"ws",0,NULL,1,-2,1,0,0},
137 {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
138 {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
139 {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
140 {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
141 {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
142 {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
143 {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
144 {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
145 {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
146 {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
147 {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
148 {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
149 {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
150 {"srandmember",srandmemberCommand,2,"rR",0,NULL,1,1,1,0,0},
151 {"sinter",sinterCommand,-2,"rS",0,NULL,1,-1,1,0,0},
152 {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
153 {"sunion",sunionCommand,-2,"rS",0,NULL,1,-1,1,0,0},
154 {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
155 {"sdiff",sdiffCommand,-2,"rS",0,NULL,1,-1,1,0,0},
156 {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,1,-1,1,0,0},
157 {"smembers",sinterCommand,2,"rS",0,NULL,1,1,1,0,0},
158 {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
159 {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
160 {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
161 {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
162 {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
163 {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
164 {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
165 {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
166 {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
167 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
168 {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
169 {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
170 {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
171 {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
172 {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
173 {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
174 {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
175 {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
176 {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
177 {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
178 {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
179 {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
180 {"hincrbyfloat",hincrbyfloatCommand,4,"wm",0,NULL,1,1,1,0,0},
181 {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
182 {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
183 {"hkeys",hkeysCommand,2,"rS",0,NULL,1,1,1,0,0},
184 {"hvals",hvalsCommand,2,"rS",0,NULL,1,1,1,0,0},
185 {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
186 {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
187 {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
188 {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
189 {"incrbyfloat",incrbyfloatCommand,3,"wm",0,NULL,1,1,1,0,0},
190 {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
191 {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
192 {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
193 {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
194 {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
195 {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
196 {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
197 {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
198 {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
199 {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
200 {"pexpire",pexpireCommand,3,"w",0,NULL,1,1,1,0,0},
201 {"pexpireat",pexpireatCommand,3,"w",0,NULL,1,1,1,0,0},
202 {"keys",keysCommand,2,"rS",0,NULL,0,0,0,0,0},
203 {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
204 {"auth",authCommand,2,"rs",0,NULL,0,0,0,0,0},
205 {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
206 {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
207 {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
208 {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
209 {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
210 {"shutdown",shutdownCommand,-1,"ar",0,NULL,0,0,0,0,0},
211 {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
212 {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
213 {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
214 {"exec",execCommand,1,"s",0,NULL,0,0,0,0,0},
215 {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
216 {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
217 {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
218 {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
219 {"sort",sortCommand,-2,"wmS",0,NULL,1,1,1,0,0},
220 {"info",infoCommand,-1,"r",0,NULL,0,0,0,0,0},
221 {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
222 {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
223 {"pttl",pttlCommand,2,"r",0,NULL,1,1,1,0,0},
224 {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
225 {"slaveof",slaveofCommand,3,"aws",0,NULL,0,0,0,0,0},
226 {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
227 {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
228 {"subscribe",subscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
229 {"unsubscribe",unsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
230 {"psubscribe",psubscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
231 {"punsubscribe",punsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
232 {"publish",publishCommand,3,"pf",0,NULL,0,0,0,0,0},
233 {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
234 {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
235 {"cluster",clusterCommand,-2,"ar",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 {"asking",askingCommand,1,"r",0,NULL,0,0,0,0,0},
239 {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
240 {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
241 {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
242 {"eval",evalCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
243 {"evalsha",evalShaCommand,-3,"s",0,zunionInterGetKeys,0,0,0,0,0},
244 {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
245 {"script",scriptCommand,-2,"ras",0,NULL,0,0,0,0,0},
246 {"time",timeCommand,1,"rR",0,NULL,0,0,0,0,0}
247};
248
249/*============================ Utility functions ============================ */
250
251/* Low level logging. To use only for very big messages, otherwise
252 * redisLog() is to prefer. */
253void redisLogRaw(int level, const char *msg) {
254 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
255 const char *c = ".-*#";
256 time_t now = time(NULL);
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 strftime(buf,sizeof(buf),"%d %b %H:%M:%S",localtime(&now));
271 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
272 }
273 fflush(fp);
274
275 if (server.logfile) fclose(fp);
276
277 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
278}
279
280/* Like redisLogRaw() but with printf-alike support. This is the funciton that
281 * is used across the code. The raw version is only used in order to dump
282 * the INFO output on crash. */
283void redisLog(int level, const char *fmt, ...) {
284 va_list ap;
285 char msg[REDIS_MAX_LOGMSG_LEN];
286
287 if ((level&0xff) < server.verbosity) return;
288
289 va_start(ap, fmt);
290 vsnprintf(msg, sizeof(msg), fmt, ap);
291 va_end(ap);
292
293 redisLogRaw(level,msg);
294}
295
296/* Redis generally does not try to recover from out of memory conditions
297 * when allocating objects or strings, it is not clear if it will be possible
298 * to report this condition to the client since the networking layer itself
299 * is based on heap allocation for send buffers, so we simply abort.
300 * At least the code will be simpler to read... */
301void oom(const char *msg) {
302 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
303 sleep(1);
304 abort();
305}
306
307/* Return the UNIX time in microseconds */
308long long ustime(void) {
309 struct timeval tv;
310 long long ust;
311
312 gettimeofday(&tv, NULL);
313 ust = ((long long)tv.tv_sec)*1000000;
314 ust += tv.tv_usec;
315 return ust;
316}
317
318/* Return the UNIX time in milliseconds */
319long long mstime(void) {
320 return ustime()/1000;
321}
322
323/*====================== Hash table type implementation ==================== */
324
325/* This is an hash table type that uses the SDS dynamic strings libary as
326 * keys and radis objects as values (objects can hold SDS strings,
327 * lists, sets). */
328
329void dictVanillaFree(void *privdata, void *val)
330{
331 DICT_NOTUSED(privdata);
332 zfree(val);
333}
334
335void dictListDestructor(void *privdata, void *val)
336{
337 DICT_NOTUSED(privdata);
338 listRelease((list*)val);
339}
340
341int dictSdsKeyCompare(void *privdata, const void *key1,
342 const void *key2)
343{
344 int l1,l2;
345 DICT_NOTUSED(privdata);
346
347 l1 = sdslen((sds)key1);
348 l2 = sdslen((sds)key2);
349 if (l1 != l2) return 0;
350 return memcmp(key1, key2, l1) == 0;
351}
352
353/* A case insensitive version used for the command lookup table. */
354int dictSdsKeyCaseCompare(void *privdata, const void *key1,
355 const void *key2)
356{
357 DICT_NOTUSED(privdata);
358
359 return strcasecmp(key1, key2) == 0;
360}
361
362void dictRedisObjectDestructor(void *privdata, void *val)
363{
364 DICT_NOTUSED(privdata);
365
366 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
367 decrRefCount(val);
368}
369
370void dictSdsDestructor(void *privdata, void *val)
371{
372 DICT_NOTUSED(privdata);
373
374 sdsfree(val);
375}
376
377int dictObjKeyCompare(void *privdata, const void *key1,
378 const void *key2)
379{
380 const robj *o1 = key1, *o2 = key2;
381 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
382}
383
384unsigned int dictObjHash(const void *key) {
385 const robj *o = key;
386 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
387}
388
389unsigned int dictSdsHash(const void *key) {
390 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
391}
392
393unsigned int dictSdsCaseHash(const void *key) {
394 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
395}
396
397int dictEncObjKeyCompare(void *privdata, const void *key1,
398 const void *key2)
399{
400 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
401 int cmp;
402
403 if (o1->encoding == REDIS_ENCODING_INT &&
404 o2->encoding == REDIS_ENCODING_INT)
405 return o1->ptr == o2->ptr;
406
407 o1 = getDecodedObject(o1);
408 o2 = getDecodedObject(o2);
409 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
410 decrRefCount(o1);
411 decrRefCount(o2);
412 return cmp;
413}
414
415unsigned int dictEncObjHash(const void *key) {
416 robj *o = (robj*) key;
417
418 if (o->encoding == REDIS_ENCODING_RAW) {
419 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
420 } else {
421 if (o->encoding == REDIS_ENCODING_INT) {
422 char buf[32];
423 int len;
424
425 len = ll2string(buf,32,(long)o->ptr);
