]> git.saurik.com Git - redis.git/blob - src/redis.c
more work done for diskstore without trying to compile, more work needed to build...
[redis.git] / src / redis.c
1 /*
2 * Copyright (c) 2009-2010, Salvatore Sanfilippo <antirez at gmail dot com>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are met:
7 *
8 * * Redistributions of source code must retain the above copyright notice,
9 * this list of conditions and the following disclaimer.
10 * * Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * * Neither the name of Redis nor the names of its contributors may be used
14 * to endorse or promote products derived from this software without
15 * specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27 * POSSIBILITY OF SUCH DAMAGE.
28 */
29
30 #include "redis.h"
31
32 #ifdef HAVE_BACKTRACE
33 #include <execinfo.h>
34 #include <ucontext.h>
35 #endif /* HAVE_BACKTRACE */
36
37 #include <time.h>
38 #include <signal.h>
39 #include <sys/wait.h>
40 #include <errno.h>
41 #include <assert.h>
42 #include <ctype.h>
43 #include <stdarg.h>
44 #include <arpa/inet.h>
45 #include <sys/stat.h>
46 #include <fcntl.h>
47 #include <sys/time.h>
48 #include <sys/resource.h>
49 #include <sys/uio.h>
50 #include <limits.h>
51 #include <float.h>
52 #include <math.h>
53 #include <pthread.h>
54 #include <sys/resource.h>
55
56 /* Our shared "common" objects */
57
58 struct sharedObjectsStruct shared;
59
60 /* Global vars that are actally used as constants. The following double
61 * values are used for double on-disk serialization, and are initialized
62 * at runtime to avoid strange compiler optimizations. */
63
64 double R_Zero, R_PosInf, R_NegInf, R_Nan;
65
66 /*================================= Globals ================================= */
67
68 /* Global vars */
69 struct redisServer server; /* server global state */
70 struct redisCommand *commandTable;
71 struct redisCommand readonlyCommandTable[] = {
72 {"get",getCommand,2,0,NULL,1,1,1},
73 {"set",setCommand,3,REDIS_CMD_DENYOOM,NULL,0,0,0},
74 {"setnx",setnxCommand,3,REDIS_CMD_DENYOOM,NULL,0,0,0},
75 {"setex",setexCommand,4,REDIS_CMD_DENYOOM,NULL,0,0,0},
76 {"append",appendCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1},
77 {"strlen",strlenCommand,2,0,NULL,1,1,1},
78 {"del",delCommand,-2,0,NULL,0,0,0},
79 {"exists",existsCommand,2,0,NULL,1,1,1},
80 {"setbit",setbitCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1},
81 {"getbit",getbitCommand,3,0,NULL,1,1,1},
82 {"setrange",setrangeCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1},
83 {"getrange",getrangeCommand,4,0,NULL,1,1,1},
84 {"substr",getrangeCommand,4,0,NULL,1,1,1},
85 {"incr",incrCommand,2,REDIS_CMD_DENYOOM,NULL,1,1,1},
86 {"decr",decrCommand,2,REDIS_CMD_DENYOOM,NULL,1,1,1},
87 {"mget",mgetCommand,-2,0,NULL,1,-1,1},
88 {"rpush",rpushCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1},
89 {"lpush",lpushCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1},
90 {"rpushx",rpushxCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1},
91 {"lpushx",lpushxCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1},
92 {"linsert",linsertCommand,5,REDIS_CMD_DENYOOM,NULL,1,1,1},
93 {"rpop",rpopCommand,2,0,NULL,1,1,1},
94 {"lpop",lpopCommand,2,0,NULL,1,1,1},
95 {"brpop",brpopCommand,-3,0,NULL,1,1,1},
96 {"brpoplpush",brpoplpushCommand,4,REDIS_CMD_DENYOOM,NULL,1,2,1},
97 {"blpop",blpopCommand,-3,0,NULL,1,1,1},
98 {"llen",llenCommand,2,0,NULL,1,1,1},
99 {"lindex",lindexCommand,3,0,NULL,1,1,1},
100 {"lset",lsetCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1},
101 {"lrange",lrangeCommand,4,0,NULL,1,1,1},
102 {"ltrim",ltrimCommand,4,0,NULL,1,1,1},
103 {"lrem",lremCommand,4,0,NULL,1,1,1},
104 {"rpoplpush",rpoplpushCommand,3,REDIS_CMD_DENYOOM,NULL,1,2,1},
105 {"sadd",saddCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1},
106 {"srem",sremCommand,3,0,NULL,1,1,1},
107 {"smove",smoveCommand,4,0,NULL,1,2,1},
108 {"sismember",sismemberCommand,3,0,NULL,1,1,1},
109 {"scard",scardCommand,2,0,NULL,1,1,1},
110 {"spop",spopCommand,2,0,NULL,1,1,1},
111 {"srandmember",srandmemberCommand,2,0,NULL,1,1,1},
112 {"sinter",sinterCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1},
113 {"sinterstore",sinterstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1},
114 {"sunion",sunionCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1},
115 {"sunionstore",sunionstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1},
116 {"sdiff",sdiffCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1},
117 {"sdiffstore",sdiffstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1},
118 {"smembers",sinterCommand,2,0,NULL,1,1,1},
119 {"zadd",zaddCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1},
120 {"zincrby",zincrbyCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1},
121 {"zrem",zremCommand,3,0,NULL,1,1,1},
122 {"zremrangebyscore",zremrangebyscoreCommand,4,0,NULL,1,1,1},
123 {"zremrangebyrank",zremrangebyrankCommand,4,0,NULL,1,1,1},
124 {"zunionstore",zunionstoreCommand,-4,REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0},
125 {"zinterstore",zinterstoreCommand,-4,REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0},
126 {"zrange",zrangeCommand,-4,0,NULL,1,1,1},
127 {"zrangebyscore",zrangebyscoreCommand,-4,0,NULL,1,1,1},
128 {"zrevrangebyscore",zrevrangebyscoreCommand,-4,0,NULL,1,1,1},
129 {"zcount",zcountCommand,4,0,NULL,1,1,1},
130 {"zrevrange",zrevrangeCommand,-4,0,NULL,1,1,1},
131 {"zcard",zcardCommand,2,0,NULL,1,1,1},
132 {"zscore",zscoreCommand,3,0,NULL,1,1,1},
133 {"zrank",zrankCommand,3,0,NULL,1,1,1},
134 {"zrevrank",zrevrankCommand,3,0,NULL,1,1,1},
135 {"hset",hsetCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1},
136 {"hsetnx",hsetnxCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1},
137 {"hget",hgetCommand,3,0,NULL,1,1,1},
138 {"hmset",hmsetCommand,-4,REDIS_CMD_DENYOOM,NULL,1,1,1},
139 {"hmget",hmgetCommand,-3,0,NULL,1,1,1},
140 {"hincrby",hincrbyCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1},
141 {"hdel",hdelCommand,3,0,NULL,1,1,1},
142 {"hlen",hlenCommand,2,0,NULL,1,1,1},
143 {"hkeys",hkeysCommand,2,0,NULL,1,1,1},
144 {"hvals",hvalsCommand,2,0,NULL,1,1,1},
145 {"hgetall",hgetallCommand,2,0,NULL,1,1,1},
146 {"hexists",hexistsCommand,3,0,NULL,1,1,1},
147 {"incrby",incrbyCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1},
148 {"decrby",decrbyCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1},
149 {"getset",getsetCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1},
150 {"mset",msetCommand,-3,REDIS_CMD_DENYOOM,NULL,1,-1,2},
151 {"msetnx",msetnxCommand,-3,REDIS_CMD_DENYOOM,NULL,1,-1,2},
152 {"randomkey",randomkeyCommand,1,0,NULL,0,0,0},
153 {"select",selectCommand,2,0,NULL,0,0,0},
154 {"move",moveCommand,3,0,NULL,1,1,1},
155 {"rename",renameCommand,3,0,NULL,1,1,1},
156 {"renamenx",renamenxCommand,3,0,NULL,1,1,1},
157 {"expire",expireCommand,3,0,NULL,0,0,0},
158 {"expireat",expireatCommand,3,0,NULL,0,0,0},
159 {"keys",keysCommand,2,0,NULL,0,0,0},
160 {"dbsize",dbsizeCommand,1,0,NULL,0,0,0},
161 {"auth",authCommand,2,0,NULL,0,0,0},
162 {"ping",pingCommand,1,0,NULL,0,0,0},
163 {"echo",echoCommand,2,0,NULL,0,0,0},
164 {"save",saveCommand,1,0,NULL,0,0,0},
165 {"bgsave",bgsaveCommand,1,0,NULL,0,0,0},
166 {"bgrewriteaof",bgrewriteaofCommand,1,0,NULL,0,0,0},
167 {"shutdown",shutdownCommand,1,0,NULL,0,0,0},
168 {"lastsave",lastsaveCommand,1,0,NULL,0,0,0},
