networking.c 61.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
/*
 * Copyright (c) 2009-2012, Salvatore Sanfilippo <antirez at gmail dot com>
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 *   * Redistributions of source code must retain the above copyright notice,
 *     this list of conditions and the following disclaimer.
 *   * Redistributions in binary form must reproduce the above copyright
 *     notice, this list of conditions and the following disclaimer in the
 *     documentation and/or other materials provided with the distribution.
 *   * Neither the name of Redis nor the names of its contributors may be used
 *     to endorse or promote products derived from this software without
 *     specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE.
 */

30
#include "server.h"
31
#include <sys/uio.h>
32
#include <math.h>
33

34 35
static void setProtocolError(redisClient *c, int pos);

36 37 38 39 40 41 42
/* Return the size consumed from the allocator, for the specified SDS string,
 * including internal fragmentation. This function is used in order to compute
 * the client output buffer size. */
size_t sdsZmallocSize(sds s) {
    void *sh = sdsAllocPtr(s);
    return zmalloc_size(sh);
}
43

44 45 46 47 48
/* Return the amount of memory used by the sds string at object->ptr
 * for a string object. */
size_t getStringObjectSdsUsedMemory(robj *o) {
    redisAssertWithInfo(NULL,o,o->type == REDIS_STRING);
    switch(o->encoding) {
49 50
    case REDIS_ENCODING_RAW: return sdsZmallocSize(o->ptr);
    case REDIS_ENCODING_EMBSTR: return zmalloc_size(o)-sizeof(robj);
51 52 53 54
    default: return 0; /* Just integer encoding for now. */
    }
}

55 56 57 58 59 60 61 62 63 64
void *dupClientReplyValue(void *o) {
    incrRefCount((robj*)o);
    return o;
}

int listMatchObjects(void *a, void *b) {
    return equalStringObjects(a,b);
}

redisClient *createClient(int fd) {
65
    redisClient *c = zmalloc(sizeof(redisClient));
66

67 68 69 70 71 72
    /* passing -1 as fd it is possible to create a non connected client.
     * This is useful since all the Redis commands needs to be executed
     * in the context of a client. When commands are executed in other
     * contexts (for instance a Lua script) we need a non connected client. */
    if (fd != -1) {
        anetNonBlock(NULL,fd);
73
        anetEnableTcpNoDelay(NULL,fd);
74 75
        if (server.tcpkeepalive)
            anetKeepAlive(NULL,fd,server.tcpkeepalive);
76 77 78 79 80 81 82
        if (aeCreateFileEvent(server.el,fd,AE_READABLE,
            readQueryFromClient, c) == AE_ERR)
        {
            close(fd);
            zfree(c);
            return NULL;
        }
83 84
    }

85
    selectDb(c,0);
86
    c->id = server.next_client_id++;
87
    c->fd = fd;
88
    c->name = NULL;
89
    c->bufpos = 0;
90
    c->querybuf = sdsempty();
91
    c->querybuf_peak = 0;
92
    c->reqtype = 0;
93 94
    c->argc = 0;
    c->argv = NULL;
95
    c->cmd = c->lastcmd = NULL;
96
    c->multibulklen = 0;
97 98 99
    c->bulklen = -1;
    c->sentlen = 0;
    c->flags = 0;
100
    c->ctime = c->lastinteraction = server.unixtime;
101 102
    c->authenticated = 0;
    c->replstate = REDIS_REPL_NONE;
A
antirez 已提交
103
    c->repl_put_online_on_ack = 0;
104
    c->reploff = 0;
A
antirez 已提交
105 106
    c->repl_ack_off = 0;
    c->repl_ack_time = 0;
A
antirez 已提交
107
    c->slave_listening_port = 0;
108
    c->reply = listCreate();
109
    c->reply_bytes = 0;
110
    c->obuf_soft_limit_reached_time = 0;
111
    listSetFreeMethod(c->reply,decrRefCountVoid);
112
    listSetDupMethod(c->reply,dupClientReplyValue);
113
    c->btype = REDIS_BLOCKED_NONE;
114
    c->bpop.timeout = 0;
115
    c->bpop.keys = dictCreate(&setDictType,NULL);
116
    c->bpop.target = NULL;
117 118
    c->bpop.numreplicas = 0;
    c->bpop.reploffset = 0;
119
    c->woff = 0;
120 121 122
    c->watched_keys = listCreate();
    c->pubsub_channels = dictCreate(&setDictType,NULL);
    c->pubsub_patterns = listCreate();
123
    c->peerid = NULL;
124
    listSetFreeMethod(c->pubsub_patterns,decrRefCountVoid);
125
    listSetMatchMethod(c->pubsub_patterns,listMatchObjects);
126
    if (fd != -1) listAddNodeTail(server.clients,c);
127 128 129 130
    initClientMultiState(c);
    return c;
}

131 132 133 134 135 136 137
/* This function is called every time we are going to transmit new data
 * to the client. The behavior is the following:
 *
 * If the client should receive new data (normal clients will) the function
 * returns REDIS_OK, and make sure to install the write handler in our event
 * loop so that when the socket is writable new data gets written.
 *
138 139 140 141 142 143 144 145 146 147 148
 * If the client should not receive new data, because it is a fake client
 * (used to load AOF in memory), a master or because the setup of the write
 * handler failed, the function returns REDIS_ERR.
 *
 * The function may return REDIS_OK without actually installing the write
 * event handler in the following cases:
 *
 * 1) The event handler should already be installed since the output buffer
 *    already contained something.
 * 2) The client is a slave but not yet online, so we want to just accumulate
 *    writes in the buffer but not actually sending them yet.
149 150 151 152 153
 *
 * Typically gets called every time a reply is built, before adding more
 * data to the clients output buffers. If the function returns REDIS_ERR no
 * data should be appended to the output buffers. */
int prepareClientToWrite(redisClient *c) {
154 155
    /* If it's the Lua client we always return ok without installing any
     * handler since there is no socket at all. */
156
    if (c->flags & REDIS_LUA_CLIENT) return REDIS_OK;
157 158 159

    /* Masters don't receive replies, unless REDIS_MASTER_FORCE_REPLY flag
     * is set. */
160 161
    if ((c->flags & REDIS_MASTER) &&
        !(c->flags & REDIS_MASTER_FORCE_REPLY)) return REDIS_ERR;
162 163 164 165 166

    if (c->fd <= 0) return REDIS_ERR; /* Fake client for AOF loading. */

    /* Only install the handler if not already installed and, in case of
     * slaves, if the client can actually receive writes. */
167
    if (c->bufpos == 0 && listLength(c->reply) == 0 &&
168
        (c->replstate == REDIS_REPL_NONE ||
169 170 171 172 173 174 175 176 177 178 179 180
         (c->replstate == REDIS_REPL_ONLINE && !c->repl_put_online_on_ack)))
    {
        /* Try to install the write handler. */
        if (aeCreateFileEvent(server.el, c->fd, AE_WRITABLE,
                sendReplyToClient, c) == AE_ERR)
        {
            freeClientAsync(c);
            return REDIS_ERR;
        }
    }

    /* Authorize the caller to queue in the output buffer of this client. */
181 182 183
    return REDIS_OK;
}

184 185 186 187 188 189 190 191 192 193 194 195 196 197
/* Create a duplicate of the last object in the reply list when
 * it is not exclusively owned by the reply list. */
robj *dupLastObjectIfNeeded(list *reply) {
    robj *new, *cur;
    listNode *ln;
    redisAssert(listLength(reply) > 0);
    ln = listLast(reply);
    cur = listNodeValue(ln);
    if (cur->refcount > 1) {
        new = dupStringObject(cur);
        decrRefCount(cur);
        listNodeValue(ln) = new;
    }
    return listNodeValue(ln);
198 199
}

200 201 202 203
/* -----------------------------------------------------------------------------
 * Low level functions to add more data to output buffers.
 * -------------------------------------------------------------------------- */

204
int _addReplyToBuffer(redisClient *c, const char *s, size_t len) {
205
    size_t available = sizeof(c->buf)-c->bufpos;
206

207 208
    if (c->flags & REDIS_CLOSE_AFTER_REPLY) return REDIS_OK;

209 210 211 212 213 214
    /* If there already are entries in the reply list, we cannot
     * add anything more to the static buffer. */
    if (listLength(c->reply) > 0) return REDIS_ERR;

    /* Check that the buffer has enough space available for this string. */
    if (len > available) return REDIS_ERR;
215

216 217 218
    memcpy(c->buf+c->bufpos,s,len);
    c->bufpos+=len;
    return REDIS_OK;
219 220
}

221 222
void _addReplyObjectToList(redisClient *c, robj *o) {
    robj *tail;
223 224 225

    if (c->flags & REDIS_CLOSE_AFTER_REPLY) return;

226 227 228
    if (listLength(c->reply) == 0) {
        incrRefCount(o);
        listAddNodeTail(c->reply,o);
229
        c->reply_bytes += getStringObjectSdsUsedMemory(o);
230 231 232 233 234
    } else {
        tail = listNodeValue(listLast(c->reply));