426 return dictGenHashFunction((unsigned char*)buf, len);
427 } else {
428 unsigned int hash;
429
430 o = getDecodedObject(o);
431 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
432 decrRefCount(o);
433 return hash;
434 }
435 }
436}
437
438/* Sets type hash table */
439dictType setDictType = {
440 dictEncObjHash, /* hash function */
441 NULL, /* key dup */
442 NULL, /* val dup */
443 dictEncObjKeyCompare, /* key compare */
444 dictRedisObjectDestructor, /* key destructor */
445 NULL /* val destructor */
446};
447
448/* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
449dictType zsetDictType = {
450 dictEncObjHash, /* hash function */
451 NULL, /* key dup */
452 NULL, /* val dup */
453 dictEncObjKeyCompare, /* key compare */
454 dictRedisObjectDestructor, /* key destructor */
455 NULL /* val destructor */
456};
457
458/* Db->dict, keys are sds strings, vals are Redis objects. */
459dictType dbDictType = {
460 dictSdsHash, /* hash function */
461 NULL, /* key dup */
462 NULL, /* val dup */
463 dictSdsKeyCompare, /* key compare */
464 dictSdsDestructor, /* key destructor */
465 dictRedisObjectDestructor /* val destructor */
466};
467
468/* Db->expires */
469dictType keyptrDictType = {
470 dictSdsHash, /* hash function */
471 NULL, /* key dup */
472 NULL, /* val dup */
473 dictSdsKeyCompare, /* key compare */
474 NULL, /* key destructor */
475 NULL /* val destructor */
476};
477
478/* Command table. sds string -> command struct pointer. */
479dictType commandTableDictType = {
480 dictSdsCaseHash, /* hash function */
481 NULL, /* key dup */
482 NULL, /* val dup */
483 dictSdsKeyCaseCompare, /* key compare */
484 dictSdsDestructor, /* key destructor */
485 NULL /* val destructor */
486};
487
488/* Hash type hash table (note that small hashes are represented with zimpaps) */
489dictType hashDictType = {
490 dictEncObjHash, /* hash function */
491 NULL, /* key dup */
492 NULL, /* val dup */
493 dictEncObjKeyCompare, /* key compare */
494 dictRedisObjectDestructor, /* key destructor */
495 dictRedisObjectDestructor /* val destructor */
496};
497
498/* Keylist hash table type has unencoded redis objects as keys and
499 * lists as values. It's used for blocking operations (BLPOP) and to
500 * map swapped keys to a list of clients waiting for this keys to be loaded. */
501dictType keylistDictType = {
502 dictObjHash, /* hash function */
503 NULL, /* key dup */
504 NULL, /* val dup */
505 dictObjKeyCompare, /* key compare */
506 dictRedisObjectDestructor, /* key destructor */
507 dictListDestructor /* val destructor */
508};
509
510/* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
511 * clusterNode structures. */
512dictType clusterNodesDictType = {
513 dictSdsHash, /* hash function */
514 NULL, /* key dup */
515 NULL, /* val dup */
516 dictSdsKeyCompare, /* key compare */
517 dictSdsDestructor, /* key destructor */
518 NULL /* val destructor */
519};
520
521int htNeedsResize(dict *dict) {
522 long long size, used;
523
524 size = dictSlots(dict);
525 used = dictSize(dict);
526 return (size && used && size > DICT_HT_INITIAL_SIZE &&
527 (used*100/size < REDIS_HT_MINFILL));
528}
529
530/* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
531 * we resize the hash table to save memory */
532void tryResizeHashTables(void) {
533 int j;
534
535 for (j = 0; j < server.dbnum; j++) {
536 if (htNeedsResize(server.db[j].dict))
537 dictResize(server.db[j].dict);
538 if (htNeedsResize(server.db[j].expires))
539 dictResize(server.db[j].expires);
540 }
541}
542
543/* Our hash table implementation performs rehashing incrementally while
544 * we write/read from the hash table. Still if the server is idle, the hash
545 * table will use two tables for a long time. So we try to use 1 millisecond
546 * of CPU time at every serverCron() loop in order to rehash some key. */
547void incrementallyRehash(void) {
548 int j;
549
550 for (j = 0; j < server.dbnum; j++) {
551 if (dictIsRehashing(server.db[j].dict)) {
552 dictRehashMilliseconds(server.db[j].dict,1);
553 break; /* already used our millisecond for this loop... */
554 }
555 }
556}
557
558/* This function is called once a background process of some kind terminates,
559 * as we want to avoid resizing the hash tables when there is a child in order
560 * to play well with copy-on-write (otherwise when a resize happens lots of
561 * memory pages are copied). The goal of this function is to update the ability
562 * for dict.c to resize the hash tables accordingly to the fact we have o not
563 * running childs. */
564void updateDictResizePolicy(void) {
565 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)
566 dictEnableResize();
567 else
568 dictDisableResize();
569}
570
571/* ======================= Cron: called every 100 ms ======================== */
572
573/* Try to expire a few timed out keys. The algorithm used is adaptive and
574 * will use few CPU cycles if there are few expiring keys, otherwise
575 * it will get more aggressive to avoid that too much memory is used by
576 * keys that can be removed from the keyspace. */
577void activeExpireCycle(void) {
578 int j;
579
580 for (j = 0; j < server.dbnum; j++) {
581 int expired;
582 redisDb *db = server.db+j;
583
584 /* Continue to expire if at the end of the cycle more than 25%
585 * of the keys were expired. */
586 do {
587 long num = dictSize(db->expires);
588 long long now = mstime();
589
590 expired = 0;
591 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
592 num = REDIS_EXPIRELOOKUPS_PER_CRON;
593 while (num--) {
594 dictEntry *de;
595 long long t;
596
597 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
598 t = dictGetSignedIntegerVal(de);
599 if (now > t) {
600 sds key = dictGetKey(de);
601 robj *keyobj = createStringObject(key,sdslen(key));
602
603 propagateExpire(db,keyobj);
604 dbDelete(db,keyobj);
605 decrRefCount(keyobj);
606 expired++;
607 server.stat_expiredkeys++;
608 }
609 }
610 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
611 }
612}
613
614void updateLRUClock(void) {
615 server.lruclock = (time(NULL)/REDIS_LRU_CLOCK_RESOLUTION) &
616 REDIS_LRU_CLOCK_MAX;
617}
618
619
620/* Add a sample to the operations per second array of samples. */
621void trackOperationsPerSecond(void) {
622 long long t = mstime() - server.ops_sec_last_sample_time;
623 long long ops = server.stat_numcommands - server.ops_sec_last_sample_ops;
624 long long ops_sec;
625
626 ops_sec = t > 0 ? (ops*1000/t) : 0;
627
628 server.ops_sec_samples[server.ops_sec_idx] = ops_sec;
629 server.ops_sec_idx = (server.ops_sec_idx+1) % REDIS_OPS_SEC_SAMPLES;
630 server.ops_sec_last_sample_time = mstime();
631 server.ops_sec_last_sample_ops = server.stat_numcommands;
632}
633
634/* Return the mean of all the samples. */
635long long getOperationsPerSecond(void) {
636 int j;
637 long long sum = 0;
638
639 for (j = 0; j < REDIS_OPS_SEC_SAMPLES; j++)
640 sum += server.ops_sec_samples[j];
641 return sum / REDIS_OPS_SEC_SAMPLES;
642}
643
644/* Check for timeouts. Returns non-zero if the client was terminated */
645int clientsCronHandleTimeout(redisClient *c) {
646 time_t now = server.unixtime;
647
648 if (server.maxidletime &&
649 !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
650 !(c->flags & REDIS_MASTER) && /* no timeout for masters */
651 !(c->flags & REDIS_BLOCKED) && /* no timeout for BLPOP */
652 dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
653 listLength(c->pubsub_patterns) == 0 &&
654 (now - c->lastinteraction > server.maxidletime))
655 {
656 redisLog(REDIS_VERBOSE,"Closing idle client");
657 freeClient(c);
658 return 1;
659 } else if (c->flags & REDIS_BLOCKED) {
660 if (c->bpop.timeout != 0 && c->bpop.timeout < now) {
661 addReply(c,shared.nullmultibulk);
662 unblockClientWaitingData(c);
663 }
664 }
665 return 0;
666}
667
668/* The client query buffer is an sds.c string that can end with a lot of
669 * free space not used, this function reclaims space if needed.
670 *
671 * The funciton always returns 0 as it never terminates the client. */
672int clientsCronResizeQueryBuffer(redisClient *c) {
673 size_t querybuf_size = sdsAllocSize(c->querybuf);
674 time_t idletime = server.unixtime - c->lastinteraction;
675
676 /* There are two conditions to resize the query buffer:
677 * 1) Query buffer is > BIG_ARG and too big for latest peak.
678 * 2) Client is inactive and the buffer is bigger than 1k. */
679 if (((querybuf_size > REDIS_MBULK_BIG_ARG) &&
680 (querybuf_size/(c->querybuf_peak+1)) > 2) ||
681 (querybuf_size > 1024 && idletime > 2))
682 {
683 /* Only resize the query buffer if it is actually wasting space. */
684 if (sdsavail(c->querybuf) > 1024) {
685 c->querybuf = sdsRemoveFreeSpace(c->querybuf);
686 }
687 }
688 /* Reset the peak again to capture the peak memory usage in the next
689 * cycle. */
690 c->querybuf_peak = 0;
691 return 0;
692}
693
694void clientsCron(void) {
695 /* Make sure to process at least 1/100 of clients per call.
696 * Since this function is called 10 times per second we are sure that
697 * in the worst case we process all the clients in 10 seconds.
698 * In normal conditions (a reasonable number of clients) we process
699 * all the clients in a shorter time. */
700 int numclients = listLength(server.clients);
701 int iterations = numclients/100;
702
703 if (iterations < 50)
704 iterations = (numclients < 50) ? numclients : 50;
705 while(listLength(server.clients) && iterations--) {
706 redisClient *c;
707 listNode *head;
708
709 /* Rotate the list, take the current head, process.
710 * This way if the client must be removed from the list it's the
711 * first element and we don't incur into O(N) computation. */
712 listRotate(server.clients);
713 head = listFirst(server.clients);
714 c = listNodeValue(head);
715 /* The following functions do different service checks on the client.
716 * The protocol is that they return non-zero if the client was
717 * terminated. */
718 if (clientsCronHandleTimeout(c)) continue;
719 if (clientsCronResizeQueryBuffer(c)) continue;
720 }
721}
722
723int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
724 int j, loops = server.cronloops;
725 REDIS_NOTUSED(eventLoop);
726 REDIS_NOTUSED(id);
727 REDIS_NOTUSED(clientData);
728
729 /* We take a cached value of the unix time in the global state because
730 * with virtual memory and aging there is to store the current time
731 * in objects at every object access, and accuracy is not needed.
732 * To access a global var is faster than calling time(NULL) */
733 server.unixtime = time(NULL);
734
735 trackOperationsPerSecond();
736
737 /* We have just 22 bits per object for LRU information.
738 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
739 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
740 *
741 * Note that even if this will wrap after 1.5 years it's not a problem,
742 * everything will still work but just some object will appear younger
743 * to Redis. But for this to happen a given object should never be touched
744 * for 1.5 years.
745 *
746 * Note that you can change the resolution altering the
747 * REDIS_LRU_CLOCK_RESOLUTION define.