169 {"type",typeCommand,2,0,NULL,1,1,1},
170 {"multi",multiCommand,1,0,NULL,0,0,0},
171 {"exec",execCommand,1,REDIS_CMD_DENYOOM,execBlockClientOnSwappedKeys,0,0,0},
172 {"discard",discardCommand,1,0,NULL,0,0,0},
173 {"sync",syncCommand,1,0,NULL,0,0,0},
174 {"flushdb",flushdbCommand,1,0,NULL,0,0,0},
175 {"flushall",flushallCommand,1,0,NULL,0,0,0},
176 {"sort",sortCommand,-2,REDIS_CMD_DENYOOM,NULL,1,1,1},
177 {"info",infoCommand,1,0,NULL,0,0,0},
178 {"monitor",monitorCommand,1,0,NULL,0,0,0},
179 {"ttl",ttlCommand,2,0,NULL,1,1,1},
180 {"persist",persistCommand,2,0,NULL,1,1,1},
181 {"slaveof",slaveofCommand,3,0,NULL,0,0,0},
182 {"debug",debugCommand,-2,0,NULL,0,0,0},
183 {"config",configCommand,-2,0,NULL,0,0,0},
184 {"subscribe",subscribeCommand,-2,0,NULL,0,0,0},
185 {"unsubscribe",unsubscribeCommand,-1,0,NULL,0,0,0},
186 {"psubscribe",psubscribeCommand,-2,0,NULL,0,0,0},
187 {"punsubscribe",punsubscribeCommand,-1,0,NULL,0,0,0},
188 {"publish",publishCommand,3,REDIS_CMD_FORCE_REPLICATION,NULL,0,0,0},
189 {"watch",watchCommand,-2,0,NULL,0,0,0},
190 {"unwatch",unwatchCommand,1,0,NULL,0,0,0}
191 };
192
193 /*============================ Utility functions ============================ */
194
195 void redisLog(int level, const char *fmt, ...) {
196 const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
197 const char *c = ".-*#";
198 time_t now = time(NULL);
199 va_list ap;
200 FILE *fp;
201 char buf[64];
202 char msg[REDIS_MAX_LOGMSG_LEN];
203
204 if (level < server.verbosity) return;
205
206 fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
207 if (!fp) return;
208
209 va_start(ap, fmt);
210 vsnprintf(msg, sizeof(msg), fmt, ap);
211 va_end(ap);
212
213 strftime(buf,sizeof(buf),"%d %b %H:%M:%S",localtime(&now));
214 fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
215 fflush(fp);
216
217 if (server.logfile) fclose(fp);
218
219 if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
220 }
221
222 /* Redis generally does not try to recover from out of memory conditions
223 * when allocating objects or strings, it is not clear if it will be possible
224 * to report this condition to the client since the networking layer itself
225 * is based on heap allocation for send buffers, so we simply abort.
226 * At least the code will be simpler to read... */
227 void oom(const char *msg) {
228 redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
229 sleep(1);
230 abort();
231 }
232
233 /*====================== Hash table type implementation ==================== */
234
235 /* This is an hash table type that uses the SDS dynamic strings libary as
236 * keys and radis objects as values (objects can hold SDS strings,
237 * lists, sets). */
238
239 void dictVanillaFree(void *privdata, void *val)
240 {
241 DICT_NOTUSED(privdata);
242 zfree(val);
243 }
244
245 void dictListDestructor(void *privdata, void *val)
246 {
247 DICT_NOTUSED(privdata);
248 listRelease((list*)val);
249 }
250
251 int dictSdsKeyCompare(void *privdata, const void *key1,
252 const void *key2)
253 {
254 int l1,l2;
255 DICT_NOTUSED(privdata);
256
257 l1 = sdslen((sds)key1);
258 l2 = sdslen((sds)key2);
259 if (l1 != l2) return 0;
260 return memcmp(key1, key2, l1) == 0;
261 }
262
263 /* A case insensitive version used for the command lookup table. */
264 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
265 const void *key2)
266 {
267 DICT_NOTUSED(privdata);
268
269 return strcasecmp(key1, key2) == 0;
270 }
271
272 void dictRedisObjectDestructor(void *privdata, void *val)
273 {
274 DICT_NOTUSED(privdata);
275
276 if (val == NULL) return; /* Values of swapped out keys as set to NULL */
277 decrRefCount(val);
278 }
279
280 void dictSdsDestructor(void *privdata, void *val)
281 {
282 DICT_NOTUSED(privdata);
283
284 sdsfree(val);
285 }
286
287 int dictObjKeyCompare(void *privdata, const void *key1,
288 const void *key2)
289 {
290 const robj *o1 = key1, *o2 = key2;
291 return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
292 }
293
294 unsigned int dictObjHash(const void *key) {
295 const robj *o = key;
296 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
297 }
298
299 unsigned int dictSdsHash(const void *key) {
300 return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
301 }
302
303 unsigned int dictSdsCaseHash(const void *key) {
304 return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
305 }
306
307 int dictEncObjKeyCompare(void *privdata, const void *key1,
308 const void *key2)
309 {
310 robj *o1 = (robj*) key1, *o2 = (robj*) key2;
311 int cmp;
312
313 if (o1->encoding == REDIS_ENCODING_INT &&
314 o2->encoding == REDIS_ENCODING_INT)
315 return o1->ptr == o2->ptr;
316
317 o1 = getDecodedObject(o1);
318 o2 = getDecodedObject(o2);
319 cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
320 decrRefCount(o1);
321 decrRefCount(o2);
322 return cmp;
323 }
324
325 unsigned int dictEncObjHash(const void *key) {
326 robj *o = (robj*) key;
327
328 if (o->encoding == REDIS_ENCODING_RAW) {
329 return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
330 } else {
331 if (o->encoding == REDIS_ENCODING_INT) {
332 char buf[32];
333 int len;
334
335 len = ll2string(buf,32,(long)o->ptr);
336 return dictGenHashFunction((unsigned char*)buf, len);
337 } else {
338 unsigned int hash;
339
340 o = getDecodedObject(o);
341 hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
342 decrRefCount(o);
343 return hash;
344 }
345 }
346 }
347
348 /* Sets type */
349 dictType setDictType = {
350 dictEncObjHash, /* hash function */
351 NULL, /* key dup */
352 NULL, /* val dup */
353 dictEncObjKeyCompare, /* key compare */
354 dictRedisObjectDestructor, /* key destructor */
355 NULL /* val destructor */
356 };
357
358 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
359 dictType zsetDictType = {
360 dictEncObjHash, /* hash function */
361 NULL, /* key dup */
362 NULL, /* val dup */
363 dictEncObjKeyCompare, /* key compare */
364 dictRedisObjectDestructor, /* key destructor */
365 NULL /* val destructor */
366 };
367
368 /* Db->dict, keys are sds strings, vals are Redis objects. */
369 dictType dbDictType = {
370 dictSdsHash, /* hash function */
371 NULL, /* key dup */
372 NULL, /* val dup */
373 dictSdsKeyCompare, /* key compare */
374 dictSdsDestructor, /* key destructor */
375 dictRedisObjectDestructor /* val destructor */
376 };
377
378 /* Db->expires */
379 dictType keyptrDictType = {
380 dictSdsHash, /* hash function */
381 NULL, /* key dup */
382 NULL, /* val dup */
383 dictSdsKeyCompare, /* key compare */
384 NULL, /* key destructor */
385 NULL /* val destructor */
386 };
387
388 /* Command table. sds string -> command struct pointer. */
389 dictType commandTableDictType = {
390 dictSdsCaseHash, /* hash function */
391 NULL, /* key dup */
392 NULL, /* val dup */
393 dictSdsKeyCaseCompare, /* key compare */
394 dictSdsDestructor, /* key destructor */
395 NULL /* val destructor */
396 };
397
398 /* Hash type hash table (note that small hashes are represented with zimpaps) */
399 dictType hashDictType = {
400 dictEncObjHash, /* hash function */
401 NULL, /* key dup */
402 NULL, /* val dup */
403 dictEncObjKeyCompare, /* key compare */
404 dictRedisObjectDestructor, /* key destructor */
405 dictRedisObjectDestructor /* val destructor */
406 };
407
408 /* Keylist hash table type has unencoded redis objects as keys and