        /* Append to this object when possible. */
        if (tail->ptr != NULL &&
235
            tail->encoding == REDIS_ENCODING_RAW &&
236 237
            sdslen(tail->ptr)+sdslen(o->ptr) <= REDIS_REPLY_CHUNK_BYTES)
        {
238
            c->reply_bytes -= sdsZmallocSize(tail->ptr);
239 240
            tail = dupLastObjectIfNeeded(c->reply);
            tail->ptr = sdscatlen(tail->ptr,o->ptr,sdslen(o->ptr));
241
            c->reply_bytes += sdsZmallocSize(tail->ptr);
242 243 244
        } else {
            incrRefCount(o);
            listAddNodeTail(c->reply,o);
245
            c->reply_bytes += getStringObjectSdsUsedMemory(o);
246 247
        }
    }
248
    asyncCloseClientOnOutputBufferLimitReached(c);
249
}
250

251 252 253 254
/* This method takes responsibility over the sds. When it is no longer
 * needed it will be free'd, otherwise it ends up in a robj. */
void _addReplySdsToList(redisClient *c, sds s) {
    robj *tail;
255

256 257 258 259
    if (c->flags & REDIS_CLOSE_AFTER_REPLY) {
        sdsfree(s);
        return;
    }
260

261 262
    if (listLength(c->reply) == 0) {
        listAddNodeTail(c->reply,createObject(REDIS_STRING,s));
263
        c->reply_bytes += sdsZmallocSize(s);
264 265 266 267
    } else {
        tail = listNodeValue(listLast(c->reply));

        /* Append to this object when possible. */
268
        if (tail->ptr != NULL && tail->encoding == REDIS_ENCODING_RAW &&
269 270
            sdslen(tail->ptr)+sdslen(s) <= REDIS_REPLY_CHUNK_BYTES)
        {
271
            c->reply_bytes -= sdsZmallocSize(tail->ptr);
272 273
            tail = dupLastObjectIfNeeded(c->reply);
            tail->ptr = sdscatlen(tail->ptr,s,sdslen(s));
274
            c->reply_bytes += sdsZmallocSize(tail->ptr);
275
            sdsfree(s);
276
        } else {
277
            listAddNodeTail(c->reply,createObject(REDIS_STRING,s));
278
            c->reply_bytes += sdsZmallocSize(s);
279
        }
280
    }
281
    asyncCloseClientOnOutputBufferLimitReached(c);
282 283
}

284
void _addReplyStringToList(redisClient *c, const char *s, size_t len) {
285
    robj *tail;
286 287 288

    if (c->flags & REDIS_CLOSE_AFTER_REPLY) return;

289
    if (listLength(c->reply) == 0) {
290 291 292
        robj *o = createStringObject(s,len);

        listAddNodeTail(c->reply,o);
293
        c->reply_bytes += getStringObjectSdsUsedMemory(o);
294
    } else {
295 296 297
        tail = listNodeValue(listLast(c->reply));

        /* Append to this object when possible. */
298
        if (tail->ptr != NULL && tail->encoding == REDIS_ENCODING_RAW &&
299 300
            sdslen(tail->ptr)+len <= REDIS_REPLY_CHUNK_BYTES)
        {
301
            c->reply_bytes -= sdsZmallocSize(tail->ptr);
302 303
            tail = dupLastObjectIfNeeded(c->reply);
            tail->ptr = sdscatlen(tail->ptr,s,len);
304
            c->reply_bytes += sdsZmallocSize(tail->ptr);
305
        } else {
306 307 308
            robj *o = createStringObject(s,len);

            listAddNodeTail(c->reply,o);
309
            c->reply_bytes += getStringObjectSdsUsedMemory(o);
310 311
        }
    }
312
    asyncCloseClientOnOutputBufferLimitReached(c);
313
}
314

315 316 317 318 319
/* -----------------------------------------------------------------------------
 * Higher level functions to queue data on the client output buffer.
 * The following functions are the ones that commands implementations will call.
 * -------------------------------------------------------------------------- */

320
void addReply(redisClient *c, robj *obj) {
321
    if (prepareClientToWrite(c) != REDIS_OK) return;
322 323 324 325 326 327 328 329

    /* This is an important place where we can avoid copy-on-write
     * when there is a saving child running, avoiding touching the
     * refcount field of the object if it's not needed.
     *
     * If the encoding is RAW and there is room in the static buffer
     * we'll be able to send the object to the client without
     * messing with its page. */
330
    if (sdsEncodedObject(obj)) {
331 332
        if (_addReplyToBuffer(c,obj->ptr,sdslen(obj->ptr)) != REDIS_OK)
            _addReplyObjectToList(c,obj);
333 334 335 336 337 338 339 340 341 342 343 344 345 346
    } else if (obj->encoding == REDIS_ENCODING_INT) {
        /* Optimization: if there is room in the static buffer for 32 bytes
         * (more than the max chars a 64 bit integer can take as string) we
         * avoid decoding the object and go for the lower level approach. */
        if (listLength(c->reply) == 0 && (sizeof(c->buf) - c->bufpos) >= 32) {
            char buf[32];
            int len;

            len = ll2string(buf,sizeof(buf),(long)obj->ptr);
            if (_addReplyToBuffer(c,buf,len) == REDIS_OK)
                return;
            /* else... continue with the normal code path, but should never
             * happen actually since we verified there is room. */
        }
347
        obj = getDecodedObject(obj);
348 349 350
        if (_addReplyToBuffer(c,obj->ptr,sdslen(obj->ptr)) != REDIS_OK)
            _addReplyObjectToList(c,obj);
        decrRefCount(obj);
351 352
    } else {
        redisPanic("Wrong obj->encoding in addReply()");
353 354 355 356
    }
}

void addReplySds(redisClient *c, sds s) {
357
    if (prepareClientToWrite(c) != REDIS_OK) {
358 359 360 361
        /* The caller expects the sds to be free'd. */
        sdsfree(s);
        return;
    }
362
    if (_addReplyToBuffer(c,s,sdslen(s)) == REDIS_OK) {
363 364
        sdsfree(s);
    } else {
365 366
        /* This method free's the sds when it is no longer needed. */
        _addReplySdsToList(c,s);
367
    }
368 369
}

370
void addReplyString(redisClient *c, const char *s, size_t len) {
371
    if (prepareClientToWrite(c) != REDIS_OK) return;
372 373
    if (_addReplyToBuffer(c,s,len) != REDIS_OK)
        _addReplyStringToList(c,s,len);
374
}
375

376
void addReplyErrorLength(redisClient *c, const char *s, size_t len) {
377 378 379
    addReplyString(c,"-ERR ",5);
    addReplyString(c,s,len);
    addReplyString(c,"\r\n",2);
380 381
}

382
void addReplyError(redisClient *c, const char *err) {
383
    addReplyErrorLength(c,err,strlen(err));
384
}
385

386
void addReplyErrorFormat(redisClient *c, const char *fmt, ...) {
387
    size_t l, j;
388 389 390 391
    va_list ap;
    va_start(ap,fmt);
    sds s = sdscatvprintf(sdsempty(),fmt,ap);
    va_end(ap);
392 393 394 395 396 397
    /* Make sure there are no newlines in the string, otherwise invalid protocol
     * is emitted. */
    l = sdslen(s);
    for (j = 0; j < l; j++) {
        if (s[j] == '\r' || s[j] == '\n') s[j] = ' ';
    }
398
    addReplyErrorLength(c,s,sdslen(s));
399 400 401
    sdsfree(s);
}

402
void addReplyStatusLength(redisClient *c, const char *s, size_t len) {
403 404 405 406 407
    addReplyString(c,"+",1);
    addReplyString(c,s,len);
    addReplyString(c,"\r\n",2);
}

408
void addReplyStatus(redisClient *c, const char *status) {
409
    addReplyStatusLength(c,status,strlen(status));
410 411 412 413 414 415 416
}

void addReplyStatusFormat(redisClient *c, const char *fmt, ...) {
    va_list ap;
    va_start(ap,fmt);
    sds s = sdscatvprintf(sdsempty(),fmt,ap);
    va_end(ap);
417
    addReplyStatusLength(c,s,sdslen(s));
418 419 420
    sdsfree(s);
}

421 422 423
/* Adds an empty object to the reply list that will contain the multi bulk
 * length, which is not known when this function is called. */
void *addDeferredMultiBulkLength(redisClient *c) {
424 425 426
    /* Note that we install the write event here even if the object is not
     * ready to be sent, since we are sure that before returning to the
     * event loop setDeferredMultiBulkLength() will be called. */
427
    if (prepareClientToWrite(c) != REDIS_OK) return NULL;
428
    listAddNodeTail(c->reply,createObject(REDIS_STRING,NULL));
429 430 431
    return listLast(c->reply);
}

G
guiquanz 已提交
432
/* Populate the length object and try gluing it to the next chunk. */
433 434 435 436 437 438 439 440 441
void setDeferredMultiBulkLength(redisClient *c, void *node, long length) {
    listNode *ln = (listNode*)node;
    robj *len, *next;

    /* Abort when *node is NULL (see addDeferredMultiBulkLength). */
    if (node == NULL) return;

    len = listNodeValue(ln);
    len->ptr = sdscatprintf(sdsempty(),"*%ld\r\n",length);
442
    len->encoding = REDIS_ENCODING_RAW; /* in case it was an EMBSTR. */
443
    c->reply_bytes += sdsZmallocSize(len->ptr);
444 445
    if (ln->next != NULL) {
        next = listNodeValue(ln->next);
446

447
        /* Only glue when the next node is non-NULL (an sds in this case) */
448
        if (next->ptr != NULL) {
449
            c->reply_bytes -= sdsZmallocSize(len->ptr);
450
            c->reply_bytes -= getStringObjectSdsUsedMemory(next);
451
            len->ptr = sdscatlen(len->ptr,next->ptr,sdslen(next->ptr));
452
            c->reply_bytes += sdsZmallocSize(len->ptr);
453 454
            listDelNode(c->reply,ln->next);
        }
455
    }
456
    asyncCloseClientOnOutputBufferLimitReached(c);
457 458
}