748 */
749 updateLRUClock();
750
751 /* Record the max memory used since the server was started. */
752 if (zmalloc_used_memory() > server.stat_peak_memory)
753 server.stat_peak_memory = zmalloc_used_memory();
754
755 /* We received a SIGTERM, shutting down here in a safe way, as it is
756 * not ok doing so inside the signal handler. */
757 if (server.shutdown_asap) {
758 if (prepareForShutdown(0) == REDIS_OK) exit(0);
759 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
760 }
761
762 /* Show some info about non-empty databases */
763 for (j = 0; j < server.dbnum; j++) {
764 long long size, used, vkeys;
765
766 size = dictSlots(server.db[j].dict);
767 used = dictSize(server.db[j].dict);
768 vkeys = dictSize(server.db[j].expires);
769 if (!(loops % 50) && (used || vkeys)) {
770 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
771 /* dictPrintStats(server.dict); */
772 }
773 }
774
775 /* We don't want to resize the hash tables while a bacground saving
776 * is in progress: the saving child is created using fork() that is
777 * implemented with a copy-on-write semantic in most modern systems, so
778 * if we resize the HT while there is the saving child at work actually
779 * a lot of memory movements in the parent will cause a lot of pages
780 * copied. */
781 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
782 if (!(loops % 10)) tryResizeHashTables();
783 if (server.activerehashing) incrementallyRehash();
784 }
785
786 /* Show information about connected clients */
787 if (!(loops % 50)) {
788 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
789 listLength(server.clients)-listLength(server.slaves),
790 listLength(server.slaves),
791 zmalloc_used_memory());
792 }
793
794 /* We need to do a few operations on clients asynchronously. */
795 clientsCron();
796
797 /* Start a scheduled AOF rewrite if this was requested by the user while
798 * a BGSAVE was in progress. */
799 if (server.rdb_child_pid == -1 && server.aof_child_pid == -1 &&
800 server.aof_rewrite_scheduled)
801 {
802 rewriteAppendOnlyFileBackground();
803 }
804
805 /* Check if a background saving or AOF rewrite in progress terminated. */
806 if (server.rdb_child_pid != -1 || server.aof_child_pid != -1) {
807 int statloc;
808 pid_t pid;
809
810 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
811 int exitcode = WEXITSTATUS(statloc);
812 int bysignal = 0;
813
814 if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
815
816 if (pid == server.rdb_child_pid) {
817 backgroundSaveDoneHandler(exitcode,bysignal);
818 } else {
819 backgroundRewriteDoneHandler(exitcode,bysignal);
820 }
821 updateDictResizePolicy();
822 }
823 } else {
824 time_t now = time(NULL);
825
826 /* If there is not a background saving/rewrite in progress check if
827 * we have to save/rewrite now */
828 for (j = 0; j < server.saveparamslen; j++) {
829 struct saveparam *sp = server.saveparams+j;
830
831 if (server.dirty >= sp->changes &&
832 now-server.lastsave > sp->seconds) {
833 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
834 sp->changes, sp->seconds);
835 rdbSaveBackground(server.rdb_filename);
836 break;
837 }
838 }
839
840 /* Trigger an AOF rewrite if needed */
841 if (server.rdb_child_pid == -1 &&
842 server.aof_child_pid == -1 &&
843 server.aof_rewrite_perc &&
844 server.aof_current_size > server.aof_rewrite_min_size)
845 {
846 long long base = server.aof_rewrite_base_size ?
847 server.aof_rewrite_base_size : 1;
848 long long growth = (server.aof_current_size*100/base) - 100;
849 if (growth >= server.aof_rewrite_perc) {
850 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
851 rewriteAppendOnlyFileBackground();
852 }
853 }
854 }
855
856
857 /* If we postponed an AOF buffer flush, let's try to do it every time the
858 * cron function is called. */
859 if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
860
861 /* Expire a few keys per cycle, only if this is a master.
862 * On slaves we wait for DEL operations synthesized by the master
863 * in order to guarantee a strict consistency. */
864 if (server.masterhost == NULL) activeExpireCycle();
865
866 /* Close clients that need to be closed asynchronous */
867 freeClientsInAsyncFreeQueue();
868
869 /* Replication cron function -- used to reconnect to master and
870 * to detect transfer failures. */
871 if (!(loops % 10)) replicationCron();
872
873 /* Run other sub-systems specific cron jobs */
874 if (server.cluster_enabled && !(loops % 10)) clusterCron();
875
876 server.cronloops++;
877 return 100;
878}
879
880/* This function gets called every time Redis is entering the
881 * main loop of the event driven library, that is, before to sleep
882 * for ready file descriptors. */
883void beforeSleep(struct aeEventLoop *eventLoop) {
884 REDIS_NOTUSED(eventLoop);
885 listNode *ln;
886 redisClient *c;
887
888 /* Try to process pending commands for clients that were just unblocked. */
889 while (listLength(server.unblocked_clients)) {
890 ln = listFirst(server.unblocked_clients);
891 redisAssert(ln != NULL);
892 c = ln->value;
893 listDelNode(server.unblocked_clients,ln);
894 c->flags &= ~REDIS_UNBLOCKED;
895
896 /* Process remaining data in the input buffer. */
897 if (c->querybuf && sdslen(c->querybuf) > 0) {
898 server.current_client = c;
899 processInputBuffer(c);
900 server.current_client = NULL;
901 }
902 }
903
904 /* Write the AOF buffer on disk */
905 flushAppendOnlyFile(0);
906}
907
908/* =========================== Server initialization ======================== */
909
910void createSharedObjects(void) {
911 int j;
912
913 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
914 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
915 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
916 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
917 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
918 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
919 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
920 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
921 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
922 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
923 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
924 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
925 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
926 "-ERR Operation against a key holding the wrong kind of value\r\n"));
927 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
928 "-ERR no such key\r\n"));
929 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
930 "-ERR syntax error\r\n"));
931 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
932 "-ERR source and destination objects are the same\r\n"));
933 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
934 "-ERR index out of range\r\n"));
935 shared.noscripterr = createObject(REDIS_STRING,sdsnew(
936 "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
937 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
938 "-LOADING Redis is loading the dataset in memory\r\n"));
939 shared.slowscripterr = createObject(REDIS_STRING,sdsnew(
940 "-BUSY Redis is busy running a script. You can only call SCRIPT KILL or SHUTDOWN NOSAVE.\r\n"));
941 shared.bgsaveerr = createObject(REDIS_STRING,sdsnew(
942 "-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"));
943 shared.roslaveerr = createObject(REDIS_STRING,sdsnew(
944 "-READONLY You can't write against a read only slave.\r\n"));
945 shared.oomerr = createObject(REDIS_STRING,sdsnew(
946 "-OOM command not allowed when used memory > 'maxmemory'.\r\n"));
947 shared.space = createObject(REDIS_STRING,sdsnew(" "));
948 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
949 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
950 shared.select0 = createStringObject("select 0\r\n",10);
951 shared.select1 = createStringObject("select 1\r\n",10);
952 shared.select2 = createStringObject("select 2\r\n",10);
953 shared.select3 = createStringObject("select 3\r\n",10);
954 shared.select4 = createStringObject("select 4\r\n",10);
955 shared.select5 = createStringObject("select 5\r\n",10);
956 shared.select6 = createStringObject("select 6\r\n",10);
957 shared.select7 = createStringObject("select 7\r\n",10);
958 shared.select8 = createStringObject("select 8\r\n",10);
959 shared.select9 = createStringObject("select 9\r\n",10);
960 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
961 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
962 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
963 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
964 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
965 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
966 shared.del = createStringObject("DEL",3);
967 shared.rpop = createStringObject("RPOP",4);
968 shared.lpop = createStringObject("LPOP",4);
969 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
970 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
971 shared.integers[j]->encoding = REDIS_ENCODING_INT;
972 }
973 for (j = 0; j < REDIS_SHARED_BULKHDR_LEN; j++) {
974 shared.mbulkhdr[j] = createObject(REDIS_STRING,
975 sdscatprintf(sdsempty(),"*%d\r\n",j));
976 shared.bulkhdr[j] = createObject(REDIS_STRING,
977 sdscatprintf(sdsempty(),"$%d\r\n",j));
978 }
979}
980
981void initServerConfig() {
982 getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
983 server.runid[REDIS_RUN_ID_SIZE] = '\0';
984 server.arch_bits = (sizeof(long) == 8) ? 64 : 32;
985 server.port = REDIS_SERVERPORT;
986 server.bindaddr = NULL;
987 server.unixsocket = NULL;
988 server.unixsocketperm = 0;
989 server.ipfd = -1;
990 server.sofd = -1;
991 server.dbnum = REDIS_DEFAULT_DBNUM;
992 server.verbosity = REDIS_NOTICE;
993 server.maxidletime = REDIS_MAXIDLETIME;
994 server.client_max_querybuf_len = REDIS_MAX_QUERYBUF_LEN;
995 server.saveparams = NULL;
996 server.loading = 0;
997 server.logfile = NULL; /* NULL = log on standard output */
998 server.syslog_enabled = 0;
999 server.syslog_ident = zstrdup("redis");
1000 server.syslog_facility = LOG_LOCAL0;
1001 server.daemonize = 0;
1002 server.aof_state = REDIS_AOF_OFF;
1003 server.aof_fsync = AOF_FSYNC_EVERYSEC;
1004 server.aof_no_fsync_on_rewrite = 0;
1005 server.aof_rewrite_perc = REDIS_AOF_REWRITE_PERC;
1006 server.aof_rewrite_min_size = REDIS_AOF_REWRITE_MIN_SIZE;
1007 server.aof_rewrite_base_size = 0;
1008 server.aof_rewrite_scheduled = 0;
1009 server.aof_last_fsync = time(NULL);
1010 server.aof_delayed_fsync = 0;
1011 server.aof_fd = -1;
1012 server.aof_selected_db = -1; /* Make sure the first time will not match */
1013 server.aof_flush_postponed_start = 0;
1014 server.pidfile = zstrdup("/var/run/redis.pid");
1015 server.rdb_filename = zstrdup("dump.rdb");
1016 server.aof_filename = zstrdup("appendonly.aof");
1017 server.requirepass = NULL;
1018 server.rdb_compression = 1;
1019 server.activerehashing = 1;
1020 server.maxclients = REDIS_MAX_CLIENTS;
1021 server.bpop_blocked_clients = 0;
1022 server.maxmemory = 0;
1023 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
1024 server.maxmemory_samples = 3;
1025 server.hash_max_ziplist_entries = REDIS_HASH_MAX_ZIPLIST_ENTRIES;
1026 server.hash_max_ziplist_value = REDIS_HASH_MAX_ZIPLIST_VALUE;
1027 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
1028 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
1029 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
1030 server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
1031 server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
1032 server.shutdown_asap = 0;
1033 server.repl_ping_slave_period = REDIS_REPL_PING_SLAVE_PERIOD;
1034 server.repl_timeout = REDIS_REPL_TIMEOUT;
1035 server.cluster_enabled = 0;
1036 server.cluster.configfile = zstrdup("nodes.conf");
1037 server.lua_caller = NULL;
1038 server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
1039 server.lua_client = NULL;
1040 server.lua_timedout = 0;
1041
1042 updateLRUClock();
1043 resetServerSaveParams();
1044
1045 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
1046 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
1047 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
1048 /* Replication related */
1049 server.masterauth = NULL;
1050 server.masterhost = NULL;
1051 server.masterport = 6379;
1052 server.master = NULL;
1053 server.repl_state = REDIS_REPL_NONE;
1054 server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
1055 server.repl_serve_stale_data = 1;
1056 server.repl_slave_ro = 1;
1057 server.repl_down_since = -1;
1058
1059 /* Client output buffer limits */
1060 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].hard_limit_bytes = 0;
1061 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_bytes = 0;
1062 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_NORMAL].soft_limit_seconds = 0;
1063 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].hard_limit_bytes = 1024*1024*256;
1064 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_bytes = 1024*1024*64;
1065 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_SLAVE].soft_limit_seconds = 60;
1066 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].hard_limit_bytes = 1024*1024*32;
1067 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_bytes = 1024*1024*8;
1068 server.client_obuf_limits[REDIS_CLIENT_LIMIT_CLASS_PUBSUB].soft_limit_seconds = 60;
1069
1070 /* Double constants initialization */
1071 R_Zero = 0.0;
1072 R_PosInf = 1.0/R_Zero;
1073 R_NegInf = -1.0/R_Zero;
1074 R_Nan = R_Zero/R_Zero;
1075
1076 /* Command table -- we intiialize it here as it is part of the
1077 * initial configuration, since command names may be changed via
1078 * redis.conf using the rename-command directive. */
1079 server.commands = dictCreate(&commandTableDictType,NULL);
1080 populateCommandTable();
1081 server.delCommand = lookupCommandByCString("del");
1082 server.multiCommand = lookupCommandByCString("multi");
1083 server.lpushCommand = lookupCommandByCString("lpush");
1084
1085 /* Slow log */
1086 server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
1087 server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
1088
1089 /* Assert */
1090 server.assert_failed = "<no assertion failed>";
1091 server.assert_file = "<no file>";
1092 server.assert_line = 0;
1093 server.bug_report_start = 0;
1094}
1095
1096/* This function will try to raise the max number of open files accordingly to
1097 * the configured max number of clients. It will also account for 32 additional
1098 * file descriptors as we need a few more for persistence, listening
1099 * sockets, log files and so forth.