409 * lists as values. It's used for blocking operations (BLPOP) and to
410 * map swapped keys to a list of clients waiting for this keys to be loaded. */
411 dictType keylistDictType = {
412 dictObjHash, /* hash function */
413 NULL, /* key dup */
414 NULL, /* val dup */
415 dictObjKeyCompare, /* key compare */
416 dictRedisObjectDestructor, /* key destructor */
417 dictListDestructor /* val destructor */
418 };
419
420 int htNeedsResize(dict *dict) {
421 long long size, used;
422
423 size = dictSlots(dict);
424 used = dictSize(dict);
425 return (size && used && size > DICT_HT_INITIAL_SIZE &&
426 (used*100/size < REDIS_HT_MINFILL));
427 }
428
429 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
430 * we resize the hash table to save memory */
431 void tryResizeHashTables(void) {
432 int j;
433
434 for (j = 0; j < server.dbnum; j++) {
435 if (htNeedsResize(server.db[j].dict))
436 dictResize(server.db[j].dict);
437 if (htNeedsResize(server.db[j].expires))
438 dictResize(server.db[j].expires);
439 }
440 }
441
442 /* Our hash table implementation performs rehashing incrementally while
443 * we write/read from the hash table. Still if the server is idle, the hash
444 * table will use two tables for a long time. So we try to use 1 millisecond
445 * of CPU time at every serverCron() loop in order to rehash some key. */
446 void incrementallyRehash(void) {
447 int j;
448
449 for (j = 0; j < server.dbnum; j++) {
450 if (dictIsRehashing(server.db[j].dict)) {
451 dictRehashMilliseconds(server.db[j].dict,1);
452 break; /* already used our millisecond for this loop... */
453 }
454 }
455 }
456
457 /* This function is called once a background process of some kind terminates,
458 * as we want to avoid resizing the hash tables when there is a child in order
459 * to play well with copy-on-write (otherwise when a resize happens lots of
460 * memory pages are copied). The goal of this function is to update the ability
461 * for dict.c to resize the hash tables accordingly to the fact we have o not
462 * running childs. */
463 void updateDictResizePolicy(void) {
464 if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1)
465 dictEnableResize();
466 else
467 dictDisableResize();
468 }
469
470 /* ======================= Cron: called every 100 ms ======================== */
471
472 /* Try to expire a few timed out keys. The algorithm used is adaptive and
473 * will use few CPU cycles if there are few expiring keys, otherwise
474 * it will get more aggressive to avoid that too much memory is used by
475 * keys that can be removed from the keyspace. */
476 void activeExpireCycle(void) {
477 int j;
478
479 for (j = 0; j < server.dbnum; j++) {
480 int expired;
481 redisDb *db = server.db+j;
482
483 /* Continue to expire if at the end of the cycle more than 25%
484 * of the keys were expired. */
485 do {
486 long num = dictSize(db->expires);
487 time_t now = time(NULL);
488
489 expired = 0;
490 if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
491 num = REDIS_EXPIRELOOKUPS_PER_CRON;
492 while (num--) {
493 dictEntry *de;
494 time_t t;
495
496 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
497 t = (time_t) dictGetEntryVal(de);
498 if (now > t) {
499 sds key = dictGetEntryKey(de);
500 robj *keyobj = createStringObject(key,sdslen(key));
501
502 propagateExpire(db,keyobj);
503 dbDelete(db,keyobj);
504 decrRefCount(keyobj);
505 expired++;
506 server.stat_expiredkeys++;
507 }
508 }
509 } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
510 }
511 }
512
513 void updateLRUClock(void) {
514 server.lruclock = (time(NULL)/REDIS_LRU_CLOCK_RESOLUTION) &
515 REDIS_LRU_CLOCK_MAX;
516 }
517
518 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
519 int j, loops = server.cronloops++;
520 REDIS_NOTUSED(eventLoop);
521 REDIS_NOTUSED(id);
522 REDIS_NOTUSED(clientData);
523
524 /* We take a cached value of the unix time in the global state because
525 * with virtual memory and aging there is to store the current time
526 * in objects at every object access, and accuracy is not needed.
527 * To access a global var is faster than calling time(NULL) */
528 server.unixtime = time(NULL);
529 /* We have just 22 bits per object for LRU information.
530 * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
531 * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
532 *
533 * Note that even if this will wrap after 1.5 years it's not a problem,
534 * everything will still work but just some object will appear younger
535 * to Redis. But for this to happen a given object should never be touched
536 * for 1.5 years.
537 *
538 * Note that you can change the resolution altering the
539 * REDIS_LRU_CLOCK_RESOLUTION define.
540 */
541 updateLRUClock();
542
543 /* We received a SIGTERM, shutting down here in a safe way, as it is
544 * not ok doing so inside the signal handler. */
545 if (server.shutdown_asap) {
546 if (prepareForShutdown() == REDIS_OK) exit(0);
547 redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
548 }
549
550 /* Show some info about non-empty databases */
551 for (j = 0; j < server.dbnum; j++) {
552 long long size, used, vkeys;
553
554 size = dictSlots(server.db[j].dict);
555 used = dictSize(server.db[j].dict);
556 vkeys = dictSize(server.db[j].expires);
557 if (!(loops % 50) && (used || vkeys)) {
558 redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
559 /* dictPrintStats(server.dict); */
560 }
561 }
562
563 /* We don't want to resize the hash tables while a bacground saving
564 * is in progress: the saving child is created using fork() that is
565 * implemented with a copy-on-write semantic in most modern systems, so
566 * if we resize the HT while there is the saving child at work actually
567 * a lot of memory movements in the parent will cause a lot of pages
568 * copied. */
569 if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1) {
570 if (!(loops % 10)) tryResizeHashTables();
571 if (server.activerehashing) incrementallyRehash();
572 }
573
574 /* Show information about connected clients */
575 if (!(loops % 50)) {
576 redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
577 listLength(server.clients)-listLength(server.slaves),
578 listLength(server.slaves),
579 zmalloc_used_memory());
580 }
581
582 /* Close connections of timedout clients */
583 if ((server.maxidletime && !(loops % 100)) || server.bpop_blocked_clients)
584 closeTimedoutClients();
585
586 /* Check if a background saving or AOF rewrite in progress terminated */
587 if (server.bgsavechildpid != -1 || server.bgrewritechildpid != -1) {
588 int statloc;
589 pid_t pid;
590
591 if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
592 if (pid == server.bgsavechildpid) {
593 backgroundSaveDoneHandler(statloc);
594 } else {
595 backgroundRewriteDoneHandler(statloc);
596 }
597 updateDictResizePolicy();
598 }
599 } else {
600 /* If there is not a background saving in progress check if
601 * we have to save now */
602 time_t now = time(NULL);
603 for (j = 0; j < server.saveparamslen; j++) {
604 struct saveparam *sp = server.saveparams+j;
605
606 if (server.dirty >= sp->changes &&
607 now-server.lastsave > sp->seconds) {
608 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
609 sp->changes, sp->seconds);
610 rdbSaveBackground(server.dbfilename);
611 break;
612 }
613 }
614 }
615
616 /* Expire a few keys per cycle, only if this is a master.