G
guiquanz 已提交
459
/* Add a double as a bulk reply */
460 461 462
void addReplyDouble(redisClient *c, double d) {
    char dbuf[128], sbuf[128];
    int dlen, slen;
463 464 465 466 467 468 469 470 471
    if (isinf(d)) {
        /* Libc in odd systems (Hi Solaris!) will format infinite in a
         * different way, so better to handle it in an explicit way. */
        addReplyBulkCString(c, d > 0 ? "inf" : "-inf");
    } else {
        dlen = snprintf(dbuf,sizeof(dbuf),"%.17g",d);
        slen = snprintf(sbuf,sizeof(sbuf),"$%d\r\n%s\r\n",dlen,dbuf);
        addReplyString(c,sbuf,slen);
    }
472 473
}

474 475
/* Add a long long as integer reply or bulk len / multi bulk count.
 * Basically this is used to output <prefix><long long><crlf>. */
476
void addReplyLongLongWithPrefix(redisClient *c, long long ll, char prefix) {
477
    char buf[128];
478
    int len;
479 480 481 482

    /* Things like $3\r\n or *2\r\n are emitted very often by the protocol
     * so we have a few shared objects to use if the integer is small
     * like it is most of the times. */
483
    if (prefix == '*' && ll < REDIS_SHARED_BULKHDR_LEN && ll >= 0) {
484 485
        addReply(c,shared.mbulkhdr[ll]);
        return;
486
    } else if (prefix == '$' && ll < REDIS_SHARED_BULKHDR_LEN && ll >= 0) {
487 488 489 490
        addReply(c,shared.bulkhdr[ll]);
        return;
    }

491
    buf[0] = prefix;
492 493 494
    len = ll2string(buf+1,sizeof(buf)-1,ll);
    buf[len+1] = '\r';
    buf[len+2] = '\n';
495
    addReplyString(c,buf,len+3);
496 497
}

498
void addReplyLongLong(redisClient *c, long long ll) {
499 500 501 502 503
    if (ll == 0)
        addReply(c,shared.czero);
    else if (ll == 1)
        addReply(c,shared.cone);
    else
504
        addReplyLongLongWithPrefix(c,ll,':');
505
}
506

507
void addReplyMultiBulkLen(redisClient *c, long length) {
508 509 510 511
    if (length < REDIS_SHARED_BULKHDR_LEN)
        addReply(c,shared.mbulkhdr[length]);
    else
        addReplyLongLongWithPrefix(c,length,'*');
512 513
}

514
/* Create the length prefix of a bulk reply, example: $2234 */
515
void addReplyBulkLen(redisClient *c, robj *obj) {
516
    size_t len;
517

518
    if (sdsEncodedObject(obj)) {
519 520 521 522 523 524 525 526 527 528 529 530 531 532
        len = sdslen(obj->ptr);
    } else {
        long n = (long)obj->ptr;

        /* Compute how many bytes will take this integer as a radix 10 string */
        len = 1;
        if (n < 0) {
            len++;
            n = -n;
        }
        while((n = n/10) != 0) {
            len++;
        }
    }
533 534 535 536 537

    if (len < REDIS_SHARED_BULKHDR_LEN)
        addReply(c,shared.bulkhdr[len]);
    else
        addReplyLongLongWithPrefix(c,len,'$');
538 539
}

540
/* Add a Redis Object as a bulk reply */
541 542 543 544 545 546
void addReplyBulk(redisClient *c, robj *obj) {
    addReplyBulkLen(c,obj);
    addReply(c,obj);
    addReply(c,shared.crlf);
}

547
/* Add a C buffer as bulk reply */
548
void addReplyBulkCBuffer(redisClient *c, const void *p, size_t len) {
549
    addReplyLongLongWithPrefix(c,len,'$');
550 551 552 553
    addReplyString(c,p,len);
    addReply(c,shared.crlf);
}

M
Matt Stancliff 已提交
554 555 556 557 558 559 560 561
/* Add sds to reply (takes ownership of sds and frees it) */
void addReplyBulkSds(redisClient *c, sds s)  {
    addReplySds(c,sdscatfmt(sdsempty(),"$%u\r\n",
        (unsigned long)sdslen(s)));
    addReplySds(c,s);
    addReply(c,shared.crlf);
}

562
/* Add a C nul term string as bulk reply */
563
void addReplyBulkCString(redisClient *c, const char *s) {
564 565 566
    if (s == NULL) {
        addReply(c,shared.nullbulk);
    } else {
567
        addReplyBulkCBuffer(c,s,strlen(s));
568 569 570
    }
}

571 572 573 574 575 576 577 578 579
/* Add a long long as a bulk reply */
void addReplyBulkLongLong(redisClient *c, long long ll) {
    char buf[64];
    int len;

    len = ll2string(buf,64,ll);
    addReplyBulkCBuffer(c,buf,len);
}

580 581 582 583 584 585 586 587
/* Copy 'src' client output buffers into 'dst' client output buffers.
 * The function takes care of freeing the old output buffers of the
 * destination client. */
void copyClientOutputBuffer(redisClient *dst, redisClient *src) {
    listRelease(dst->reply);
    dst->reply = listDup(src->reply);
    memcpy(dst->buf,src->buf,src->bufpos);
    dst->bufpos = src->bufpos;
588
    dst->reply_bytes = src->reply_bytes;
589 590
}

A
antirez 已提交
591
#define MAX_ACCEPTS_PER_CALL 1000
592
static void acceptCommonHandler(int fd, int flags) {
593
    redisClient *c;
594
    if ((c = createClient(fd)) == NULL) {
A
antirez 已提交
595
        serverLog(REDIS_WARNING,
596 597
            "Error registering fd event for the new client: %s (fd=%d)",
            strerror(errno),fd);
598
        close(fd); /* May be already closed, just ignore errors */
599 600 601 602
        return;
    }
    /* If maxclient directive is set and this is one client more... close the
     * connection. Note that we create the client instead to check before
G
guiquanz 已提交
603
     * for this condition, since now the socket is already set in non-blocking
604
     * mode and we can send an error for free using the Kernel I/O */
605
    if (listLength(server.clients) > server.maxclients) {
606 607 608 609 610 611
        char *err = "-ERR max number of clients reached\r\n";

        /* That's a best effort error message, don't check write errors */
        if (write(c->fd,err,strlen(err)) == -1) {
            /* Nothing to do, Just to avoid the warning... */
        }
612
        server.stat_rejected_conn++;
613 614 615 616
        freeClient(c);
        return;
    }
    server.stat_numconnections++;
617
    c->flags |= flags;
618 619
}

620
void acceptTcpHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
A
antirez 已提交
621
    int cport, cfd, max = MAX_ACCEPTS_PER_CALL;
622
    char cip[REDIS_IP_STR_LEN];
623 624 625 626
    REDIS_NOTUSED(el);
    REDIS_NOTUSED(mask);
    REDIS_NOTUSED(privdata);

A
antirez 已提交
627 628 629 630
    while(max--) {
        cfd = anetTcpAccept(server.neterr, fd, cip, sizeof(cip), &cport);
        if (cfd == ANET_ERR) {
            if (errno != EWOULDBLOCK)
A
antirez 已提交
631
                serverLog(REDIS_WARNING,
A
antirez 已提交
632 633 634
                    "Accepting client connection: %s", server.neterr);
            return;
        }
A
antirez 已提交
635
        serverLog(REDIS_VERBOSE,"Accepted %s:%d", cip, cport);
A
antirez 已提交
636
        acceptCommonHandler(cfd,0);
637 638 639 640
    }
}

void acceptUnixHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
A
antirez 已提交
641
    int cfd, max = MAX_ACCEPTS_PER_CALL;
642 643 644 645
    REDIS_NOTUSED(el);
    REDIS_NOTUSED(mask);
    REDIS_NOTUSED(privdata);

A
antirez 已提交
646 647 648 649
    while(max--) {
        cfd = anetUnixAccept(server.neterr, fd);
        if (cfd == ANET_ERR) {
            if (errno != EWOULDBLOCK)
A
antirez 已提交
650
                serverLog(REDIS_WARNING,
A
antirez 已提交
651 652 653
                    "Accepting client connection: %s", server.neterr);
            return;
        }
A
antirez 已提交
654
        serverLog(REDIS_VERBOSE,"Accepted connection to %s", server.unixsocket);
A
antirez 已提交
655
        acceptCommonHandler(cfd,REDIS_UNIX_SOCKET);
656 657 658
    }
}

659 660 661 662 663
static void freeClientArgv(redisClient *c) {
    int j;
    for (j = 0; j < c->argc; j++)
        decrRefCount(c->argv[j]);
    c->argc = 0;
664
    c->cmd = NULL;
665 666
}

667 668 669 670 671 672 673 674 675 676
/* Close all the slaves connections. This is useful in chained replication
 * when we resync with our own master and want to force all our slaves to
 * resync with us as well. */
void disconnectSlaves(void) {
    while (listLength(server.slaves)) {
        listNode *ln = listFirst(server.slaves);
        freeClient((redisClient*)ln->value);
    }
}