1100 *
1101 * If it will not be possible to set the limit accordingly to the configured
1102 * max number of clients, the function will do the reverse setting
1103 * server.maxclients to the value that we can actually handle. */
1104void adjustOpenFilesLimit(void) {
1105 rlim_t maxfiles = server.maxclients+32;
1106 struct rlimit limit;
1107
1108 if (maxfiles < 1024) maxfiles = 1024;
1109 if (getrlimit(RLIMIT_NOFILE,&limit) == -1) {
1110 redisLog(REDIS_WARNING,"Unable to obtain the current NOFILE limit (%s), assuming 1024 and setting the max clients configuration accordingly.",
1111 strerror(errno));
1112 server.maxclients = 1024-32;
1113 } else {
1114 rlim_t oldlimit = limit.rlim_cur;
1115
1116 /* Set the max number of files if the current limit is not enough
1117 * for our needs. */
1118 if (oldlimit < maxfiles) {
1119 limit.rlim_cur = maxfiles;
1120 limit.rlim_max = maxfiles;
1121 if (setrlimit(RLIMIT_NOFILE,&limit) == -1) {
1122 server.maxclients = oldlimit-32;
1123 redisLog(REDIS_WARNING,"Unable to set the max number of files limit to %d (%s), setting the max clients configuration to %d.",
1124 (int) maxfiles, strerror(errno), (int) server.maxclients);
1125 } else {
1126 redisLog(REDIS_NOTICE,"Max number of open files set to %d",
1127 (int) maxfiles);
1128 }
1129 }
1130 }
1131}
1132
1133void initServer() {
1134 int j;
1135
1136 signal(SIGHUP, SIG_IGN);
1137 signal(SIGPIPE, SIG_IGN);
1138 setupSignalHandlers();
1139
1140 if (server.syslog_enabled) {
1141 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
1142 server.syslog_facility);
1143 }
1144
1145 server.current_client = NULL;
1146 server.clients = listCreate();
1147 server.clients_to_close = listCreate();
1148 server.slaves = listCreate();
1149 server.monitors = listCreate();
1150 server.unblocked_clients = listCreate();
1151
1152 createSharedObjects();
1153 adjustOpenFilesLimit();
1154 server.el = aeCreateEventLoop(server.maxclients+1024);
1155 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
1156
1157 if (server.port != 0) {
1158 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
1159 if (server.ipfd == ANET_ERR) {
1160 redisLog(REDIS_WARNING, "Opening port %d: %s",
1161 server.port, server.neterr);
1162 exit(1);
1163 }
1164 }
1165 if (server.unixsocket != NULL) {
1166 unlink(server.unixsocket); /* don't care if this fails */
1167 server.sofd = anetUnixServer(server.neterr,server.unixsocket,server.unixsocketperm);
1168 if (server.sofd == ANET_ERR) {
1169 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
1170 exit(1);
1171 }
1172 }
1173 if (server.ipfd < 0 && server.sofd < 0) {
1174 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
1175 exit(1);
1176 }
1177 for (j = 0; j < server.dbnum; j++) {
1178 server.db[j].dict = dictCreate(&dbDictType,NULL);
1179 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
1180 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
1181 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
1182 server.db[j].id = j;
1183 }
1184 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
1185 server.pubsub_patterns = listCreate();
1186 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
1187 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
1188 server.cronloops = 0;
1189 server.rdb_child_pid = -1;
1190 server.aof_child_pid = -1;
1191 server.aof_rewrite_buf = sdsempty();
1192 server.aof_buf = sdsempty();
1193 server.lastsave = time(NULL);
1194 server.dirty = 0;
1195 server.stat_numcommands = 0;
1196 server.stat_numconnections = 0;
1197 server.stat_expiredkeys = 0;
1198 server.stat_evictedkeys = 0;
1199 server.stat_starttime = time(NULL);
1200 server.stat_keyspace_misses = 0;
1201 server.stat_keyspace_hits = 0;
1202 server.stat_peak_memory = 0;
1203 server.stat_fork_time = 0;
1204 server.stat_rejected_conn = 0;
1205 memset(server.ops_sec_samples,0,sizeof(server.ops_sec_samples));
1206 server.ops_sec_idx = 0;
1207 server.ops_sec_last_sample_time = mstime();
1208 server.ops_sec_last_sample_ops = 0;
1209 server.unixtime = time(NULL);
1210 server.lastbgsave_status = REDIS_OK;
1211 server.stop_writes_on_bgsave_err = 1;
1212 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
1213 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
1214 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
1215 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
1216 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
1217
1218 if (server.aof_state == REDIS_AOF_ON) {
1219 server.aof_fd = open(server.aof_filename,
1220 O_WRONLY|O_APPEND|O_CREAT,0644);
1221 if (server.aof_fd == -1) {
1222 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
1223 strerror(errno));
1224 exit(1);
1225 }
1226 }
1227
1228 /* 32 bit instances are limited to 4GB of address space, so if there is
1229 * no explicit limit in the user provided configuration we set a limit
1230 * at 3.5GB using maxmemory with 'noeviction' policy'. This saves
1231 * useless crashes of the Redis instance. */
1232 if (server.arch_bits == 32 && server.maxmemory == 0) {
1233 redisLog(REDIS_WARNING,"Warning: 32 bit instance detected but no memory limit set. Setting 3.5 GB maxmemory limit with 'noeviction' policy now.");
1234 server.maxmemory = 3584LL*(1024*1024); /* 3584 MB = 3.5 GB */
1235 server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;
1236 }
1237
1238 if (server.cluster_enabled) clusterInit();
1239 scriptingInit();
1240 slowlogInit();
1241 bioInit();
1242}
1243
1244/* Populates the Redis Command Table starting from the hard coded list
1245 * we have on top of redis.c file. */
1246void populateCommandTable(void) {
1247 int j;
1248 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1249
1250 for (j = 0; j < numcommands; j++) {
1251 struct redisCommand *c = redisCommandTable+j;
1252 char *f = c->sflags;
1253 int retval;
1254
1255 while(*f != '\0') {
1256 switch(*f) {
1257 case 'w': c->flags |= REDIS_CMD_WRITE; break;
1258 case 'r': c->flags |= REDIS_CMD_READONLY; break;
1259 case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1260 case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1261 case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1262 case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1263 case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1264 case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1265 case 'S': c->flags |= REDIS_CMD_SORT_FOR_SCRIPT; break;
1266 default: redisPanic("Unsupported command flag"); break;
1267 }
1268 f++;
1269 }
1270
1271 retval = dictAdd(server.commands, sdsnew(c->name), c);
1272 assert(retval == DICT_OK);
1273 }
1274}
1275
1276void resetCommandTableStats(void) {
1277 int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1278 int j;
1279
1280 for (j = 0; j < numcommands; j++) {
1281 struct redisCommand *c = redisCommandTable+j;
1282
1283 c->microseconds = 0;
1284 c->calls = 0;
1285 }
1286}
1287
1288/* ========================== Redis OP Array API ============================ */
1289
1290void redisOpArrayInit(redisOpArray *oa) {
1291 oa->ops = NULL;
1292 oa->numops = 0;
1293}
1294
1295int redisOpArrayAppend(redisOpArray *oa, struct redisCommand *cmd, int dbid,
1296 robj **argv, int argc, int target)
1297{
1298 redisOp *op;
1299
1300 oa->ops = zrealloc(oa->ops,sizeof(redisOp)*(oa->numops+1));
1301 op = oa->ops+oa->numops;
1302 op->cmd = cmd;
1303 op->dbid = dbid;
1304 op->argv = argv;
1305 op->argc = argc;
1306 op->target = target;
1307 oa->numops++;
1308 return oa->numops;
1309}
1310
1311void redisOpArrayFree(redisOpArray *oa) {
1312 while(oa->numops) {
1313 int j;
1314 redisOp *op;
1315
1316 oa->numops--;
1317 op = oa->ops+oa->numops;
1318 for (j = 0; j < op->argc; j++)
1319 decrRefCount(op->argv[j]);
1320 zfree(op->argv);
1321 }
1322 zfree(oa->ops);
1323}
1324
1325/* ====================== Commands lookup and execution ===================== */
1326
1327struct redisCommand *lookupCommand(sds name) {
1328 return dictFetchValue(server.commands, name);
1329}
1330
1331struct redisCommand *lookupCommandByCString(char *s) {
1332 struct redisCommand *cmd;
1333 sds name = sdsnew(s);
1334
1335 cmd = dictFetchValue(server.commands, name);
1336 sdsfree(name);
1337 return cmd;
1338}
1339
1340/* Propagate the specified command (in the context of the specified database id)
1341 * to AOF, Slaves and Monitors.