617 * On slaves we wait for DEL operations synthesized by the master
618 * in order to guarantee a strict consistency. */
619 if (server.masterhost == NULL) activeExpireCycle();
620
621 /* Remove a few cached objects from memory if we are over the
622 * configured memory limit */
623 if (server.ds_enabled) cacheCron();
624
625 /* Replication cron function -- used to reconnect to master and
626 * to detect transfer failures. */
627 if (!(loops % 10)) replicationCron();
628
629 return 100;
630 }
631
632 /* This function gets called every time Redis is entering the
633 * main loop of the event driven library, that is, before to sleep
634 * for ready file descriptors. */
635 void beforeSleep(struct aeEventLoop *eventLoop) {
636 REDIS_NOTUSED(eventLoop);
637 listNode *ln;
638 redisClient *c;
639
640 /* Awake clients that got all the on disk keys they requested */
641 if (server.ds_enabled && listLength(server.io_ready_clients)) {
642 listIter li;
643
644 listRewind(server.io_ready_clients,&li);
645 while((ln = listNext(&li))) {
646 c = ln->value;
647 struct redisCommand *cmd;
648
649 /* Resume the client. */
650 listDelNode(server.io_ready_clients,ln);
651 c->flags &= (~REDIS_IO_WAIT);
652 server.cache_blocked_clients--;
653 aeCreateFileEvent(server.el, c->fd, AE_READABLE,
654 readQueryFromClient, c);
655 cmd = lookupCommand(c->argv[0]->ptr);
656 redisAssert(cmd != NULL);
657 call(c,cmd);
658 resetClient(c);
659 /* There may be more data to process in the input buffer. */
660 if (c->querybuf && sdslen(c->querybuf) > 0)
661 processInputBuffer(c);
662 }
663 }
664
665 /* Try to process pending commands for clients that were just unblocked. */
666 while (listLength(server.unblocked_clients)) {
667 ln = listFirst(server.unblocked_clients);
668 redisAssert(ln != NULL);
669 c = ln->value;
670 listDelNode(server.unblocked_clients,ln);
671
672 /* Process remaining data in the input buffer. */
673 if (c->querybuf && sdslen(c->querybuf) > 0)
674 processInputBuffer(c);
675 }
676
677 /* Write the AOF buffer on disk */
678 flushAppendOnlyFile();
679 }
680
681 /* =========================== Server initialization ======================== */
682
683 void createSharedObjects(void) {
684 int j;
685
686 shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
687 shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
688 shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
689 shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
690 shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
691 shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
692 shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
693 shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
694 shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
695 shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
696 shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
697 shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
698 shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
699 "-ERR Operation against a key holding the wrong kind of value\r\n"));
700 shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
701 "-ERR no such key\r\n"));
702 shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
703 "-ERR syntax error\r\n"));
704 shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
705 "-ERR source and destination objects are the same\r\n"));
706 shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
707 "-ERR index out of range\r\n"));
708 shared.loadingerr = createObject(REDIS_STRING,sdsnew(
709 "-LOADING Redis is loading the dataset in memory\r\n"));
710 shared.space = createObject(REDIS_STRING,sdsnew(" "));
711 shared.colon = createObject(REDIS_STRING,sdsnew(":"));
712 shared.plus = createObject(REDIS_STRING,sdsnew("+"));
713 shared.select0 = createStringObject("select 0\r\n",10);
714 shared.select1 = createStringObject("select 1\r\n",10);
715 shared.select2 = createStringObject("select 2\r\n",10);
716 shared.select3 = createStringObject("select 3\r\n",10);
717 shared.select4 = createStringObject("select 4\r\n",10);
718 shared.select5 = createStringObject("select 5\r\n",10);
719 shared.select6 = createStringObject("select 6\r\n",10);
720 shared.select7 = createStringObject("select 7\r\n",10);
721 shared.select8 = createStringObject("select 8\r\n",10);
722 shared.select9 = createStringObject("select 9\r\n",10);
723 shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
724 shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
725 shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
726 shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
727 shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
728 shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
729 shared.mbulk3 = createStringObject("*3\r\n",4);
730 shared.mbulk4 = createStringObject("*4\r\n",4);
731 for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
732 shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
733 shared.integers[j]->encoding = REDIS_ENCODING_INT;
734 }
735 }
736
737 void initServerConfig() {
738 server.port = REDIS_SERVERPORT;
739 server.bindaddr = NULL;
740 server.unixsocket = NULL;
741 server.ipfd = -1;
742 server.sofd = -1;
743 server.dbnum = REDIS_DEFAULT_DBNUM;
744 server.verbosity = REDIS_VERBOSE;
745 server.maxidletime = REDIS_MAXIDLETIME;
746 server.saveparams = NULL;
747 server.loading = 0;
748 server.logfile = NULL; /* NULL = log on standard output */
749 server.syslog_enabled = 0;
750 server.syslog_ident = zstrdup("redis");
751 server.syslog_facility = LOG_LOCAL0;
752 server.glueoutputbuf = 1;
753 server.daemonize = 0;
754 server.appendonly = 0;
755 server.appendfsync = APPENDFSYNC_EVERYSEC;
756 server.no_appendfsync_on_rewrite = 0;
757 server.lastfsync = time(NULL);
758 server.appendfd = -1;
759 server.appendseldb = -1; /* Make sure the first time will not match */
760 server.pidfile = zstrdup("/var/run/redis.pid");
761 server.dbfilename = zstrdup("dump.rdb");
762 server.appendfilename = zstrdup("appendonly.aof");
763 server.requirepass = NULL;
764 server.rdbcompression = 1;
765 server.activerehashing = 1;
766 server.maxclients = 0;
767 server.bpop_blocked_clients = 0;
768 server.maxmemory = 0;
769 server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
770 server.maxmemory_samples = 3;
771 server.ds_enabled = 0;
772 server.ds_path = sdsnew("/tmp/redis.ds");
773 server.cache_max_memory = 64LL*1024*1024; /* 64 MB of RAM */
774 server.cache_blocked_clients = 0;
775 server.hash_max_zipmap_entries = REDIS_HASH_MAX_ZIPMAP_ENTRIES;
776 server.hash_max_zipmap_value = REDIS_HASH_MAX_ZIPMAP_VALUE;
777 server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
778 server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
779 server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
780 server.shutdown_asap = 0;
781
782 updateLRUClock();
783 resetServerSaveParams();
784
785 appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
786 appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