677 678 679 680 681 682 683 684 685 686 687 688 689 690
/* This function is called when the slave lose the connection with the
 * master into an unexpected way. */
void replicationHandleMasterDisconnection(void) {
    server.master = NULL;
    server.repl_state = REDIS_REPL_CONNECT;
    server.repl_down_since = server.unixtime;
    /* We lost connection with our master, force our slaves to resync
     * with us as well to load the new data set.
     *
     * If server.masterhost is NULL the user called SLAVEOF NO ONE so
     * slave resync is not needed. */
    if (server.masterhost != NULL) disconnectSlaves();
}

691 692 693
void freeClient(redisClient *c) {
    listNode *ln;

694 695 696
    /* If this is marked as current client unset it */
    if (server.current_client == c) server.current_client = NULL;

697 698 699 700 701
    /* If it is our master that's beging disconnected we should make sure
     * to cache the state to try a partial resynchronization later.
     *
     * Note that before doing this we make sure that the client is not in
     * some unexpected state, by checking its flags. */
702
    if (server.master && c->flags & REDIS_MASTER) {
A
antirez 已提交
703
        serverLog(REDIS_WARNING,"Connection with master lost.");
704 705 706 707 708 709 710 711
        if (!(c->flags & (REDIS_CLOSE_AFTER_REPLY|
                          REDIS_CLOSE_ASAP|
                          REDIS_BLOCKED|
                          REDIS_UNBLOCKED)))
        {
            replicationCacheMaster(c);
            return;
        }
712 713
    }

714
    /* Log link disconnection with slave */
715
    if ((c->flags & REDIS_SLAVE) && !(c->flags & REDIS_MONITOR)) {
A
antirez 已提交
716
        serverLog(REDIS_WARNING,"Connection with slave %s lost.",
717
            replicationGetSlaveName(c));
718 719
    }

720
    /* Free the query buffer */
721 722
    sdsfree(c->querybuf);
    c->querybuf = NULL;
723 724

    /* Deallocate structures used to block on blocking ops. */
725
    if (c->flags & REDIS_BLOCKED) unblockClient(c);
726
    dictRelease(c->bpop.keys);
727 728 729 730

    /* UNWATCH all the keys */
    unwatchAllKeys(c);
    listRelease(c->watched_keys);
731

732 733 734 735 736
    /* Unsubscribe from all the pubsub channels */
    pubsubUnsubscribeAllChannels(c,0);
    pubsubUnsubscribeAllPatterns(c,0);
    dictRelease(c->pubsub_channels);
    listRelease(c->pubsub_patterns);
737

738 739 740 741 742 743 744
    /* Close socket, unregister events, and remove list of replies and
     * accumulated arguments. */
    if (c->fd != -1) {
        aeDeleteFileEvent(server.el,c->fd,AE_READABLE);
        aeDeleteFileEvent(server.el,c->fd,AE_WRITABLE);
        close(c->fd);
    }
745 746
    listRelease(c->reply);
    freeClientArgv(c);
747

748
    /* Remove from the list of clients */
749 750 751 752 753
    if (c->fd != -1) {
        ln = listSearchKey(server.clients,c);
        redisAssert(ln != NULL);
        listDelNode(server.clients,ln);
    }
754

755
    /* When client was just unblocked because of a blocking operation,
A
antirez 已提交
756
     * remove it from the list of unblocked clients. */
757 758 759 760 761
    if (c->flags & REDIS_UNBLOCKED) {
        ln = listSearchKey(server.unblocked_clients,c);
        redisAssert(ln != NULL);
        listDelNode(server.unblocked_clients,ln);
    }
762 763 764

    /* Master/slave cleanup Case 1:
     * we lost the connection with a slave. */
765
    if (c->flags & REDIS_SLAVE) {
766 767 768 769
        if (c->replstate == REDIS_REPL_SEND_BULK) {
            if (c->repldbfd != -1) close(c->repldbfd);
            if (c->replpreamble) sdsfree(c->replpreamble);
        }
770 771 772 773
        list *l = (c->flags & REDIS_MONITOR) ? server.monitors : server.slaves;
        ln = listSearchKey(l,c);
        redisAssert(ln != NULL);
        listDelNode(l,ln);
774 775 776 777 778
        /* We need to remember the time when we started to have zero
         * attached slaves, as after some time we'll free the replication
         * backlog. */
        if (c->flags & REDIS_SLAVE && listLength(server.slaves) == 0)
            server.repl_no_slaves_since = server.unixtime;
779
        refreshGoodSlavesCount();
780
    }
781

782 783
    /* Master/slave cleanup Case 2:
     * we lost the connection with the master. */
784
    if (c->flags & REDIS_MASTER) replicationHandleMasterDisconnection();
785 786 787 788 789 790 791 792 793

    /* If this client was scheduled for async freeing we need to remove it
     * from the queue. */
    if (c->flags & REDIS_CLOSE_ASAP) {
        ln = listSearchKey(server.clients_to_close,c);
        redisAssert(ln != NULL);
        listDelNode(server.clients_to_close,ln);
    }

794 795
    /* Release other dynamically allocated client structure fields,
     * and finally release the client structure itself. */
796
    if (c->name) decrRefCount(c->name);
797 798
    zfree(c->argv);
    freeClientMultiState(c);
799
    sdsfree(c->peerid);
800 801 802
    zfree(c);
}

803 804 805 806 807
/* Schedule a client to free it at a safe time in the serverCron() function.
 * This function is useful when we need to terminate a client but we are in
 * a context where calling freeClient() is not possible, because the client
 * should be valid for the continuation of the flow of the program. */
void freeClientAsync(redisClient *c) {
808
    if (c->flags & REDIS_CLOSE_ASAP || c->flags & REDIS_LUA_CLIENT) return;
809 810 811 812 813 814 815 816 817 818 819 820 821 822 823
    c->flags |= REDIS_CLOSE_ASAP;
    listAddNodeTail(server.clients_to_close,c);
}

void freeClientsInAsyncFreeQueue(void) {
    while (listLength(server.clients_to_close)) {
        listNode *ln = listFirst(server.clients_to_close);
        redisClient *c = listNodeValue(ln);

        c->flags &= ~REDIS_CLOSE_ASAP;
        freeClient(c);
        listDelNode(server.clients_to_close,ln);
    }
}

824 825
void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) {
    redisClient *c = privdata;
826 827
    ssize_t nwritten = 0, totwritten = 0;
    size_t objlen;
828
    size_t objmem;
829 830 831 832
    robj *o;
    REDIS_NOTUSED(el);
    REDIS_NOTUSED(mask);

833 834
    while(c->bufpos > 0 || listLength(c->reply)) {
        if (c->bufpos > 0) {
835 836
            nwritten = write(fd,c->buf+c->sentlen,c->bufpos-c->sentlen);
            if (nwritten <= 0) break;
837 838 839 840 841
            c->sentlen += nwritten;
            totwritten += nwritten;

            /* If the buffer was sent, set bufpos to zero to continue with
             * the remainder of the reply. */
842
            if ((int)c->sentlen == c->bufpos) {
843 844 845 846 847 848
                c->bufpos = 0;
                c->sentlen = 0;
            }
        } else {
            o = listNodeValue(listFirst(c->reply));
            objlen = sdslen(o->ptr);
849
            objmem = getStringObjectSdsUsedMemory(o);
850

851 852
            if (objlen == 0) {
                listDelNode(c->reply,listFirst(c->reply));
853
                c->reply_bytes -= objmem;
854 855
                continue;
            }
856

857 858
            nwritten = write(fd, ((char*)o->ptr)+c->sentlen,objlen-c->sentlen);
            if (nwritten <= 0) break;
859 860
            c->sentlen += nwritten;
            totwritten += nwritten;
861

862 863 864 865
            /* If we fully sent the object on head go to the next one */
            if (c->sentlen == objlen) {
                listDelNode(c->reply,listFirst(c->reply));
                c->sentlen = 0;
866
                c->reply_bytes -= objmem;
867
            }
868
        }
869
        /* Note that we avoid to send more than REDIS_MAX_WRITE_PER_EVENT
870 871 872
         * bytes, in a single threaded server it's a good idea to serve
         * other clients as well, even if a very large request comes from
         * super fast link that is always able to accept data (in real world
873 874 875 876
         * scenario think about 'KEYS *' against the loopback interface).
         *
         * However if we are over the maxmemory limit we ignore that and
         * just deliver as much data as it is possible to deliver. */
877
        server.stat_net_output_bytes += totwritten;
878 879 880
        if (totwritten > REDIS_MAX_WRITE_PER_EVENT &&
            (server.maxmemory == 0 ||
             zmalloc_used_memory() < server.maxmemory)) break;
881 882 883 884 885
    }
    if (nwritten == -1) {
        if (errno == EAGAIN) {
            nwritten = 0;
        } else {
A
antirez 已提交
886
            serverLog(REDIS_VERBOSE,
887 888 889 890 891
                "Error writing to client: %s", strerror(errno));
            freeClient(c);
            return;
        }
    }
A
antirez 已提交
892 893 894 895 896 897 898
    if (totwritten > 0) {
        /* For clients representing masters we don't count sending data
         * as an interaction, since we always send REPLCONF ACK commands
         * that take some time to just fill the socket output buffer.
         * We just rely on data / pings received for timeout detection. */
        if (!(c->flags & REDIS_MASTER)) c->lastinteraction = server.unixtime;
    }
899
    if (c->bufpos == 0 && listLength(c->reply) == 0) {
900 901
        c->sentlen = 0;
        aeDeleteFileEvent(server.el,c->fd,AE_WRITABLE);
P
Pieter Noordhuis 已提交
902 903