1342 *
1343 * flags are an xor between:
1344 * + REDIS_PROPAGATE_NONE (no propagation of command at all)
1345 * + REDIS_PROPAGATE_AOF (propagate into the AOF file if is enabled)
1346 * + REDIS_PROPAGATE_REPL (propagate into the replication link)
1347 */
1348void propagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1349 int flags)
1350{
1351 if (server.aof_state != REDIS_AOF_OFF && flags & REDIS_PROPAGATE_AOF)
1352 feedAppendOnlyFile(cmd,dbid,argv,argc);
1353 if (flags & REDIS_PROPAGATE_REPL && listLength(server.slaves))
1354 replicationFeedSlaves(server.slaves,dbid,argv,argc);
1355}
1356
1357/* Used inside commands to schedule the propagation of additional commands
1358 * after the current command is propagated to AOF / Replication. */
1359void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
1360 int target)
1361{
1362 redisOpArrayAppend(&server.also_propagate,cmd,dbid,argv,argc,target);
1363}
1364
1365/* Call() is the core of Redis execution of a command */
1366void call(redisClient *c, int flags) {
1367 long long dirty, start = ustime(), duration;
1368
1369 /* Sent the command to clients in MONITOR mode, only if the commands are
1370 * not geneated from reading an AOF. */
1371 if (listLength(server.monitors) && !server.loading)
1372 replicationFeedMonitors(c,server.monitors,c->db->id,c->argv,c->argc);
1373
1374 /* Call the command. */
1375 redisOpArrayInit(&server.also_propagate);
1376 dirty = server.dirty;
1377 c->cmd->proc(c);
1378 dirty = server.dirty-dirty;
1379 duration = ustime()-start;
1380
1381 /* When EVAL is called loading the AOF we don't want commands called
1382 * from Lua to go into the slowlog or to populate statistics. */
1383 if (server.loading && c->flags & REDIS_LUA_CLIENT)
1384 flags &= ~(REDIS_CALL_SLOWLOG | REDIS_CALL_STATS);
1385
1386 /* Log the command into the Slow log if needed, and populate the
1387 * per-command statistics that we show in INFO commandstats. */
1388 if (flags & REDIS_CALL_SLOWLOG)
1389 slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1390 if (flags & REDIS_CALL_STATS) {
1391 c->cmd->microseconds += duration;
1392 c->cmd->calls++;
1393 }
1394
1395 /* Propagate the command into the AOF and replication link */
1396 if (flags & REDIS_CALL_PROPAGATE) {
1397 int flags = REDIS_PROPAGATE_NONE;
1398
1399 if (c->cmd->flags & REDIS_CMD_FORCE_REPLICATION)
1400 flags |= REDIS_PROPAGATE_REPL;
1401 if (dirty)
1402 flags |= (REDIS_PROPAGATE_REPL | REDIS_PROPAGATE_AOF);
1403 if (flags != REDIS_PROPAGATE_NONE)
1404 propagate(c->cmd,c->db->id,c->argv,c->argc,flags);
1405 }
1406 /* Commands such as LPUSH or BRPOPLPUSH may propagate an additional
1407 * PUSH command. */
1408 if (server.also_propagate.numops) {
1409 int j;
1410 redisOp *rop;
1411
1412 for (j = 0; j < server.also_propagate.numops; j++) {
1413 rop = &server.also_propagate.ops[j];
1414 propagate(rop->cmd, rop->dbid, rop->argv, rop->argc, rop->target);
1415 }
1416 redisOpArrayFree(&server.also_propagate);
1417 }
1418 server.stat_numcommands++;
1419}
1420
1421/* If this function gets called we already read a whole
1422 * command, argments are in the client argv/argc fields.
1423 * processCommand() execute the command or prepare the
1424 * server for a bulk read from the client.
1425 *
1426 * If 1 is returned the client is still alive and valid and
1427 * and other operations can be performed by the caller. Otherwise
1428 * if 0 is returned the client was destroied (i.e. after QUIT). */
1429int processCommand(redisClient *c) {
1430 /* The QUIT command is handled separately. Normal command procs will
1431 * go through checking for replication and QUIT will cause trouble
1432 * when FORCE_REPLICATION is enabled and would be implemented in
1433 * a regular command proc. */
1434 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1435 addReply(c,shared.ok);
1436 c->flags |= REDIS_CLOSE_AFTER_REPLY;
1437 return REDIS_ERR;
1438 }
1439
1440 /* Now lookup the command and check ASAP about trivial error conditions
1441 * such as wrong arity, bad command name and so forth. */
1442 c->cmd = c->lastcmd = lookupCommand(c->argv[0]->ptr);
1443 if (!c->cmd) {
1444 addReplyErrorFormat(c,"unknown command '%s'",
1445 (char*)c->argv[0]->ptr);
1446 return REDIS_OK;
1447 } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1448 (c->argc < -c->cmd->arity)) {
1449 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1450 c->cmd->name);
1451 return REDIS_OK;
1452 }
1453
1454 /* Check if the user is authenticated */
1455 if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1456 {
1457 addReplyError(c,"operation not permitted");
1458 return REDIS_OK;
1459 }
1460
1461 /* If cluster is enabled, redirect here */
1462 if (server.cluster_enabled &&
1463 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1464 int hashslot;
1465
1466 if (server.cluster.state != REDIS_CLUSTER_OK) {
1467 addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1468 return REDIS_OK;
1469 } else {
1470 int ask;
1471 clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1472 if (n == NULL) {
1473 addReplyError(c,"Multi keys request invalid in cluster");
1474 return REDIS_OK;
1475 } else if (n != server.cluster.myself) {
1476 addReplySds(c,sdscatprintf(sdsempty(),
1477 "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1478 hashslot,n->ip,n->port));
1479 return REDIS_OK;
1480 }
1481 }
1482 }
1483
1484 /* Handle the maxmemory directive.
1485 *
1486 * First we try to free some memory if possible (if there are volatile
1487 * keys in the dataset). If there are not the only thing we can do
1488 * is returning an error. */
1489 if (server.maxmemory) {
1490 int retval = freeMemoryIfNeeded();
1491 if ((c->cmd->flags & REDIS_CMD_DENYOOM) && retval == REDIS_ERR) {
1492 addReply(c, shared.oomerr);
1493 return REDIS_OK;
1494 }
1495 }
1496
1497 /* Don't accept write commands if there are problems persisting on disk. */
1498 if (server.stop_writes_on_bgsave_err &&
1499 server.saveparamslen > 0
1500 && server.lastbgsave_status == REDIS_ERR &&
1501 c->cmd->flags & REDIS_CMD_WRITE)
1502 {
1503 addReply(c, shared.bgsaveerr);
1504 return REDIS_OK;
1505 }
1506
1507 /* Don't accept wirte commands if this is a read only slave. But
1508 * accept write commands if this is our master. */
1509 if (server.masterhost && server.repl_slave_ro &&
1510 !(c->flags & REDIS_MASTER) &&
1511 c->cmd->flags & REDIS_CMD_WRITE)
1512 {
1513 addReply(c, shared.roslaveerr);
1514 return REDIS_OK;
1515 }
1516
1517 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1518 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1519 &&
1520 c->cmd->proc != subscribeCommand &&
1521 c->cmd->proc != unsubscribeCommand &&
1522 c->cmd->proc != psubscribeCommand &&
1523 c->cmd->proc != punsubscribeCommand) {
1524 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1525 return REDIS_OK;
1526 }
1527
1528 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1529 * we are a slave with a broken link with master. */
1530 if (server.masterhost && server.repl_state != REDIS_REPL_CONNECTED &&
1531 server.repl_serve_stale_data == 0 &&
1532 c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1533 {
1534 addReplyError(c,
1535 "link with MASTER is down and slave-serve-stale-data is set to no");
1536 return REDIS_OK;
1537 }
1538
1539 /* Loading DB? Return an error if the command is not INFO */
1540 if (server.loading && c->cmd->proc != infoCommand) {
1541 addReply(c, shared.loadingerr);
1542 return REDIS_OK;
1543 }
1544
1545 /* Lua script too slow? Only allow SHUTDOWN NOSAVE and SCRIPT KILL. */
1546 if (server.lua_timedout &&
1547 !(c->cmd->proc != shutdownCommand &&
1548 c->argc == 2 &&
1549 tolower(((char*)c->argv[1]->ptr)[0]) == 'n') &&
1550 !(c->cmd->proc == scriptCommand &&
1551 c->argc == 2 &&
1552 tolower(((char*)c->argv[1]->ptr)[0]) == 'k'))
1553 {
1554 addReply(c, shared.slowscripterr);
1555 return REDIS_OK;
1556 }
1557
1558 /* Exec the command */
1559 if (c->flags & REDIS_MULTI &&
1560 c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1561 c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1562 {
1563 queueMultiCommand(c);
1564 addReply(c,shared.queued);
1565 } else {
1566 call(c,REDIS_CALL_FULL);
1567 }
1568 return REDIS_OK;
1569}
1570
1571/*================================== Shutdown =============================== */
1572
1573int prepareForShutdown(int flags) {
1574 int save = flags & REDIS_SHUTDOWN_SAVE;
1575 int nosave = flags & REDIS_SHUTDOWN_NOSAVE;
1576
1577 redisLog(REDIS_WARNING,"User requested shutdown...");
1578 /* Kill the saving child if there is a background saving in progress.