787 appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
788 /* Replication related */
789 server.isslave = 0;
790 server.masterauth = NULL;
791 server.masterhost = NULL;
792 server.masterport = 6379;
793 server.master = NULL;
794 server.replstate = REDIS_REPL_NONE;
795 server.repl_serve_stale_data = 1;
796
797 /* Double constants initialization */
798 R_Zero = 0.0;
799 R_PosInf = 1.0/R_Zero;
800 R_NegInf = -1.0/R_Zero;
801 R_Nan = R_Zero/R_Zero;
802
803 /* Command table -- we intiialize it here as it is part of the
804 * initial configuration, since command names may be changed via
805 * redis.conf using the rename-command directive. */
806 server.commands = dictCreate(&commandTableDictType,NULL);
807 populateCommandTable();
808 server.delCommand = lookupCommandByCString("del");
809 server.multiCommand = lookupCommandByCString("multi");
810 }
811
812 void initServer() {
813 int j;
814
815 signal(SIGHUP, SIG_IGN);
816 signal(SIGPIPE, SIG_IGN);
817 setupSigSegvAction();
818
819 if (server.syslog_enabled) {
820 openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
821 server.syslog_facility);
822 }
823
824 server.mainthread = pthread_self();
825 server.clients = listCreate();
826 server.slaves = listCreate();
827 server.monitors = listCreate();
828 server.unblocked_clients = listCreate();
829 server.cache_flush_queue = listCreate();
830 server.cache_flush_delay = 0;
831
832 createSharedObjects();
833 server.el = aeCreateEventLoop();
834 server.db = zmalloc(sizeof(redisDb)*server.dbnum);
835 server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
836 if (server.ipfd == ANET_ERR) {
837 redisLog(REDIS_WARNING, "Opening port: %s", server.neterr);
838 exit(1);
839 }
840 if (server.unixsocket != NULL) {
841 unlink(server.unixsocket); /* don't care if this fails */
842 server.sofd = anetUnixServer(server.neterr,server.unixsocket);
843 if (server.sofd == ANET_ERR) {
844 redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
845 exit(1);
846 }
847 }
848 if (server.ipfd < 0 && server.sofd < 0) {
849 redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
850 exit(1);
851 }
852 for (j = 0; j < server.dbnum; j++) {
853 server.db[j].dict = dictCreate(&dbDictType,NULL);
854 server.db[j].expires = dictCreate(&keyptrDictType,NULL);
855 server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
856 server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
857 if (server.ds_enabled)
858 server.db[j].io_keys = dictCreate(&keylistDictType,NULL);
859 server.db[j].id = j;
860 }
861 server.pubsub_channels = dictCreate(&keylistDictType,NULL);
862 server.pubsub_patterns = listCreate();
863 listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
864 listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
865 server.cronloops = 0;
866 server.bgsavechildpid = -1;
867 server.bgrewritechildpid = -1;
868 server.bgrewritebuf = sdsempty();
869 server.aofbuf = sdsempty();
870 server.lastsave = time(NULL);
871 server.dirty = 0;
872 server.stat_numcommands = 0;
873 server.stat_numconnections = 0;
874 server.stat_expiredkeys = 0;
875 server.stat_evictedkeys = 0;
876 server.stat_starttime = time(NULL);
877 server.stat_keyspace_misses = 0;
878 server.stat_keyspace_hits = 0;
879 server.unixtime = time(NULL);
880 aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
881 if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
882 acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
883 if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
884 acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
885
886 if (server.appendonly) {
887 server.appendfd = open(server.appendfilename,O_WRONLY|O_APPEND|O_CREAT,0644);
888 if (server.appendfd == -1) {
889 redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
890 strerror(errno));
891 exit(1);
892 }
893 }
894
895 if (server.ds_enabled) dsInit();
896 }
897
898 /* Populates the Redis Command Table starting from the hard coded list
899 * we have on top of redis.c file. */
900 void populateCommandTable(void) {
901 int j;
902 int numcommands = sizeof(readonlyCommandTable)/sizeof(struct redisCommand);
903
904 for (j = 0; j < numcommands; j++) {
905 struct redisCommand *c = readonlyCommandTable+j;
906 int retval;
907
908 retval = dictAdd(server.commands, sdsnew(c->name), c);
909 assert(retval == DICT_OK);
910 }
911 }
912
913 /* ====================== Commands lookup and execution ===================== */
914
915 struct redisCommand *lookupCommand(sds name) {
916 return dictFetchValue(server.commands, name);
917 }
918
919 struct redisCommand *lookupCommandByCString(char *s) {
920 struct redisCommand *cmd;
921 sds name = sdsnew(s);
922
923 cmd = dictFetchValue(server.commands, name);
924 sdsfree(name);
925 return cmd;
926 }
927
928 /* Call() is the core of Redis execution of a command */
929 void call(redisClient *c, struct redisCommand *cmd) {
930 long long dirty;
931
932 dirty = server.dirty;
933 cmd->proc(c);
934 dirty = server.dirty-dirty;
935
936 if (server.appendonly && dirty)
937 feedAppendOnlyFile(cmd,c->db->id,c->argv,c->argc);
938 if ((dirty || cmd->flags & REDIS_CMD_FORCE_REPLICATION) &&
939 listLength(server.slaves))
940 replicationFeedSlaves(server.slaves,c->db->id,c->argv,c->argc);
941 if (listLength(server.monitors))
942 replicationFeedMonitors(server.monitors,c->db->id,c->argv,c->argc);
943 server.stat_numcommands++;
944 }
945
946 /* If this function gets called we already read a whole
947 * command, argments are in the client argv/argc fields.
948 * processCommand() execute the command or prepare the
949 * server for a bulk read from the client.
950 *
951 * If 1 is returned the client is still alive and valid and
952 * and other operations can be performed by the caller. Otherwise
953 * if 0 is returned the client was destroied (i.e. after QUIT). */
954 int processCommand(redisClient *c) {
955 struct redisCommand *cmd;
956
957 /* The QUIT command is handled separately. Normal command procs will
958 * go through checking for replication and QUIT will cause trouble
959 * when FORCE_REPLICATION is enabled and would be implemented in
960 * a regular command proc. */
961 if (!strcasecmp(c->argv[0]->ptr,"quit")) {
962 addReply(c,shared.ok);
963 c->flags |= REDIS_CLOSE_AFTER_REPLY;
964 return REDIS_ERR;
965 }
966
967 /* Now lookup the command and check ASAP about trivial error conditions
968 * such wrong arity, bad command name and so forth. */
969 cmd = lookupCommand(c->argv[0]->ptr);
970 if (!cmd) {
971 addReplyErrorFormat(c,"unknown command '%s'",
972 (char*)c->argv[0]->ptr);
973 return REDIS_OK;
974 } else if ((cmd->arity > 0 && cmd->arity != c->argc) ||
975 (c->argc < -cmd->arity)) {
976 addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
977 cmd->name);
978 return REDIS_OK;
979 }
980
981 /* Check if the user is authenticated */
982 if (server.requirepass && !c->authenticated && cmd->proc != authCommand) {
983 addReplyError(c,"operation not permitted");
984 return REDIS_OK;
985 }
986
987 /* Handle the maxmemory directive.