        /* Close connection after entire reply has been sent. */
904
        if (c->flags & REDIS_CLOSE_AFTER_REPLY) freeClient(c);
905 906 907 908 909
    }
}

/* resetClient prepare the client to process the next command */
void resetClient(redisClient *c) {
910 911
    redisCommandProc *prevcmd = c->cmd ? c->cmd->proc : NULL;

912
    freeClientArgv(c);
913 914
    c->reqtype = 0;
    c->multibulklen = 0;
915
    c->bulklen = -1;
916 917 918 919
    /* We clear the ASKING flag as well if we are not inside a MULTI, and
     * if what we just executed is not the ASKING command itself. */
    if (!(c->flags & REDIS_MULTI) && prevcmd != askingCommand)
        c->flags &= (~REDIS_ASKING);
920 921
}

922
int processInlineBuffer(redisClient *c) {
923
    char *newline;
924
    int argc, j;
925
    sds *argv, aux;
926 927
    size_t querylen;

928 929 930
    /* Search for end of line */
    newline = strchr(c->querybuf,'\n');

931
    /* Nothing to do without a \r\n */
932 933 934 935 936
    if (newline == NULL) {
        if (sdslen(c->querybuf) > REDIS_INLINE_MAX_SIZE) {
            addReplyError(c,"Protocol error: too big inline request");
            setProtocolError(c,0);
        }
937
        return REDIS_ERR;
938
    }
939

940 941 942 943
    /* Handle the \r\n case. */
    if (newline && newline != c->querybuf && *(newline-1) == '\r')
        newline--;

944 945
    /* Split the input buffer up to the \r\n */
    querylen = newline-(c->querybuf);
946 947 948
    aux = sdsnewlen(c->querybuf,querylen);
    argv = sdssplitargs(aux,&argc);
    sdsfree(aux);
949 950 951 952 953
    if (argv == NULL) {
        addReplyError(c,"Protocol error: unbalanced quotes in request");
        setProtocolError(c,0);
        return REDIS_ERR;
    }
954

955 956 957 958 959 960
    /* Newline from slaves can be used to refresh the last ACK time.
     * This is useful for a slave to ping back while loading a big
     * RDB file. */
    if (querylen == 0 && c->flags & REDIS_SLAVE)
        c->repl_ack_time = server.unixtime;

961
    /* Leave data after the first line of the query in the buffer */
962
    sdsrange(c->querybuf,querylen+2,-1);
963 964

    /* Setup argv array on client structure */
965 966 967 968
    if (argc) {
        if (c->argv) zfree(c->argv);
        c->argv = zmalloc(sizeof(robj*)*argc);
    }
969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985

    /* Create redis objects for all arguments. */
    for (c->argc = 0, j = 0; j < argc; j++) {
        if (sdslen(argv[j])) {
            c->argv[c->argc] = createObject(REDIS_STRING,argv[j]);
            c->argc++;
        } else {
            sdsfree(argv[j]);
        }
    }
    zfree(argv);
    return REDIS_OK;
}

/* Helper function. Trims query buffer to make the function that processes
 * multi bulk requests idempotent. */
static void setProtocolError(redisClient *c, int pos) {
986
    if (server.verbosity <= REDIS_VERBOSE) {
987
        sds client = catClientInfoString(sdsempty(),c);
A
antirez 已提交
988
        serverLog(REDIS_VERBOSE,
989 990 991
            "Protocol error from client: %s", client);
        sdsfree(client);
    }
992
    c->flags |= REDIS_CLOSE_AFTER_REPLY;
993
    sdsrange(c->querybuf,pos,-1);
994 995 996 997
}

int processMultibulkBuffer(redisClient *c) {
    char *newline = NULL;
998 999
    int pos = 0, ok;
    long long ll;
1000 1001 1002

    if (c->multibulklen == 0) {
        /* The client should have been reset */
1003
        redisAssertWithInfo(c,NULL,c->argc == 0);
1004 1005

        /* Multi bulk length cannot be read without a \r\n */
1006
        newline = strchr(c->querybuf,'\r');
1007 1008 1009 1010 1011
        if (newline == NULL) {
            if (sdslen(c->querybuf) > REDIS_INLINE_MAX_SIZE) {
                addReplyError(c,"Protocol error: too big mbulk count string");
                setProtocolError(c,0);
            }
1012
            return REDIS_ERR;
1013
        }
1014

P
Pieter Noordhuis 已提交
1015 1016 1017 1018
        /* Buffer should also contain \n */
        if (newline-(c->querybuf) > ((signed)sdslen(c->querybuf)-2))
            return REDIS_ERR;

1019 1020
        /* We know for sure there is a whole line since newline != NULL,
         * so go ahead and find out the multi bulk length. */
1021
        redisAssertWithInfo(c,NULL,c->querybuf[0] == '*');
1022 1023
        ok = string2ll(c->querybuf+1,newline-(c->querybuf+1),&ll);
        if (!ok || ll > 1024*1024) {
1024 1025 1026
            addReplyError(c,"Protocol error: invalid multibulk length");
            setProtocolError(c,pos);
            return REDIS_ERR;
1027
        }
P
Pieter Noordhuis 已提交
1028 1029 1030

        pos = (newline-c->querybuf)+2;
        if (ll <= 0) {
1031
            sdsrange(c->querybuf,pos,-1);
P
Pieter Noordhuis 已提交
1032 1033 1034
            return REDIS_OK;
        }

1035
        c->multibulklen = ll;
1036 1037 1038 1039 1040 1041

        /* Setup argv array on client structure */
        if (c->argv) zfree(c->argv);
        c->argv = zmalloc(sizeof(robj*)*c->multibulklen);
    }

1042
    redisAssertWithInfo(c,NULL,c->multibulklen > 0);
1043 1044 1045
    while(c->multibulklen) {
        /* Read bulk length if unknown */
        if (c->bulklen == -1) {
1046
            newline = strchr(c->querybuf+pos,'\r');
1047 1048
            if (newline == NULL) {
                if (sdslen(c->querybuf) > REDIS_INLINE_MAX_SIZE) {
1049 1050
                    addReplyError(c,
                        "Protocol error: too big bulk count string");
1051
                    setProtocolError(c,0);
1052
                    return REDIS_ERR;
1053
                }
P
Pieter Noordhuis 已提交
1054
                break;
1055
            }
P
Pieter Noordhuis 已提交
1056 1057 1058

            /* Buffer should also contain \n */
            if (newline-(c->querybuf) > ((signed)sdslen(c->querybuf)-2))
1059
                break;
P
Pieter Noordhuis 已提交
1060 1061 1062 1063 1064 1065 1066

            if (c->querybuf[pos] != '$') {
                addReplyErrorFormat(c,
                    "Protocol error: expected '$', got '%c'",
                    c->querybuf[pos]);
                setProtocolError(c,pos);
                return REDIS_ERR;
1067
            }
P
Pieter Noordhuis 已提交
1068 1069 1070 1071 1072 1073 1074 1075 1076

            ok = string2ll(c->querybuf+pos+1,newline-(c->querybuf+pos+1),&ll);
            if (!ok || ll < 0 || ll > 512*1024*1024) {
                addReplyError(c,"Protocol error: invalid bulk length");
                setProtocolError(c,pos);
                return REDIS_ERR;
            }

            pos += newline-(c->querybuf+pos)+2;
1077
            if (ll >= REDIS_MBULK_BIG_ARG) {
1078 1079
                size_t qblen;

1080 1081
                /* If we are going to read a large object from network
                 * try to make it likely that it will start at c->querybuf
1082
                 * boundary so that we can optimize object creation
1083
                 * avoiding a large copy of data. */
1084
                sdsrange(c->querybuf,pos,-1);
1085
                pos = 0;
1086
                qblen = sdslen(c->querybuf);
A
antirez 已提交
1087 1088
                /* Hint the sds library about the amount of bytes this string is
                 * going to contain. */
1089
                if (qblen < (size_t)ll+2)
1090
                    c->querybuf = sdsMakeRoomFor(c->querybuf,ll+2-qblen);
A
antirez 已提交
1091
            }
P
Pieter Noordhuis 已提交
1092
            c->bulklen = ll;
1093 1094 1095 1096 1097 1098 1099
        }

        /* Read bulk argument */
        if (sdslen(c->querybuf)-pos < (unsigned)(c->bulklen+2)) {
            /* Not enough data (+2 == trailing \r\n) */
            break;
        } else {
1100
            /* Optimization: if the buffer contains JUST our bulk element
1101 1102 1103
             * instead of creating a new object by *copying* the sds we
             * just use the current sds string. */
            if (pos == 0 &&
1104
                c->bulklen >= REDIS_MBULK_BIG_ARG &&
1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
                (signed) sdslen(c->querybuf) == c->bulklen+2)
            {
                c->argv[c->argc++] = createObject(REDIS_STRING,c->querybuf);
                sdsIncrLen(c->querybuf,-2); /* remove CRLF */
                c->querybuf = sdsempty();
                /* Assume that if we saw a fat argument we'll see another one
                 * likely... */
                c->querybuf = sdsMakeRoomFor(c->querybuf,c->bulklen+2);
                pos = 0;
            } else {
                c->argv[c->argc++] =
                    createStringObject(c->querybuf+pos,c->bulklen);
                pos += c->bulklen+2;
            }
1119 1120 1121 1122 1123 1124
            c->bulklen = -1;
            c->multibulklen--;
        }
    }