1579 We want to avoid race conditions, for instance our saving child may
1580 overwrite the synchronous saving did by SHUTDOWN. */
1581 if (server.rdb_child_pid != -1) {
1582 redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1583 kill(server.rdb_child_pid,SIGKILL);
1584 rdbRemoveTempFile(server.rdb_child_pid);
1585 }
1586 if (server.aof_state != REDIS_AOF_OFF) {
1587 /* Kill the AOF saving child as the AOF we already have may be longer
1588 * but contains the full dataset anyway. */
1589 if (server.aof_child_pid != -1) {
1590 redisLog(REDIS_WARNING,
1591 "There is a child rewriting the AOF. Killing it!");
1592 kill(server.aof_child_pid,SIGKILL);
1593 }
1594 /* Append only file: fsync() the AOF and exit */
1595 redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1596 aof_fsync(server.aof_fd);
1597 }
1598 if ((server.saveparamslen > 0 && !nosave) || save) {
1599 redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1600 /* Snapshotting. Perform a SYNC SAVE and exit */
1601 if (rdbSave(server.rdb_filename) != REDIS_OK) {
1602 /* Ooops.. error saving! The best we can do is to continue
1603 * operating. Note that if there was a background saving process,
1604 * in the next cron() Redis will be notified that the background
1605 * saving aborted, handling special stuff like slaves pending for
1606 * synchronization... */
1607 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1608 return REDIS_ERR;
1609 }
1610 }
1611 if (server.daemonize) {
1612 redisLog(REDIS_NOTICE,"Removing the pid file.");
1613 unlink(server.pidfile);
1614 }
1615 /* Close the listening sockets. Apparently this allows faster restarts. */
1616 if (server.ipfd != -1) close(server.ipfd);
1617 if (server.sofd != -1) close(server.sofd);
1618 if (server.unixsocket) {
1619 redisLog(REDIS_NOTICE,"Removing the unix socket file.");
1620 unlink(server.unixsocket); /* don't care if this fails */
1621 }
1622
1623 redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1624 return REDIS_OK;
1625}
1626
1627/*================================== Commands =============================== */
1628
1629void authCommand(redisClient *c) {
1630 if (!server.requirepass) {
1631 addReplyError(c,"Client sent AUTH, but no password is set");
1632 } else if (!strcmp(c->argv[1]->ptr, server.requirepass)) {
1633 c->authenticated = 1;
1634 addReply(c,shared.ok);
1635 } else {
1636 c->authenticated = 0;
1637 addReplyError(c,"invalid password");
1638 }
1639}
1640
1641void pingCommand(redisClient *c) {
1642 addReply(c,shared.pong);
1643}
1644
1645void echoCommand(redisClient *c) {
1646 addReplyBulk(c,c->argv[1]);
1647}
1648
1649void timeCommand(redisClient *c) {
1650 struct timeval tv;
1651
1652 /* gettimeofday() can only fail if &tv is a bad addresss so we
1653 * don't check for errors. */
1654 gettimeofday(&tv,NULL);
1655 addReplyMultiBulkLen(c,2);
1656 addReplyBulkLongLong(c,tv.tv_sec);
1657 addReplyBulkLongLong(c,tv.tv_usec);
1658}
1659
1660/* Convert an amount of bytes into a human readable string in the form
1661 * of 100B, 2G, 100M, 4K, and so forth. */
1662void bytesToHuman(char *s, unsigned long long n) {
1663 double d;
1664
1665 if (n < 1024) {
1666 /* Bytes */
1667 sprintf(s,"%lluB",n);
1668 return;
1669 } else if (n < (1024*1024)) {
1670 d = (double)n/(1024);
1671 sprintf(s,"%.2fK",d);
1672 } else if (n < (1024LL*1024*1024)) {
1673 d = (double)n/(1024*1024);
1674 sprintf(s,"%.2fM",d);
1675 } else if (n < (1024LL*1024*1024*1024)) {
1676 d = (double)n/(1024LL*1024*1024);
1677 sprintf(s,"%.2fG",d);
1678 }
1679}
1680
1681/* Create the string returned by the INFO command. This is decoupled
1682 * by the INFO command itself as we need to report the same information
1683 * on memory corruption problems. */
1684sds genRedisInfoString(char *section) {
1685 sds info = sdsempty();
1686 time_t uptime = time(NULL)-server.stat_starttime;
1687 int j, numcommands;
1688 struct rusage self_ru, c_ru;
1689 unsigned long lol, bib;
1690 int allsections = 0, defsections = 0;
1691 int sections = 0;
1692
1693 if (section) {
1694 allsections = strcasecmp(section,"all") == 0;
1695 defsections = strcasecmp(section,"default") == 0;
1696 }
1697
1698 getrusage(RUSAGE_SELF, &self_ru);
1699 getrusage(RUSAGE_CHILDREN, &c_ru);
1700 getClientsMaxBuffers(&lol,&bib);
1701
1702 /* Server */
1703 if (allsections || defsections || !strcasecmp(section,"server")) {
1704 if (sections++) info = sdscat(info,"\r\n");
1705 info = sdscatprintf(info,
1706 "# Server\r\n"
1707 "redis_version:%s\r\n"
1708 "redis_git_sha1:%s\r\n"
1709 "redis_git_dirty:%d\r\n"
1710 "arch_bits:%d\r\n"
1711 "multiplexing_api:%s\r\n"
1712 "gcc_version:%d.%d.%d\r\n"
1713 "process_id:%ld\r\n"
1714 "run_id:%s\r\n"
1715 "tcp_port:%d\r\n"
1716 "uptime_in_seconds:%ld\r\n"
1717 "uptime_in_days:%ld\r\n"
1718 "lru_clock:%ld\r\n",
1719 REDIS_VERSION,
1720 redisGitSHA1(),
1721 strtol(redisGitDirty(),NULL,10) > 0,
1722 server.arch_bits,
1723 aeGetApiName(),
1724#ifdef __GNUC__
1725 __GNUC__,__GNUC_MINOR__,__GNUC_PATCHLEVEL__,
1726#else
1727 0,0,0,
1728#endif
1729 (long) getpid(),
1730 server.runid,
1731 server.port,
1732 uptime,
1733 uptime/(3600*24),
1734 (unsigned long) server.lruclock);
1735 }
1736
1737 /* Clients */
1738 if (allsections || defsections || !strcasecmp(section,"clients")) {
1739 if (sections++) info = sdscat(info,"\r\n");
1740 info = sdscatprintf(info,
1741 "# Clients\r\n"
1742 "connected_clients:%lu\r\n"
1743 "client_longest_output_list:%lu\r\n"
1744 "client_biggest_input_buf:%lu\r\n"
1745 "blocked_clients:%d\r\n",
1746 listLength(server.clients)-listLength(server.slaves),
1747 lol, bib,
1748 server.bpop_blocked_clients);
1749 }
1750
1751 /* Memory */
1752 if (allsections || defsections || !strcasecmp(section,"memory")) {
1753 char hmem[64];
1754 char peak_hmem[64];
1755
1756 bytesToHuman(hmem,zmalloc_used_memory());
1757 bytesToHuman(peak_hmem,server.stat_peak_memory);
1758 if (sections++) info = sdscat(info,"\r\n");
1759 info = sdscatprintf(info,
1760 "# Memory\r\n"
1761 "used_memory:%zu\r\n"
1762 "used_memory_human:%s\r\n"
1763 "used_memory_rss:%zu\r\n"
1764 "used_memory_peak:%zu\r\n"
1765 "used_memory_peak_human:%s\r\n"
1766 "used_memory_lua:%lld\r\n"
1767 "mem_fragmentation_ratio:%.2f\r\n"
1768 "mem_allocator:%s\r\n",
1769 zmalloc_used_memory(),
1770 hmem,
1771 zmalloc_get_rss(),
1772 server.stat_peak_memory,
1773 peak_hmem,
1774 ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1775 zmalloc_get_fragmentation_ratio(),
1776 ZMALLOC_LIB
1777 );
1778 }
1779
1780 /* Persistence */
1781 if (allsections || defsections || !strcasecmp(section,"persistence")) {
1782 if (sections++) info = sdscat(info,"\r\n");
1783 info = sdscatprintf(info,
1784 "# Persistence\r\n"
1785 "loading:%d\r\n"
1786 "aof_enabled:%d\r\n"
1787 "changes_since_last_save:%lld\r\n"
1788 "bgsave_in_progress:%d\r\n"
1789 "last_save_time:%ld\r\n"
1790 "last_bgsave_status:%s\r\n"
1791 "bgrewriteaof_in_progress:%d\r\n",
1792 server.loading,
1793 server.aof_state != REDIS_AOF_OFF,
1794 server.dirty,
1795 server.rdb_child_pid != -1,
1796 server.lastsave,
1797 server.lastbgsave_status == REDIS_OK ? "ok" : "err",
1798 server.aof_child_pid != -1);
1799
1800 if (server.aof_state != REDIS_AOF_OFF) {
1801 info = sdscatprintf(info,
1802 "aof_current_size:%lld\r\n"
1803 "aof_base_size:%lld\r\n"
1804 "aof_pending_rewrite:%d\r\n"
1805 "aof_buffer_length:%zu\r\n"
1806 "aof_pending_bio_fsync:%llu\r\n"
1807 "aof_delayed_fsync:%lu\r\n",
1808 (long long) server.aof_current_size,
1809 (long long) server.aof_rewrite_base_size,
1810 server.aof_rewrite_scheduled,
1811 sdslen(server.aof_buf),
1812 bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC),
1813 server.aof_delayed_fsync);
1814 }
1815
1816 if (server.loading) {
1817 double perc;
1818 time_t eta, elapsed;
1819 off_t remaining_bytes = server.loading_total_bytes-
1820 server.loading_loaded_bytes;
1821
1822 perc = ((double)server.loading_loaded_bytes /
1823 server.loading_total_bytes) * 100;
1824
1825 elapsed = time(NULL)-server.loading_start_time;
1826 if (elapsed == 0) {
1827 eta = 1; /* A fake 1 second figure if we don't have
1828 enough info */
1829 } else {
1830 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1831 }
1832
1833 info = sdscatprintf(info,
1834 "loading_start_time:%ld\r\n"
1835 "loading_total_bytes:%llu\r\n"
1836 "loading_loaded_bytes:%llu\r\n"
1837 "loading_loaded_perc:%.2f\r\n"
1838 "loading_eta_seconds:%ld\r\n"
1839 ,(unsigned long) server.loading_start_time,
1840 (unsigned long long) server.loading_total_bytes,
1841 (unsigned long long) server.loading_loaded_bytes,
1842 perc,
1843 eta
1844 );
1845 }
1846 }
1847
1848 /* Stats */
1849 if (allsections || defsections || !strcasecmp(section,"stats")) {
1850 if (sections++) info = sdscat(info,"\r\n");
1851 info = sdscatprintf(info,
1852 "# Stats\r\n"
1853 "total_connections_received:%lld\r\n"
1854 "total_commands_processed:%lld\r\n"
1855 "instantaneous_ops_per_sec:%lld\r\n"
1856 "rejected_connections:%lld\r\n"
1857 "expired_keys:%lld\r\n"
1858 "evicted_keys:%lld\r\n"
1859 "keyspace_hits:%lld\r\n"
1860 "keyspace_misses:%lld\r\n"
1861 "pubsub_channels:%ld\r\n"
1862 "pubsub_patterns:%lu\r\n"
1863 "latest_fork_usec:%lld\r\n",
1864 server.stat_numconnections,
1865 server.stat_numcommands,
1866 getOperationsPerSecond(),
1867 server.stat_rejected_conn,
1868 server.stat_expiredkeys,
1869 server.stat_evictedkeys,
1870 server.stat_keyspace_hits,
1871 server.stat_keyspace_misses,
1872 dictSize(server.pubsub_channels),
1873 listLength(server.pubsub_patterns),
1874 server.stat_fork_time);
1875 }
1876
1877 /* Replication */
1878 if (allsections || defsections || !strcasecmp(section,"replication")) {
1879 if (sections++) info = sdscat(info,"\r\n");
1880 info = sdscatprintf(info,
1881 "# Replication\r\n"
1882 "role:%s\r\n",
1883 server.masterhost == NULL ? "master" : "slave");
1884 if (server.masterhost) {
1885 info = sdscatprintf(info,
1886 "master_host:%s\r\n"
1887 "master_port:%d\r\n"
1888 "master_link_status:%s\r\n"
1889 "master_last_io_seconds_ago:%d\r\n"
1890 "master_sync_in_progress:%d\r\n"
1891 ,server.masterhost,
1892 server.masterport,
1893 (server.repl_state == REDIS_REPL_CONNECTED) ?