988 *
989 * First we try to free some memory if possible (if there are volatile
990 * keys in the dataset). If there are not the only thing we can do
991 * is returning an error. */
992 if (server.maxmemory) freeMemoryIfNeeded();
993 if (server.maxmemory && (cmd->flags & REDIS_CMD_DENYOOM) &&
994 zmalloc_used_memory() > server.maxmemory)
995 {
996 addReplyError(c,"command not allowed when used memory > 'maxmemory'");
997 return REDIS_OK;
998 }
999
1000 /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1001 if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1002 &&
1003 cmd->proc != subscribeCommand && cmd->proc != unsubscribeCommand &&
1004 cmd->proc != psubscribeCommand && cmd->proc != punsubscribeCommand) {
1005 addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1006 return REDIS_OK;
1007 }
1008
1009 /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1010 * we are a slave with a broken link with master. */
1011 if (server.masterhost && server.replstate != REDIS_REPL_CONNECTED &&
1012 server.repl_serve_stale_data == 0 &&
1013 cmd->proc != infoCommand && cmd->proc != slaveofCommand)
1014 {
1015 addReplyError(c,
1016 "link with MASTER is down and slave-serve-stale-data is set to no");
1017 return REDIS_OK;
1018 }
1019
1020 /* Loading DB? Return an error if the command is not INFO */
1021 if (server.loading && cmd->proc != infoCommand) {
1022 addReply(c, shared.loadingerr);
1023 return REDIS_OK;
1024 }
1025
1026 /* Exec the command */
1027 if (c->flags & REDIS_MULTI &&
1028 cmd->proc != execCommand && cmd->proc != discardCommand &&
1029 cmd->proc != multiCommand && cmd->proc != watchCommand)
1030 {
1031 queueMultiCommand(c,cmd);
1032 addReply(c,shared.queued);
1033 } else {
1034 if (server.ds_enabled && blockClientOnSwappedKeys(c,cmd))
1035 return REDIS_ERR;
1036 call(c,cmd);
1037 }
1038 return REDIS_OK;
1039 }
1040
1041 /*================================== Shutdown =============================== */
1042
1043 int prepareForShutdown() {
1044 redisLog(REDIS_WARNING,"User requested shutdown, saving DB...");
1045 /* Kill the saving child if there is a background saving in progress.
1046 We want to avoid race conditions, for instance our saving child may
1047 overwrite the synchronous saving did by SHUTDOWN. */
1048 if (server.bgsavechildpid != -1) {
1049 redisLog(REDIS_WARNING,"There is a live saving child. Killing it!");
1050 kill(server.bgsavechildpid,SIGKILL);
1051 rdbRemoveTempFile(server.bgsavechildpid);
1052 }
1053 if (server.ds_enabled) {
1054 /* FIXME: flush all objects on disk */
1055 } else if (server.appendonly) {
1056 /* Append only file: fsync() the AOF and exit */
1057 aof_fsync(server.appendfd);
1058 } else if (server.saveparamslen > 0) {
1059 /* Snapshotting. Perform a SYNC SAVE and exit */
1060 if (rdbSave(server.dbfilename) != REDIS_OK) {
1061 /* Ooops.. error saving! The best we can do is to continue
1062 * operating. Note that if there was a background saving process,
1063 * in the next cron() Redis will be notified that the background
1064 * saving aborted, handling special stuff like slaves pending for
1065 * synchronization... */
1066 redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit");
1067 return REDIS_ERR;
1068 }
1069 } else {
1070 redisLog(REDIS_WARNING,"Not saving DB.");
1071 }
1072 if (server.daemonize) unlink(server.pidfile);
1073 redisLog(REDIS_WARNING,"Server exit now, bye bye...");
1074 return REDIS_OK;
1075 }
1076
1077 /*================================== Commands =============================== */
1078
1079 void authCommand(redisClient *c) {
1080 if (!server.requirepass || !strcmp(c->argv[1]->ptr, server.requirepass)) {
1081 c->authenticated = 1;
1082 addReply(c,shared.ok);
1083 } else {
1084 c->authenticated = 0;
1085 addReplyError(c,"invalid password");
1086 }
1087 }
1088
1089 void pingCommand(redisClient *c) {
1090 addReply(c,shared.pong);
1091 }
1092
1093 void echoCommand(redisClient *c) {
1094 addReplyBulk(c,c->argv[1]);
1095 }
1096
1097 /* Convert an amount of bytes into a human readable string in the form
1098 * of 100B, 2G, 100M, 4K, and so forth. */
1099 void bytesToHuman(char *s, unsigned long long n) {
1100 double d;
1101
1102 if (n < 1024) {
1103 /* Bytes */
1104 sprintf(s,"%lluB",n);
1105 return;
1106 } else if (n < (1024*1024)) {
1107 d = (double)n/(1024);
1108 sprintf(s,"%.2fK",d);
1109 } else if (n < (1024LL*1024*1024)) {
1110 d = (double)n/(1024*1024);
1111 sprintf(s,"%.2fM",d);
1112 } else if (n < (1024LL*1024*1024*1024)) {
1113 d = (double)n/(1024LL*1024*1024);
1114 sprintf(s,"%.2fG",d);
1115 }
1116 }
1117
1118 /* Create the string returned by the INFO command. This is decoupled
1119 * by the INFO command itself as we need to report the same information
1120 * on memory corruption problems. */
1121 sds genRedisInfoString(void) {
1122 sds info;
1123 time_t uptime = time(NULL)-server.stat_starttime;
1124 int j;
1125 char hmem[64];
1126 struct rusage self_ru, c_ru;
1127
1128 getrusage(RUSAGE_SELF, &self_ru);
1129 getrusage(RUSAGE_CHILDREN, &c_ru);
1130
1131 bytesToHuman(hmem,zmalloc_used_memory());
1132 info = sdscatprintf(sdsempty(),
1133 "redis_version:%s\r\n"
1134 "redis_git_sha1:%s\r\n"
1135 "redis_git_dirty:%d\r\n"
1136 "arch_bits:%s\r\n"
1137 "multiplexing_api:%s\r\n"
1138 "process_id:%ld\r\n"
1139 "uptime_in_seconds:%ld\r\n"
1140 "uptime_in_days:%ld\r\n"
1141 "lru_clock:%ld\r\n"
1142 "used_cpu_sys:%.2f\r\n"
1143 "used_cpu_user:%.2f\r\n"
1144 "used_cpu_sys_childrens:%.2f\r\n"
1145 "used_cpu_user_childrens:%.2f\r\n"
1146 "connected_clients:%d\r\n"
1147 "connected_slaves:%d\r\n"
1148 "blocked_clients:%d\r\n"
1149 "used_memory:%zu\r\n"
1150 "used_memory_human:%s\r\n"
1151 "used_memory_rss:%zu\r\n"
1152 "mem_fragmentation_ratio:%.2f\r\n"
1153 "use_tcmalloc:%d\r\n"
1154 "loading:%d\r\n"
1155 "aof_enabled:%d\r\n"
1156 "changes_since_last_save:%lld\r\n"
1157 "bgsave_in_progress:%d\r\n"
1158 "last_save_time:%ld\r\n"
1159 "bgrewriteaof_in_progress:%d\r\n"
1160 "total_connections_received:%lld\r\n"
1161 "total_commands_processed:%lld\r\n"
1162 "expired_keys:%lld\r\n"
1163 "evicted_keys:%lld\r\n"
1164 "keyspace_hits:%lld\r\n"
1165 "keyspace_misses:%lld\r\n"
1166 "hash_max_zipmap_entries:%zu\r\n"
1167 "hash_max_zipmap_value:%zu\r\n"
1168 "pubsub_channels:%ld\r\n"
1169 "pubsub_patterns:%u\r\n"
1170 "ds_enabled:%d\r\n"
1171 "role:%s\r\n"
1172 ,REDIS_VERSION,
1173 redisGitSHA1(),
1174 strtol(redisGitDirty(),NULL,10) > 0,
1175 (sizeof(long) == 8) ? "64" : "32",
1176 aeGetApiName(),
1177 (long) getpid(),
1178 uptime,
1179 uptime/(3600*24),
1180 (unsigned long) server.lruclock,
1181 (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
1182 (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
1183 (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000,
1184 (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
1185 listLength(server.clients)-listLength(server.slaves),
1186 listLength(server.slaves),
1187 server.bpop_blocked_clients,
1188 zmalloc_used_memory(),
1189 hmem,
1190 zmalloc_get_rss(),
1191 zmalloc_get_fragmentation_ratio(),
1192 #ifdef USE_TCMALLOC
1193 1,
1194 #else
1195 0,
1196 #endif
1197 server.loading,
1198 server.appendonly,
1199 server.dirty,
1200 server.bgsavechildpid != -1,
1201 server.lastsave,
1202 server.bgrewritechildpid != -1,
1203 server.stat_numconnections,
1204 server.stat_numcommands,
1205 server.stat_expiredkeys,
1206 server.stat_evictedkeys,
1207 server.stat_keyspace_hits,
1208 server.stat_keyspace_misses,
1209 server.hash_max_zipmap_entries,
1210 server.hash_max_zipmap_value,
1211 dictSize(server.pubsub_channels),
1212 listLength(server.pubsub_patterns),
1213 server.ds_enabled != 0,
1214 server.masterhost == NULL ? "master" : "slave"
1215 );
1216 if (server.masterhost) {
1217 info = sdscatprintf(info,
1218 "master_host:%s\r\n"
1219 "master_port:%d\r\n"
1220 "master_link_status:%s\r\n"
1221 "master_last_io_seconds_ago:%d\r\n"
1222 "master_sync_in_progress:%d\r\n"
1223 ,server.masterhost,
1224 server.masterport,
1225 (server.replstate == REDIS_REPL_CONNECTED) ?