    /* Trim to pos */
1125
    if (pos) sdsrange(c->querybuf,pos,-1);
1126 1127

    /* We're done when c->multibulk == 0 */
1128 1129 1130
    if (c->multibulklen == 0) return REDIS_OK;

    /* Still not read to process the command */
1131 1132 1133 1134
    return REDIS_ERR;
}

void processInputBuffer(redisClient *c) {
1135
    server.current_client = c;
1136 1137
    /* Keep processing while there is something in the input buffer */
    while(sdslen(c->querybuf)) {
1138
        /* Return if clients are paused. */
1139
        if (!(c->flags & REDIS_SLAVE) && clientsArePaused()) break;
1140

1141
        /* Immediately abort if the client is in the middle of something. */
1142
        if (c->flags & REDIS_BLOCKED) break;
1143

1144 1145 1146
        /* REDIS_CLOSE_AFTER_REPLY closes the connection once the reply is
         * written to the client. Make sure to not let the reply grow after
         * this flag has been set (i.e. don't process more commands). */
1147
        if (c->flags & REDIS_CLOSE_AFTER_REPLY) break;
1148 1149 1150 1151 1152

        /* Determine request type when unknown. */
        if (!c->reqtype) {
            if (c->querybuf[0] == '*') {
                c->reqtype = REDIS_REQ_MULTIBULK;
1153
            } else {
1154
                c->reqtype = REDIS_REQ_INLINE;
1155 1156
            }
        }
1157 1158 1159 1160 1161 1162 1163

        if (c->reqtype == REDIS_REQ_INLINE) {
            if (processInlineBuffer(c) != REDIS_OK) break;
        } else if (c->reqtype == REDIS_REQ_MULTIBULK) {
            if (processMultibulkBuffer(c) != REDIS_OK) break;
        } else {
            redisPanic("Unknown request type");
1164
        }
1165 1166

        /* Multibulk processing could see a <= 0 length. */
1167 1168 1169 1170 1171 1172 1173
        if (c->argc == 0) {
            resetClient(c);
        } else {
            /* Only reset the client when the command was executed. */
            if (processCommand(c) == REDIS_OK)
                resetClient(c);
        }
1174
    }
1175
    server.current_client = NULL;
1176 1177 1178 1179
}

void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask) {
    redisClient *c = (redisClient*) privdata;
1180
    int nread, readlen;
1181
    size_t qblen;
1182 1183 1184
    REDIS_NOTUSED(el);
    REDIS_NOTUSED(mask);

1185 1186
    readlen = REDIS_IOBUF_LEN;
    /* If this is a multi bulk request, and we are processing a bulk reply
1187 1188 1189
     * that is large enough, try to maximize the probability that the query
     * buffer contains exactly the SDS string representing the object, even
     * at the risk of requiring more read(2) calls. This way the function
1190 1191 1192
     * processMultiBulkBuffer() can avoid copying buffers to create the
     * Redis Object representing the argument. */
    if (c->reqtype == REDIS_REQ_MULTIBULK && c->multibulklen && c->bulklen != -1
1193
        && c->bulklen >= REDIS_MBULK_BIG_ARG)
1194 1195 1196 1197 1198 1199
    {
        int remaining = (unsigned)(c->bulklen+2)-sdslen(c->querybuf);

        if (remaining < readlen) readlen = remaining;
    }

1200
    qblen = sdslen(c->querybuf);
1201
    if (c->querybuf_peak < qblen) c->querybuf_peak = qblen;
1202 1203
    c->querybuf = sdsMakeRoomFor(c->querybuf, readlen);
    nread = read(fd, c->querybuf+qblen, readlen);
1204 1205
    if (nread == -1) {
        if (errno == EAGAIN) {
1206
            return;
1207
        } else {
A
antirez 已提交
1208
            serverLog(REDIS_VERBOSE, "Reading from client: %s",strerror(errno));
1209 1210 1211 1212
            freeClient(c);
            return;
        }
    } else if (nread == 0) {
A
antirez 已提交
1213
        serverLog(REDIS_VERBOSE, "Client closed connection");
1214 1215 1216
        freeClient(c);
        return;
    }
1217 1218 1219 1220 1221

    sdsIncrLen(c->querybuf,nread);
    c->lastinteraction = server.unixtime;
    if (c->flags & REDIS_MASTER) c->reploff += nread;
    server.stat_net_input_bytes += nread;
1222
    if (sdslen(c->querybuf) > server.client_max_querybuf_len) {
1223
        sds ci = catClientInfoString(sdsempty(),c), bytes = sdsempty();
1224 1225

        bytes = sdscatrepr(bytes,c->querybuf,64);
A
antirez 已提交
1226
        serverLog(REDIS_WARNING,"Closing client that reached max query buffer length: %s (qbuf initial bytes: %s)", ci, bytes);
1227
        sdsfree(ci);
1228
        sdsfree(bytes);
1229 1230 1231
        freeClient(c);
        return;
    }
1232 1233
    processInputBuffer(c);
}
1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252

void getClientsMaxBuffers(unsigned long *longest_output_list,
                          unsigned long *biggest_input_buffer) {
    redisClient *c;
    listNode *ln;
    listIter li;
    unsigned long lol = 0, bib = 0;

    listRewind(server.clients,&li);
    while ((ln = listNext(&li)) != NULL) {
        c = listNodeValue(ln);

        if (listLength(c->reply) > lol) lol = listLength(c->reply);
        if (sdslen(c->querybuf) > bib) bib = sdslen(c->querybuf);
    }
    *longest_output_list = lol;
    *biggest_input_buffer = bib;
}

A
antirez 已提交
1253
/* A Redis "Peer ID" is a colon separated ip:port pair.
G
Gregory Petrosyan 已提交
1254
 * For IPv4 it's in the form x.y.z.k:port, example: "127.0.0.1:1234".
A
antirez 已提交
1255
 * For IPv6 addresses we use [] around the IP part, like in "[::1]:1234".
G
Gregory Petrosyan 已提交
1256
 * For Unix sockets we use path:0, like in "/tmp/redis:0".
A
antirez 已提交
1257 1258 1259 1260
 *
 * A Peer ID always fits inside a buffer of REDIS_PEER_ID_LEN bytes, including
 * the null term.
 *
A
antirez 已提交
1261 1262 1263
 * On failure the function still populates 'peerid' with the "?:0" string
 * in case you want to relax error checking or need to display something
 * anyway (see anetPeerToString implementation for more info). */
A
antirez 已提交
1264
void genClientPeerId(redisClient *client, char *peerid,
M
Matt Stancliff 已提交
1265
                            size_t peerid_len) {
A
antirez 已提交
1266 1267 1268 1269 1270
    if (client->flags & REDIS_UNIX_SOCKET) {
        /* Unix socket client. */
        snprintf(peerid,peerid_len,"%s:0",server.unixsocket);
    } else {
        /* TCP client. */
M
Matt Stancliff 已提交
1271
        anetFormatPeer(client->fd,peerid,peerid_len);
A
antirez 已提交
1272 1273 1274
    }
}

1275
/* This function returns the client peer id, by creating and caching it
G
Gregory Petrosyan 已提交
1276
 * if client->peerid is NULL, otherwise returning the cached value.
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292
 * The Peer ID never changes during the life of the client, however it
 * is expensive to compute. */
char *getClientPeerId(redisClient *c) {
    char peerid[REDIS_PEER_ID_LEN];

    if (c->peerid == NULL) {
        genClientPeerId(c,peerid,sizeof(peerid));
        c->peerid = sdsnew(peerid);
    }
    return c->peerid;
}

/* Concatenate a string representing the state of a client in an human
 * readable format, into the sds string 's'. */
sds catClientInfoString(sds s, redisClient *client) {
    char flags[16], events[3], *p;
1293
    int emask;
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307

    p = flags;
    if (client->flags & REDIS_SLAVE) {
        if (client->flags & REDIS_MONITOR)
            *p++ = 'O';
        else
            *p++ = 'S';
    }
    if (client->flags & REDIS_MASTER) *p++ = 'M';
    if (client->flags & REDIS_MULTI) *p++ = 'x';
    if (client->flags & REDIS_BLOCKED) *p++ = 'b';
    if (client->flags & REDIS_DIRTY_CAS) *p++ = 'd';
    if (client->flags & REDIS_CLOSE_AFTER_REPLY) *p++ = 'c';
    if (client->flags & REDIS_UNBLOCKED) *p++ = 'u';
1308
    if (client->flags & REDIS_CLOSE_ASAP) *p++ = 'A';
1309
    if (client->flags & REDIS_UNIX_SOCKET) *p++ = 'U';
1310
    if (client->flags & REDIS_READONLY) *p++ = 'r';
1311
    if (p == flags) *p++ = 'N';
1312
    *p++ = '\0';
1313 1314 1315 1316 1317 1318

    emask = client->fd == -1 ? 0 : aeGetFileEvents(server.el,client->fd);
    p = events;
    if (emask & AE_READABLE) *p++ = 'r';
    if (emask & AE_WRITABLE) *p++ = 'w';
    *p = '\0';
1319
    return sdscatfmt(s,
1320 1321
        "id=%U addr=%s fd=%i name=%s age=%I idle=%I flags=%s db=%i sub=%i psub=%i multi=%i qbuf=%U qbuf-free=%U obl=%U oll=%U omem=%U events=%s cmd=%s",
        (unsigned long long) client->id,
1322
        getClientPeerId(client),
1323 1324
        client->fd,
        client->name ? (char*)client->name->ptr : "",
1325 1326
        (long long)(server.unixtime - client->ctime),
        (long long)(server.unixtime - client->lastinteraction),
1327 1328 1329
        flags,
        client->db->id,
        (int) dictSize(client->pubsub_channels),
1330
        (int) listLength(client->pubsub_patterns),
1331
        (client->flags & REDIS_MULTI) ? client->mstate.count : -1,
1332 1333 1334 1335 1336
        (unsigned long long) sdslen(client->querybuf),
        (unsigned long long) sdsavail(client->querybuf),
        (unsigned long long) client->bufpos,
        (unsigned long long) listLength(client->reply),
        (unsigned long long) getClientOutputBufferMemoryUsage(client),
1337 1338
        events,
        client->lastcmd ? client->lastcmd->name : "NULL");
1339 1340
}