1894 "up" : "down",
1895 server.master ?
1896 ((int)(time(NULL)-server.master->lastinteraction)) : -1,
1897 server.repl_state == REDIS_REPL_TRANSFER
1898 );
1899
1900 if (server.repl_state == REDIS_REPL_TRANSFER) {
1901 info = sdscatprintf(info,
1902 "master_sync_left_bytes:%ld\r\n"
1903 "master_sync_last_io_seconds_ago:%d\r\n"
1904 ,(long)server.repl_transfer_left,
1905 (int)(time(NULL)-server.repl_transfer_lastio)
1906 );
1907 }
1908
1909 if (server.repl_state != REDIS_REPL_CONNECTED) {
1910 info = sdscatprintf(info,
1911 "master_link_down_since_seconds:%ld\r\n",
1912 (long)time(NULL)-server.repl_down_since);
1913 }
1914 }
1915 info = sdscatprintf(info,
1916 "connected_slaves:%lu\r\n",
1917 listLength(server.slaves));
1918 if (listLength(server.slaves)) {
1919 int slaveid = 0;
1920 listNode *ln;
1921 listIter li;
1922
1923 listRewind(server.slaves,&li);
1924 while((ln = listNext(&li))) {
1925 redisClient *slave = listNodeValue(ln);
1926 char *state = NULL;
1927 char ip[32];
1928 int port;
1929
1930 if (anetPeerToString(slave->fd,ip,&port) == -1) continue;
1931 switch(slave->replstate) {
1932 case REDIS_REPL_WAIT_BGSAVE_START:
1933 case REDIS_REPL_WAIT_BGSAVE_END:
1934 state = "wait_bgsave";
1935 break;
1936 case REDIS_REPL_SEND_BULK:
1937 state = "send_bulk";
1938 break;
1939 case REDIS_REPL_ONLINE:
1940 state = "online";
1941 break;
1942 }
1943 if (state == NULL) continue;
1944 info = sdscatprintf(info,"slave%d:%s,%d,%s\r\n",
1945 slaveid,ip,port,state);
1946 slaveid++;
1947 }
1948 }
1949 }
1950
1951 /* CPU */
1952 if (allsections || defsections || !strcasecmp(section,"cpu")) {
1953 if (sections++) info = sdscat(info,"\r\n");
1954 info = sdscatprintf(info,
1955 "# CPU\r\n"
1956 "used_cpu_sys:%.2f\r\n"
1957 "used_cpu_user:%.2f\r\n"
1958 "used_cpu_sys_children:%.2f\r\n"
1959 "used_cpu_user_children:%.2f\r\n",
1960 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
1961 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
1962 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
1963 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000);
1964 }
1965
1966 /* cmdtime */
1967 if (allsections || !strcasecmp(section,"commandstats")) {
1968 if (sections++) info = sdscat(info,"\r\n");
1969 info = sdscatprintf(info, "# Commandstats\r\n");
1970 numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1971 for (j = 0; j < numcommands; j++) {
1972 struct redisCommand *c = redisCommandTable+j;
1973
1974 if (!c->calls) continue;
1975 info = sdscatprintf(info,
1976 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
1977 c->name, c->calls, c->microseconds,
1978 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
1979 }
1980 }
1981
1982 /* Clusetr */
1983 if (allsections || defsections || !strcasecmp(section,"cluster")) {
1984 if (sections++) info = sdscat(info,"\r\n");
1985 info = sdscatprintf(info,
1986 "# Cluster\r\n"
1987 "cluster_enabled:%d\r\n",
1988 server.cluster_enabled);
1989 }
1990
1991 /* Key space */
1992 if (allsections || defsections || !strcasecmp(section,"keyspace")) {
1993 if (sections++) info = sdscat(info,"\r\n");
1994 info = sdscatprintf(info, "# Keyspace\r\n");
1995 for (j = 0; j < server.dbnum; j++) {
1996 long long keys, vkeys;
1997
1998 keys = dictSize(server.db[j].dict);
1999 vkeys = dictSize(server.db[j].expires);
2000 if (keys || vkeys) {
2001 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
2002 j, keys, vkeys);
2003 }
2004 }
2005 }
2006 return info;
2007}
2008
2009void infoCommand(redisClient *c) {
2010 char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
2011
2012 if (c->argc > 2) {
2013 addReply(c,shared.syntaxerr);
2014 return;
2015 }
2016 sds info = genRedisInfoString(section);
2017 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
2018 (unsigned long)sdslen(info)));
2019 addReplySds(c,info);
2020 addReply(c,shared.crlf);
2021}
2022
2023void monitorCommand(redisClient *c) {
2024 /* ignore MONITOR if aleady slave or in monitor mode */
2025 if (c->flags & REDIS_SLAVE) return;
2026
2027 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
2028 c->slaveseldb = 0;
2029 listAddNodeTail(server.monitors,c);
2030 addReply(c,shared.ok);
2031}
2032
2033/* ============================ Maxmemory directive ======================== */
2034
2035/* This function gets called when 'maxmemory' is set on the config file to limit
2036 * the max memory used by the server, before processing a command.
2037 *
2038 * The goal of the function is to free enough memory to keep Redis under the
2039 * configured memory limit.
2040 *
2041 * The function starts calculating how many bytes should be freed to keep
2042 * Redis under the limit, and enters a loop selecting the best keys to
2043 * evict accordingly to the configured policy.
2044 *
2045 * If all the bytes needed to return back under the limit were freed the
2046 * function returns REDIS_OK, otherwise REDIS_ERR is returned, and the caller
2047 * should block the execution of commands that will result in more memory
2048 * used by the server.
2049 */
2050int freeMemoryIfNeeded(void) {
2051 size_t mem_used, mem_tofree, mem_freed;
2052 int slaves = listLength(server.slaves);
2053
2054 /* Remove the size of slaves output buffers and AOF buffer from the
2055 * count of used memory. */
2056 mem_used = zmalloc_used_memory();
2057 if (slaves) {
2058 listIter li;
2059 listNode *ln;
2060
2061 listRewind(server.slaves,&li);
2062 while((ln = listNext(&li))) {
2063 redisClient *slave = listNodeValue(ln);
2064 unsigned long obuf_bytes = getClientOutputBufferMemoryUsage(slave);
2065 if (obuf_bytes > mem_used)
2066 mem_used = 0;
2067 else
2068 mem_used -= obuf_bytes;
2069 }
2070 }
2071 if (server.aof_state != REDIS_AOF_OFF) {
2072 mem_used -= sdslen(server.aof_buf);
2073 mem_used -= sdslen(server.aof_rewrite_buf);
2074 }
2075
2076 /* Check if we are over the memory limit. */
2077 if (mem_used <= server.maxmemory) return REDIS_OK;
2078
2079 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION)
2080 return REDIS_ERR; /* We need to free memory, but policy forbids. */
2081
2082 /* Compute how much memory we need to free. */
2083 mem_tofree = mem_used - server.maxmemory;
2084 mem_freed = 0;
2085 while (mem_freed < mem_tofree) {
2086 int j, k, keys_freed = 0;
2087
2088 for (j = 0; j < server.dbnum; j++) {
2089 long bestval = 0; /* just to prevent warning */
2090 sds bestkey = NULL;
2091 struct dictEntry *de;
2092 redisDb *db = server.db+j;
2093 dict *dict;
2094
2095 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2096 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
2097 {
2098 dict = server.db[j].dict;
2099 } else {
2100 dict = server.db[j].expires;
2101 }
2102 if (dictSize(dict) == 0) continue;
2103
2104 /* volatile-random and allkeys-random policy */
2105 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
2106 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
2107 {
2108 de = dictGetRandomKey(dict);
2109 bestkey = dictGetKey(de);
2110 }
2111
2112 /* volatile-lru and allkeys-lru policy */
2113 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
2114 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2115 {
2116 for (k = 0; k < server.maxmemory_samples; k++) {
2117 sds thiskey;
2118 long thisval;
2119 robj *o;
2120
2121 de = dictGetRandomKey(dict);
2122 thiskey = dictGetKey(de);
2123 /* When policy is volatile-lru we need an additonal lookup
2124 * to locate the real key, as dict is set to db->expires. */
2125 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
2126 de = dictFind(db->dict, thiskey);
2127 o = dictGetVal(de);
2128 thisval = estimateObjectIdleTime(o);
2129
2130 /* Higher idle time is better candidate for deletion */
2131 if (bestkey == NULL || thisval > bestval) {
2132 bestkey = thiskey;
2133 bestval = thisval;
2134 }
2135 }
2136 }
2137
2138 /* volatile-ttl */
2139 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
2140 for (k = 0; k < server.maxmemory_samples; k++) {
2141 sds thiskey;
2142 long thisval;
2143
2144 de = dictGetRandomKey(dict);
2145 thiskey = dictGetKey(de);
2146 thisval = (long) dictGetVal(de);
2147
2148 /* Expire sooner (minor expire unix timestamp) is better
2149 * candidate for deletion */
2150 if (bestkey == NULL || thisval < bestval) {
2151 bestkey = thiskey;
2152 bestval = thisval;
2153 }
2154 }
2155 }
2156
2157 /* Finally remove the selected key. */
2158 if (bestkey) {
2159 long long delta;
2160
2161 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
2162 propagateExpire(db,keyobj);
2163 /* We compute the amount of memory freed by dbDelete() alone.