1226 "up" : "down",
1227 server.master ? ((int)(time(NULL)-server.master->lastinteraction)) : -1,
1228 server.replstate == REDIS_REPL_TRANSFER
1229 );
1230
1231 if (server.replstate == REDIS_REPL_TRANSFER) {
1232 info = sdscatprintf(info,
1233 "master_sync_left_bytes:%ld\r\n"
1234 "master_sync_last_io_seconds_ago:%d\r\n"
1235 ,(long)server.repl_transfer_left,
1236 (int)(time(NULL)-server.repl_transfer_lastio)
1237 );
1238 }
1239 }
1240 if (server.ds_enabled) {
1241 lockThreadedIO();
1242 info = sdscatprintf(info,
1243 "cache_max_memory:%llu\r\n"
1244 "cache_blocked_clients:%lu\r\n"
1245 ,(unsigned long long) server.cache_max_memory,
1246 (unsigned long) server.cache_blocked_clients
1247 );
1248 unlockThreadedIO();
1249 }
1250 if (server.loading) {
1251 double perc;
1252 time_t eta, elapsed;
1253 off_t remaining_bytes = server.loading_total_bytes-
1254 server.loading_loaded_bytes;
1255
1256 perc = ((double)server.loading_loaded_bytes /
1257 server.loading_total_bytes) * 100;
1258
1259 elapsed = time(NULL)-server.loading_start_time;
1260 if (elapsed == 0) {
1261 eta = 1; /* A fake 1 second figure if we don't have enough info */
1262 } else {
1263 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1264 }
1265
1266 info = sdscatprintf(info,
1267 "loading_start_time:%ld\r\n"
1268 "loading_total_bytes:%llu\r\n"
1269 "loading_loaded_bytes:%llu\r\n"
1270 "loading_loaded_perc:%.2f\r\n"
1271 "loading_eta_seconds:%ld\r\n"
1272 ,(unsigned long) server.loading_start_time,
1273 (unsigned long long) server.loading_total_bytes,
1274 (unsigned long long) server.loading_loaded_bytes,
1275 perc,
1276 eta
1277 );
1278 }
1279 for (j = 0; j < server.dbnum; j++) {
1280 long long keys, vkeys;
1281
1282 keys = dictSize(server.db[j].dict);
1283 vkeys = dictSize(server.db[j].expires);
1284 if (keys || vkeys) {
1285 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
1286 j, keys, vkeys);
1287 }
1288 }
1289 return info;
1290 }
1291
1292 void infoCommand(redisClient *c) {
1293 sds info = genRedisInfoString();
1294 addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
1295 (unsigned long)sdslen(info)));
1296 addReplySds(c,info);
1297 addReply(c,shared.crlf);
1298 }
1299
1300 void monitorCommand(redisClient *c) {
1301 /* ignore MONITOR if aleady slave or in monitor mode */
1302 if (c->flags & REDIS_SLAVE) return;
1303
1304 c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
1305 c->slaveseldb = 0;
1306 listAddNodeTail(server.monitors,c);
1307 addReply(c,shared.ok);
1308 }
1309
1310 /* ============================ Maxmemory directive ======================== */
1311
1312 /* This function gets called when 'maxmemory' is set on the config file to limit
1313 * the max memory used by the server, and we are out of memory.
1314 * This function will try to, in order:
1315 *
1316 * - Free objects from the free list
1317 * - Try to remove keys with an EXPIRE set
1318 *
1319 * It is not possible to free enough memory to reach used-memory < maxmemory
1320 * the server will start refusing commands that will enlarge even more the
1321 * memory usage.
1322 */
1323 void freeMemoryIfNeeded(void) {
1324 /* Remove keys accordingly to the active policy as long as we are
1325 * over the memory limit. */
1326 if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION) return;
1327
1328 while (server.maxmemory && zmalloc_used_memory() > server.maxmemory) {
1329 int j, k, freed = 0;
1330
1331 for (j = 0; j < server.dbnum; j++) {
1332 long bestval = 0; /* just to prevent warning */
1333 sds bestkey = NULL;
1334 struct dictEntry *de;
1335 redisDb *db = server.db+j;
1336 dict *dict;
1337
1338 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
1339 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
1340 {
1341 dict = server.db[j].dict;
1342 } else {
1343 dict = server.db[j].expires;
1344 }
1345 if (dictSize(dict) == 0) continue;
1346
1347 /* volatile-random and allkeys-random policy */
1348 if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
1349 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
1350 {
1351 de = dictGetRandomKey(dict);
1352 bestkey = dictGetEntryKey(de);
1353 }
1354
1355 /* volatile-lru and allkeys-lru policy */
1356 else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
1357 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
1358 {
1359 for (k = 0; k < server.maxmemory_samples; k++) {
1360 sds thiskey;
1361 long thisval;
1362 robj *o;
1363
1364 de = dictGetRandomKey(dict);
1365 thiskey = dictGetEntryKey(de);
1366 /* When policy is volatile-lru we need an additonal lookup
1367 * to locate the real key, as dict is set to db->expires. */
1368 if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
1369 de = dictFind(db->dict, thiskey);
1370 o = dictGetEntryVal(de);
1371 thisval = estimateObjectIdleTime(o);
1372
1373 /* Higher idle time is better candidate for deletion */
1374 if (bestkey == NULL || thisval > bestval) {
1375 bestkey = thiskey;
1376 bestval = thisval;
1377 }
1378 }
1379 }
1380
1381 /* volatile-ttl */
1382 else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
1383 for (k = 0; k < server.maxmemory_samples; k++) {
1384 sds thiskey;
1385 long thisval;
1386
1387 de = dictGetRandomKey(dict);
1388 thiskey = dictGetEntryKey(de);
1389 thisval = (long) dictGetEntryVal(de);
1390
1391 /* Expire sooner (minor expire unix timestamp) is better
1392 * candidate for deletion */
1393 if (bestkey == NULL || thisval < bestval) {
1394 bestkey = thiskey;
1395 bestval = thisval;
1396 }
1397 }
1398 }
1399
1400 /* Finally remove the selected key. */
1401 if (bestkey) {
1402 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
1403 dbDelete(db,keyobj);
1404 server.stat_evictedkeys++;
1405 decrRefCount(keyobj);
1406 freed++;
1407 }
1408 }
1409 if (!freed) return; /* nothing to free... */
1410 }
1411 }
1412
1413 /* =================================== Main! ================================ */
1414
1415 #ifdef __linux__
1416 int linuxOvercommitMemoryValue(void) {
1417 FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
1418 char buf[64];
1419
1420 if (!fp) return -1;
1421 if (fgets(buf,64,fp) == NULL) {
1422 fclose(fp);
1423 return -1;
1424 }
1425 fclose(fp);
1426
1427 return atoi(buf);
1428 }
1429
1430 void linuxOvercommitMemoryWarning(void) {
1431 if (linuxOvercommitMemoryValue() == 0) {