1341 1342 1343 1344 1345 1346
sds getAllClientsInfoString(void) {
    listNode *ln;
    listIter li;
    redisClient *client;
    sds o = sdsempty();

1347
    o = sdsMakeRoomFor(o,200*listLength(server.clients));
1348 1349 1350
    listRewind(server.clients,&li);
    while ((ln = listNext(&li)) != NULL) {
        client = listNodeValue(ln);
1351
        o = catClientInfoString(o,client);
1352 1353 1354 1355 1356
        o = sdscatlen(o,"\n",1);
    }
    return o;
}

A
antirez 已提交
1357
void clientCommand(redisClient *c) {
A
antirez 已提交
1358 1359 1360 1361
    listNode *ln;
    listIter li;
    redisClient *client;

A
antirez 已提交
1362
    if (!strcasecmp(c->argv[1]->ptr,"list") && c->argc == 2) {
A
antirez 已提交
1363
        /* CLIENT LIST */
1364
        sds o = getAllClientsInfoString();
A
antirez 已提交
1365 1366
        addReplyBulkCBuffer(c,o,sdslen(o));
        sdsfree(o);
A
antirez 已提交
1367 1368
    } else if (!strcasecmp(c->argv[1]->ptr,"kill")) {
        /* CLIENT KILL <ip:port>
A
antirez 已提交
1369
         * CLIENT KILL <option> [value] ... <option> [value] */
A
antirez 已提交
1370 1371 1372
        char *addr = NULL;
        int type = -1;
        uint64_t id = 0;
A
antirez 已提交
1373 1374
        int skipme = 1;
        int killed = 0, close_this_client = 0;
A
antirez 已提交
1375 1376

        if (c->argc == 3) {
A
antirez 已提交
1377
            /* Old style syntax: CLIENT KILL <addr> */
A
antirez 已提交
1378
            addr = c->argv[2]->ptr;
1379
            skipme = 0; /* With the old form, you can kill yourself. */
A
antirez 已提交
1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412
        } else if (c->argc > 3) {
            int i = 2; /* Next option index. */

            /* New style syntax: parse options. */
            while(i < c->argc) {
                int moreargs = c->argc > i+1;

                if (!strcasecmp(c->argv[i]->ptr,"id") && moreargs) {
                    long long tmp;

                    if (getLongLongFromObjectOrReply(c,c->argv[i+1],&tmp,NULL)
                        != REDIS_OK) return;
                    id = tmp;
                } else if (!strcasecmp(c->argv[i]->ptr,"type") && moreargs) {
                    type = getClientTypeByName(c->argv[i+1]->ptr);
                    if (type == -1) {
                        addReplyErrorFormat(c,"Unknown client type '%s'",
                            (char*) c->argv[i+1]->ptr);
                        return;
                    }
                } else if (!strcasecmp(c->argv[i]->ptr,"addr") && moreargs) {
                    addr = c->argv[i+1]->ptr;
                } else if (!strcasecmp(c->argv[i]->ptr,"skipme") && moreargs) {
                    if (!strcasecmp(c->argv[i+1]->ptr,"yes")) {
                        skipme = 1;
                    } else if (!strcasecmp(c->argv[i+1]->ptr,"no")) {
                        skipme = 0;
                    } else {
                        addReply(c,shared.syntaxerr);
                        return;
                    }
                } else {
                    addReply(c,shared.syntaxerr);
A
antirez 已提交
1413 1414
                    return;
                }
A
antirez 已提交
1415
                i += 2;
A
antirez 已提交
1416 1417 1418 1419 1420 1421 1422
            }
        } else {
            addReply(c,shared.syntaxerr);
            return;
        }

        /* Iterate clients killing all the matching clients. */
A
antirez 已提交
1423 1424 1425
        listRewind(server.clients,&li);
        while ((ln = listNext(&li)) != NULL) {
            client = listNodeValue(ln);
A
antirez 已提交
1426
            if (addr && strcmp(getClientPeerId(client),addr) != 0) continue;
1427 1428 1429
            if (type != -1 &&
                (client->flags & REDIS_MASTER ||
                 getClientType(client) != type)) continue;
A
antirez 已提交
1430
            if (id != 0 && client->id != id) continue;
A
antirez 已提交
1431
            if (c == client && skipme) continue;
A
antirez 已提交
1432 1433 1434

            /* Kill it. */
            if (c == client) {
1435
                close_this_client = 1;
A
antirez 已提交
1436 1437
            } else {
                freeClient(client);
A
antirez 已提交
1438
            }
A
antirez 已提交
1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449
            killed++;
        }

        /* Reply according to old/new format. */
        if (c->argc == 3) {
            if (killed == 0)
                addReplyError(c,"No such client");
            else
                addReply(c,shared.ok);
        } else {
            addReplyLongLong(c,killed);
A
antirez 已提交
1450
        }
1451 1452 1453 1454

        /* If this client has to be closed, flag it as CLOSE_AFTER_REPLY
         * only after we queued the reply to its output buffers. */
        if (close_this_client) c->flags |= REDIS_CLOSE_AFTER_REPLY;
1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471
    } else if (!strcasecmp(c->argv[1]->ptr,"setname") && c->argc == 3) {
        int j, len = sdslen(c->argv[2]->ptr);
        char *p = c->argv[2]->ptr;

        /* Setting the client name to an empty string actually removes
         * the current name. */
        if (len == 0) {
            if (c->name) decrRefCount(c->name);
            c->name = NULL;
            addReply(c,shared.ok);
            return;
        }

        /* Otherwise check if the charset is ok. We need to do this otherwise
         * CLIENT LIST format will break. You should always be able to
         * split by space to get the different fields. */
        for (j = 0; j < len; j++) {
A
antirez 已提交
1472
            if (p[j] < '!' || p[j] > '~') { /* ASCII is assumed. */
1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487
                addReplyError(c,
                    "Client names cannot contain spaces, "
                    "newlines or special characters.");
                return;
            }
        }
        if (c->name) decrRefCount(c->name);
        c->name = c->argv[2];
        incrRefCount(c->name);
        addReply(c,shared.ok);
    } else if (!strcasecmp(c->argv[1]->ptr,"getname") && c->argc == 2) {
        if (c->name)
            addReplyBulk(c,c->name);
        else
            addReply(c,shared.nullbulk);
1488 1489 1490 1491 1492 1493 1494
    } else if (!strcasecmp(c->argv[1]->ptr,"pause") && c->argc == 3) {
        long long duration;

        if (getTimeoutFromObjectOrReply(c,c->argv[2],&duration,UNIT_MILLISECONDS)
                                        != REDIS_OK) return;
        pauseClients(duration);
        addReply(c,shared.ok);
A
antirez 已提交
1495
    } else {
B
bitterb 已提交
1496
        addReplyError(c, "Syntax error, try CLIENT (LIST | KILL ip:port | GETNAME | SETNAME connection-name)");
A
antirez 已提交
1497 1498
    }
}
1499

1500 1501 1502
/* Rewrite the command vector of the client. All the new objects ref count
 * is incremented. The old command vector is freed, and the old objects
 * ref count is decremented. */
1503 1504 1505 1506 1507 1508 1509 1510 1511
void rewriteClientCommandVector(redisClient *c, int argc, ...) {
    va_list ap;
    int j;
    robj **argv; /* The new argument vector */

    argv = zmalloc(sizeof(robj*)*argc);
    va_start(ap,argc);
    for (j = 0; j < argc; j++) {
        robj *a;
1512

1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524
        a = va_arg(ap, robj*);
        argv[j] = a;
        incrRefCount(a);
    }
    /* We free the objects in the original vector at the end, so we are
     * sure that if the same objects are reused in the new vector the
     * refcount gets incremented before it gets decremented. */
    for (j = 0; j < c->argc; j++) decrRefCount(c->argv[j]);
    zfree(c->argv);
    /* Replace argv and argc with our new versions. */
    c->argv = argv;
    c->argc = argc;
1525
    c->cmd = lookupCommandOrOriginal(c->argv[0]->ptr);
1526
    redisAssertWithInfo(c,NULL,c->cmd != NULL);
1527 1528
    va_end(ap);
}
1529

1530 1531 1532 1533 1534 1535 1536 1537 1538 1539
/* Completely replace the client command vector with the provided one. */
void replaceClientCommandVector(redisClient *c, int argc, robj **argv) {
    freeClientArgv(c);
    zfree(c->argv);
    c->argv = argv;
    c->argc = argc;
    c->cmd = lookupCommandOrOriginal(c->argv[0]->ptr);
    redisAssertWithInfo(c,NULL,c->cmd != NULL);
}