2164 * It is possible that actually the memory needed to propagate
2165 * the DEL in AOF and replication link is greater than the one
2166 * we are freeing removing the key, but we can't account for
2167 * that otherwise we would never exit the loop.
2168 *
2169 * AOF and Output buffer memory will be freed eventually so
2170 * we only care about memory used by the key space. */
2171 delta = (long long) zmalloc_used_memory();
2172 dbDelete(db,keyobj);
2173 delta -= (long long) zmalloc_used_memory();
2174 mem_freed += delta;
2175 server.stat_evictedkeys++;
2176 decrRefCount(keyobj);
2177 keys_freed++;
2178
2179 /* When the memory to free starts to be big enough, we may
2180 * start spending so much time here that is impossible to
2181 * deliver data to the slaves fast enough, so we force the
2182 * transmission here inside the loop. */
2183 if (slaves) flushSlavesOutputBuffers();
2184 }
2185 }
2186 if (!keys_freed) return REDIS_ERR; /* nothing to free... */
2187 }
2188 return REDIS_OK;
2189}
2190
2191/* =================================== Main! ================================ */
2192
2193#ifdef __linux__
2194int linuxOvercommitMemoryValue(void) {
2195 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
2196 char buf[64];
2197
2198 if (!fp) return -1;
2199 if (fgets(buf,64,fp) == NULL) {
2200 fclose(fp);
2201 return -1;
2202 }
2203 fclose(fp);
2204
2205 return atoi(buf);
2206}
2207
2208void linuxOvercommitMemoryWarning(void) {
2209 if (linuxOvercommitMemoryValue() == 0) {
2210 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.");
2211 }
2212}
2213#endif /* __linux__ */
2214
2215void createPidFile(void) {
2216 /* Try to write the pid file in a best-effort way. */
2217 FILE *fp = fopen(server.pidfile,"w");
2218 if (fp) {
2219 fprintf(fp,"%d\n",(int)getpid());
2220 fclose(fp);
2221 }
2222}
2223
2224void daemonize(void) {
2225 int fd;
2226
2227 if (fork() != 0) exit(0); /* parent exits */
2228 setsid(); /* create a new session */
2229
2230 /* Every output goes to /dev/null. If Redis is daemonized but
2231 * the 'logfile' is set to 'stdout' in the configuration file
2232 * it will not log at all. */
2233 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
2234 dup2(fd, STDIN_FILENO);
2235 dup2(fd, STDOUT_FILENO);
2236 dup2(fd, STDERR_FILENO);
2237 if (fd > STDERR_FILENO) close(fd);
2238 }
2239}
2240
2241void version() {
2242 printf("Redis server v=%s sha=%s:%d malloc=%s\n", REDIS_VERSION,
2243 redisGitSHA1(), atoi(redisGitDirty()) > 0, ZMALLOC_LIB);
2244 exit(0);
2245}
2246
2247void usage() {
2248 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
2249 fprintf(stderr," ./redis-server - (read config from stdin)\n");
2250 fprintf(stderr," ./redis-server -v or --version\n");
2251 fprintf(stderr," ./redis-server -h or --help\n");
2252 fprintf(stderr," ./redis-server --test-memory <megabytes>\n\n");
2253 fprintf(stderr,"Examples:\n");
2254 fprintf(stderr," ./redis-server (run the server with default conf)\n");
2255 fprintf(stderr," ./redis-server /etc/redis/6379.conf\n");
2256 fprintf(stderr," ./redis-server --port 7777\n");
2257 fprintf(stderr," ./redis-server --port 7777 --slaveof 127.0.0.1 8888\n");
2258 fprintf(stderr," ./redis-server /etc/myredis.conf --loglevel verbose\n");
2259 exit(1);
2260}
2261
2262void redisAsciiArt(void) {
2263#include "asciilogo.h"
2264 char *buf = zmalloc(1024*16);
2265
2266 snprintf(buf,1024*16,ascii_logo,
2267 REDIS_VERSION,
2268 redisGitSHA1(),
2269 strtol(redisGitDirty(),NULL,10) > 0,
2270 (sizeof(long) == 8) ? "64" : "32",
2271 server.cluster_enabled ? "cluster" : "stand alone",
2272 server.port,
2273 (long) getpid()
2274 );
2275 redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
2276 zfree(buf);
2277}
2278
2279static void sigtermHandler(int sig) {
2280 REDIS_NOTUSED(sig);
2281
2282 redisLog(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
2283 server.shutdown_asap = 1;
2284}
2285
2286void setupSignalHandlers(void) {
2287 struct sigaction act;
2288
2289 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
2290 * Otherwise, sa_handler is used. */
2291 sigemptyset(&act.sa_mask);
2292 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
2293 act.sa_handler = sigtermHandler;
2294 sigaction(SIGTERM, &act, NULL);
2295
2296#ifdef HAVE_BACKTRACE
2297 sigemptyset(&act.sa_mask);
2298 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
2299 act.sa_sigaction = sigsegvHandler;
2300 sigaction(SIGSEGV, &act, NULL);
2301 sigaction(SIGBUS, &act, NULL);
2302 sigaction(SIGFPE, &act, NULL);
2303 sigaction(SIGILL, &act, NULL);
2304#endif
2305 return;
2306}
2307
2308void memtest(size_t megabytes, int passes);
2309
2310int main(int argc, char **argv) {
2311 long long start;
2312 struct timeval tv;
2313
2314 /* We need to initialize our libraries, and the server configuration. */
2315 zmalloc_enable_thread_safeness();
2316 srand(time(NULL)^getpid());
2317 gettimeofday(&tv,NULL);
2318 dictSetHashFunctionSeed(tv.tv_sec^tv.tv_usec^getpid());
2319 initServerConfig();
2320
2321 if (argc >= 2) {
2322 int j = 1; /* First option to parse in argv[] */
2323 sds options = sdsempty();
2324 char *configfile = NULL;
2325
2326 /* Handle special options --help and --version */
2327 if (strcmp(argv[1], "-v") == 0 ||
2328 strcmp(argv[1], "--version") == 0) version();
2329 if (strcmp(argv[1], "--help") == 0 ||
2330 strcmp(argv[1], "-h") == 0) usage();
2331 if (strcmp(argv[1], "--test-memory") == 0) {
2332 if (argc == 3) {
2333 memtest(atoi(argv[2]),50);
2334 exit(0);
2335 } else {
2336 fprintf(stderr,"Please specify the amount of memory to test in megabytes.\n");
2337 fprintf(stderr,"Example: ./redis-server --test-memory 4096\n\n");
2338 exit(1);
2339 }
2340 }
2341
2342 /* First argument is the config file name? */
2343 if (argv[j][0] != '-' || argv[j][1] != '-')
2344 configfile = argv[j++];
2345 /* All the other options are parsed and conceptually appended to the
2346 * configuration file. For instance --port 6380 will generate the
2347 * string "port 6380\n" to be parsed after the actual file name
2348 * is parsed, if any. */
2349 while(j != argc) {
2350 if (argv[j][0] == '-' && argv[j][1] == '-') {
2351 /* Option name */
2352 if (sdslen(options)) options = sdscat(options,"\n");
2353 options = sdscat(options,argv[j]+2);
2354 options = sdscat(options," ");
2355 } else {
2356 /* Option argument */
2357 options = sdscatrepr(options,argv[j],strlen(argv[j]));
2358 options = sdscat(options," ");
2359 }
2360 j++;
2361 }
2362 resetServerSaveParams();
2363 loadServerConfig(configfile,options);
2364 sdsfree(options);
2365 } else {
2366 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'");
2367 }
2368 if (server.daemonize) daemonize();
2369 initServer();
2370 if (server.daemonize) createPidFile();
2371 redisAsciiArt();
2372 redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
2373#ifdef __linux__
2374 linuxOvercommitMemoryWarning();
2375#endif
2376 start = ustime();
2377 if (server.aof_state == REDIS_AOF_ON) {
2378 if (loadAppendOnlyFile(server.aof_filename) == REDIS_OK)
2379 redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
2380 } else {
2381 if (rdbLoad(server.rdb_filename) == REDIS_OK) {
2382 redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",
2383 (float)(ustime()-start)/1000000);
2384 } else if (errno != ENOENT) {
2385 redisLog(REDIS_WARNING,"Fatal error loading the DB. Exiting.");
2386 exit(1);
2387 }
2388 }
2389 if (server.ipfd > 0)
2390 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
2391 if (server.sofd > 0)
2392 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
2393 aeSetBeforeSleepProc(server.el,beforeSleep);
2394 aeMain(server.el);
2395 aeDeleteEventLoop(server.el);
2396 return 0;
2397}
2398
2399/* The End */