1432 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.");
1433 }
1434 }
1435 #endif /* __linux__ */
1436
1437 void createPidFile(void) {
1438 /* Try to write the pid file in a best-effort way. */
1439 FILE *fp = fopen(server.pidfile,"w");
1440 if (fp) {
1441 fprintf(fp,"%d\n",getpid());
1442 fclose(fp);
1443 }
1444 }
1445
1446 void daemonize(void) {
1447 int fd;
1448
1449 if (fork() != 0) exit(0); /* parent exits */
1450 setsid(); /* create a new session */
1451
1452 /* Every output goes to /dev/null. If Redis is daemonized but
1453 * the 'logfile' is set to 'stdout' in the configuration file
1454 * it will not log at all. */
1455 if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
1456 dup2(fd, STDIN_FILENO);
1457 dup2(fd, STDOUT_FILENO);
1458 dup2(fd, STDERR_FILENO);
1459 if (fd > STDERR_FILENO) close(fd);
1460 }
1461 }
1462
1463 void version() {
1464 printf("Redis server version %s (%s:%d)\n", REDIS_VERSION,
1465 redisGitSHA1(), atoi(redisGitDirty()) > 0);
1466 exit(0);
1467 }
1468
1469 void usage() {
1470 fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf]\n");
1471 fprintf(stderr," ./redis-server - (read config from stdin)\n");
1472 exit(1);
1473 }
1474
1475 int main(int argc, char **argv) {
1476 time_t start;
1477
1478 initServerConfig();
1479 if (argc == 2) {
1480 if (strcmp(argv[1], "-v") == 0 ||
1481 strcmp(argv[1], "--version") == 0) version();
1482 if (strcmp(argv[1], "--help") == 0) usage();
1483 resetServerSaveParams();
1484 loadServerConfig(argv[1]);
1485 } else if ((argc > 2)) {
1486 usage();
1487 } else {
1488 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'");
1489 }
1490 if (server.daemonize) daemonize();
1491 initServer();
1492 if (server.daemonize) createPidFile();
1493 redisLog(REDIS_NOTICE,"Server started, Redis version " REDIS_VERSION);
1494 #ifdef __linux__
1495 linuxOvercommitMemoryWarning();
1496 #endif
1497 start = time(NULL);
1498 if (server.ds_enabled) {
1499 redisLog(REDIS_NOTICE,"DB not loaded (running with disk back end)");
1500 } else if (server.appendonly) {
1501 if (loadAppendOnlyFile(server.appendfilename) == REDIS_OK)
1502 redisLog(REDIS_NOTICE,"DB loaded from append only file: %ld seconds",time(NULL)-start);
1503 } else {
1504 if (rdbLoad(server.dbfilename) == REDIS_OK)
1505 redisLog(REDIS_NOTICE,"DB loaded from disk: %ld seconds",time(NULL)-start);
1506 }
1507 if (server.ipfd > 0)
1508 redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
1509 if (server.sofd > 0)
1510 redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
1511 aeSetBeforeSleepProc(server.el,beforeSleep);
1512 aeMain(server.el);
1513 aeDeleteEventLoop(server.el);
1514 return 0;
1515 }
1516
1517 /* ============================= Backtrace support ========================= */
1518
1519 #ifdef HAVE_BACKTRACE
1520 void *getMcontextEip(ucontext_t *uc) {
1521 #if defined(__FreeBSD__)
1522 return (void*) uc->uc_mcontext.mc_eip;
1523 #elif defined(__dietlibc__)
1524 return (void*) uc->uc_mcontext.eip;
1525 #elif defined(__APPLE__) && !defined(MAC_OS_X_VERSION_10_6)
1526 #if __x86_64__
1527 return (void*) uc->uc_mcontext->__ss.__rip;
1528 #else
1529 return (void*) uc->uc_mcontext->__ss.__eip;
1530 #endif
1531 #elif defined(__APPLE__) && defined(MAC_OS_X_VERSION_10_6)
1532 #if defined(_STRUCT_X86_THREAD_STATE64) && !defined(__i386__)
1533 return (void*) uc->uc_mcontext->__ss.__rip;
1534 #else
1535 return (void*) uc->uc_mcontext->__ss.__eip;
1536 #endif
1537 #elif defined(__i386__)
1538 return (void*) uc->uc_mcontext.gregs[14]; /* Linux 32 */
1539 #elif defined(__X86_64__) || defined(__x86_64__)
1540 return (void*) uc->uc_mcontext.gregs[16]; /* Linux 64 */
1541 #elif defined(__ia64__) /* Linux IA64 */
1542 return (void*) uc->uc_mcontext.sc_ip;
1543 #else
1544 return NULL;
1545 #endif
1546 }
1547
1548 void segvHandler(int sig, siginfo_t *info, void *secret) {
1549 void *trace[100];
1550 char **messages = NULL;
1551 int i, trace_size = 0;
1552 ucontext_t *uc = (ucontext_t*) secret;
1553 sds infostring;
1554 struct sigaction act;
1555 REDIS_NOTUSED(info);
1556
1557 redisLog(REDIS_WARNING,
1558 "======= Ooops! Redis %s got signal: -%d- =======", REDIS_VERSION, sig);
1559 infostring = genRedisInfoString();
1560 redisLog(REDIS_WARNING, "%s",infostring);
1561 /* It's not safe to sdsfree() the returned string under memory
1562 * corruption conditions. Let it leak as we are going to abort */
1563
1564 trace_size = backtrace(trace, 100);
1565 /* overwrite sigaction with caller's address */
1566 if (getMcontextEip(uc) != NULL) {
1567 trace[1] = getMcontextEip(uc);
1568 }
1569 messages = backtrace_symbols(trace, trace_size);
1570
1571 for (i=1; i<trace_size; ++i)
1572 redisLog(REDIS_WARNING,"%s", messages[i]);
1573
1574 /* free(messages); Don't call free() with possibly corrupted memory. */
1575 if (server.daemonize) unlink(server.pidfile);
1576
1577 /* Make sure we exit with the right signal at the end. So for instance
1578 * the core will be dumped if enabled. */
1579 sigemptyset (&act.sa_mask);
1580 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction
1581 * is used. Otherwise, sa_handler is used */
1582 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
1583 act.sa_handler = SIG_DFL;
1584 sigaction (sig, &act, NULL);
1585 kill(getpid(),sig);
1586 }
1587
1588 void sigtermHandler(int sig) {
1589 REDIS_NOTUSED(sig);
1590
1591 redisLog(REDIS_WARNING,"SIGTERM received, scheduling shutting down...");
1592 server.shutdown_asap = 1;
1593 }
1594
1595 void setupSigSegvAction(void) {
1596 struct sigaction act;
1597
1598 sigemptyset (&act.sa_mask);
1599 /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction
1600 * is used. Otherwise, sa_handler is used */
1601 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
1602 act.sa_sigaction = segvHandler;
1603 sigaction (SIGSEGV, &act, NULL);
1604 sigaction (SIGBUS, &act, NULL);
1605 sigaction (SIGFPE, &act, NULL);
1606 sigaction (SIGILL, &act, NULL);
1607 sigaction (SIGBUS, &act, NULL);
1608
1609 act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
1610 act.sa_handler = sigtermHandler;
1611 sigaction (SIGTERM, &act, NULL);
1612 return;
1613 }
1614
1615 #else /* HAVE_BACKTRACE */
1616 void setupSigSegvAction(void) {
1617 }
1618 #endif /* HAVE_BACKTRACE */
1619
1620 /* The End */