1540 1541 1542 1543
/* Rewrite a single item in the command vector.
 * The new val ref count is incremented, and the old decremented. */
void rewriteClientCommandArgument(redisClient *c, int i, robj *newval) {
    robj *oldval;
1544

1545
    redisAssertWithInfo(c,NULL,i < c->argc);
1546 1547 1548 1549 1550 1551 1552
    oldval = c->argv[i];
    c->argv[i] = newval;
    incrRefCount(newval);
    decrRefCount(oldval);

    /* If this is the command name make sure to fix c->cmd. */
    if (i == 0) {
1553
        c->cmd = lookupCommandOrOriginal(c->argv[0]->ptr);
1554
        redisAssertWithInfo(c,NULL,c->cmd != NULL);
1555 1556
    }
}
1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569

/* This function returns the number of bytes that Redis is virtually
 * using to store the reply still not read by the client.
 * It is "virtual" since the reply output list may contain objects that
 * are shared and are not really using additional memory.
 *
 * The function returns the total sum of the length of all the objects
 * stored in the output list, plus the memory used to allocate every
 * list node. The static reply buffer is not taken into account since it
 * is allocated anyway.
 *
 * Note: this function is very fast so can be called as many time as
 * the caller wishes. The main usage of this function currently is
A
antirez 已提交
1570
 * enforcing the client output length limits. */
1571
unsigned long getClientOutputBufferMemoryUsage(redisClient *c) {
1572
    unsigned long list_item_size = sizeof(listNode)+sizeof(robj);
1573 1574 1575

    return c->reply_bytes + (list_item_size*listLength(c->reply));
}
1576

1577
/* Get the class of a client, used in order to enforce limits to different
1578 1579 1580
 * classes of clients.
 *
 * The function will return one of the following:
A
antirez 已提交
1581 1582 1583
 * REDIS_CLIENT_TYPE_NORMAL -> Normal client
 * REDIS_CLIENT_TYPE_SLAVE  -> Slave or client executing MONITOR command
 * REDIS_CLIENT_TYPE_PUBSUB -> Client subscribed to Pub/Sub channels
1584
 */
A
antirez 已提交
1585 1586 1587
int getClientType(redisClient *c) {
    if ((c->flags & REDIS_SLAVE) && !(c->flags & REDIS_MONITOR))
        return REDIS_CLIENT_TYPE_SLAVE;
1588
    if (c->flags & REDIS_PUBSUB)
A
antirez 已提交
1589 1590
        return REDIS_CLIENT_TYPE_PUBSUB;
    return REDIS_CLIENT_TYPE_NORMAL;
1591
}
1592

A
antirez 已提交
1593 1594 1595 1596
int getClientTypeByName(char *name) {
    if (!strcasecmp(name,"normal")) return REDIS_CLIENT_TYPE_NORMAL;
    else if (!strcasecmp(name,"slave")) return REDIS_CLIENT_TYPE_SLAVE;
    else if (!strcasecmp(name,"pubsub")) return REDIS_CLIENT_TYPE_PUBSUB;
1597 1598 1599
    else return -1;
}

A
antirez 已提交
1600
char *getClientTypeName(int class) {
1601
    switch(class) {
A
antirez 已提交
1602 1603 1604 1605
    case REDIS_CLIENT_TYPE_NORMAL: return "normal";
    case REDIS_CLIENT_TYPE_SLAVE:  return "slave";
    case REDIS_CLIENT_TYPE_PUBSUB: return "pubsub";
    default:                       return NULL;
1606 1607 1608
    }
}

1609 1610 1611 1612 1613 1614 1615 1616 1617 1618
/* The function checks if the client reached output buffer soft or hard
 * limit, and also update the state needed to check the soft limit as
 * a side effect.
 *
 * Return value: non-zero if the client reached the soft or the hard limit.
 *               Otherwise zero is returned. */
int checkClientOutputBufferLimits(redisClient *c) {
    int soft = 0, hard = 0, class;
    unsigned long used_mem = getClientOutputBufferMemoryUsage(c);

A
antirez 已提交
1619
    class = getClientType(c);
1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649
    if (server.client_obuf_limits[class].hard_limit_bytes &&
        used_mem >= server.client_obuf_limits[class].hard_limit_bytes)
        hard = 1;
    if (server.client_obuf_limits[class].soft_limit_bytes &&
        used_mem >= server.client_obuf_limits[class].soft_limit_bytes)
        soft = 1;

    /* We need to check if the soft limit is reached continuously for the
     * specified amount of seconds. */
    if (soft) {
        if (c->obuf_soft_limit_reached_time == 0) {
            c->obuf_soft_limit_reached_time = server.unixtime;
            soft = 0; /* First time we see the soft limit reached */
        } else {
            time_t elapsed = server.unixtime - c->obuf_soft_limit_reached_time;

            if (elapsed <=
                server.client_obuf_limits[class].soft_limit_seconds) {
                soft = 0; /* The client still did not reached the max number of
                             seconds for the soft limit to be considered
                             reached. */
            }
        }
    } else {
        c->obuf_soft_limit_reached_time = 0;
    }
    return soft || hard;
}

/* Asynchronously close a client if soft or hard limit is reached on the
1650 1651
 * output buffer size. The caller can check if the client will be closed
 * checking if the client REDIS_CLOSE_ASAP flag is set.
1652 1653 1654 1655
 *
 * Note: we need to close the client asynchronously because this function is
 * called from contexts where the client can't be freed safely, i.e. from the
 * lower level functions pushing data inside the client output buffers. */
1656
void asyncCloseClientOnOutputBufferLimitReached(redisClient *c) {
1657
    redisAssert(c->reply_bytes < SIZE_MAX-(1024*64));
1658
    if (c->reply_bytes == 0 || c->flags & REDIS_CLOSE_ASAP) return;
1659
    if (checkClientOutputBufferLimits(c)) {
1660
        sds client = catClientInfoString(sdsempty(),c);
1661 1662

        freeClientAsync(c);
A
antirez 已提交
1663
        serverLog(REDIS_WARNING,"Client %s scheduled to be closed ASAP for overcoming of output buffer limits.", client);
1664 1665 1666
        sdsfree(client);
    }
}
A
antirez 已提交
1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687

/* Helper function used by freeMemoryIfNeeded() in order to flush slaves
 * output buffers without returning control to the event loop. */
void flushSlavesOutputBuffers(void) {
    listIter li;
    listNode *ln;

    listRewind(server.slaves,&li);
    while((ln = listNext(&li))) {
        redisClient *slave = listNodeValue(ln);
        int events;

        events = aeGetFileEvents(server.el,slave->fd);
        if (events & AE_WRITABLE &&
            slave->replstate == REDIS_REPL_ONLINE &&
            listLength(slave->reply))
        {
            sendReplyToClient(server.el,slave->fd,slave,0);
        }
    }
}
1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714

/* Pause clients up to the specified unixtime (in ms). While clients
 * are paused no command is processed from clients, so the data set can't
 * change during that time.
 *
 * However while this function pauses normal and Pub/Sub clients, slaves are
 * still served, so this function can be used on server upgrades where it is
 * required that slaves process the latest bytes from the replication stream
 * before being turned to masters.
 *
 * This function is also internally used by Redis Cluster for the manual
 * failover procedure implemented by CLUSTER FAILOVER.
 *
 * The function always succeed, even if there is already a pause in progress.
 * In such a case, the pause is extended if the duration is more than the
 * time left for the previous duration. However if the duration is smaller
 * than the time left for the previous pause, no change is made to the
 * left duration. */
void pauseClients(mstime_t end) {
    if (!server.clients_paused || end > server.clients_pause_end_time)
        server.clients_pause_end_time = end;
    server.clients_paused = 1;
}

/* Return non-zero if clients are currently paused. As a side effect the
 * function checks if the pause time was reached and clear it. */
int clientsArePaused(void) {
1715 1716 1717
    if (server.clients_paused &&
        server.clients_pause_end_time < server.mstime)
    {
1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729
        listNode *ln;
        listIter li;
        redisClient *c;

        server.clients_paused = 0;

        /* Put all the clients in the unblocked clients queue in order to
         * force the re-processing of the input buffer if any. */
        listRewind(server.clients,&li);
        while ((ln = listNext(&li)) != NULL) {
            c = listNodeValue(ln);

1730 1731
            /* Don't touch slaves and blocked clients. The latter pending
             * requests be processed when unblocked. */
1732
            if (c->flags & (REDIS_SLAVE|REDIS_BLOCKED)) continue;
1733
            c->flags |= REDIS_UNBLOCKED;
1734 1735 1736 1737 1738
            listAddNodeTail(server.unblocked_clients,c);
        }
    }
    return server.clients_paused;
}
1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761

/* This function is called by Redis in order to process a few events from
 * time to time while blocked into some not interruptible operation.
 * This allows to reply to clients with the -LOADING error while loading the
 * data set at startup or after a full resynchronization with the master
 * and so forth.
 *
 * It calls the event loop in order to process a few events. Specifically we
 * try to call the event loop for times as long as we receive acknowledge that
 * some event was processed, in order to go forward with the accept, read,
 * write, close sequence needed to serve a client.
 *
 * The function returns the total number of events processed. */
int processEventsWhileBlocked(void) {
    int iterations = 4; /* See the function top-comment. */
    int count = 0;
    while (iterations--) {
        int events = aeProcessEvents(server.el, AE_FILE_EVENTS|AE_DONT_WAIT);
        if (!events) break;
        count += events;
    }
    return count;
}