* POSSIBILITY OF SUCH DAMAGE.
*/
-#define REDIS_VERSION "1.050"
+#define REDIS_VERSION "1.100"
#include "fmacros.h"
#include "config.h"
#include <fcntl.h>
#include <sys/time.h>
#include <sys/resource.h>
+#include <sys/uio.h>
#include <limits.h>
#include <math.h>
#define REDIS_MAX_SYNC_TIME 60 /* Slave can't take more to sync */
#define REDIS_EXPIRELOOKUPS_PER_CRON 100 /* try to expire 100 keys/second */
#define REDIS_MAX_WRITE_PER_EVENT (1024*64)
-#define REDIS_REQUEST_MAX_SIZE (1024*1024*256) /* max bytes in inline command */
+#define REDIS_REQUEST_MAX_SIZE (1024*1024*256) /* max bytes in inline command */
+
+/* If more then REDIS_WRITEV_THRESHOLD write packets are pending use writev */
+#define REDIS_WRITEV_THRESHOLD 3
+/* Max number of iovecs used for each writev call */
+#define REDIS_WRITEV_IOVEC_COUNT 256
/* Hash table parameters */
#define REDIS_HT_MINFILL 10 /* Minimal hash table fill 10% */
int appendfd;
int appendseldb;
char *pidfile;
- int bgsaveinprogress;
pid_t bgsavechildpid;
+ pid_t bgrewritechildpid;
+ sds bgrewritebuf; /* buffer taken by parent during oppend only rewrite */
struct saveparam *saveparams;
int saveparamslen;
char *logfile;
static int syncWithMaster(void);
static robj *tryObjectSharing(robj *o);
static int tryObjectEncoding(robj *o);
-static robj *getDecodedObject(const robj *o);
+static robj *getDecodedObject(robj *o);
static int removeExpire(redisDb *db, robj *key);
static int expireIfNeeded(redisDb *db, robj *key);
static int deleteIfVolatile(redisDb *db, robj *key);
static int processCommand(redisClient *c);
static void setupSigSegvAction(void);
static void rdbRemoveTempFile(pid_t childpid);
+static void aofRemoveTempFile(pid_t childpid);
static size_t stringObjectLen(robj *o);
static void processInputBuffer(redisClient *c);
static zskiplist *zslCreate(void);
static void zslFree(zskiplist *zsl);
static void zslInsert(zskiplist *zsl, double score, robj *obj);
+static void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int mask);
static void authCommand(redisClient *c);
static void pingCommand(redisClient *c);
static void lastsaveCommand(redisClient *c);
static void saveCommand(redisClient *c);
static void bgsaveCommand(redisClient *c);
+static void bgrewriteaofCommand(redisClient *c);
static void shutdownCommand(redisClient *c);
static void moveCommand(redisClient *c);
static void renameCommand(redisClient *c);
static void msetCommand(redisClient *c);
static void msetnxCommand(redisClient *c);
static void zaddCommand(redisClient *c);
+static void zincrbyCommand(redisClient *c);
static void zrangeCommand(redisClient *c);
static void zrangebyscoreCommand(redisClient *c);
static void zrevrangeCommand(redisClient *c);
{"lrange",lrangeCommand,4,REDIS_CMD_INLINE},
{"ltrim",ltrimCommand,4,REDIS_CMD_INLINE},
{"lrem",lremCommand,4,REDIS_CMD_BULK},
- {"rpoplpush",rpoplpushcommand,3,REDIS_CMD_BULK},
+ {"rpoplpush",rpoplpushcommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
{"sadd",saddCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
{"srem",sremCommand,3,REDIS_CMD_BULK},
{"smove",smoveCommand,4,REDIS_CMD_BULK},
{"sdiffstore",sdiffstoreCommand,-3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
{"smembers",sinterCommand,2,REDIS_CMD_INLINE},
{"zadd",zaddCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
+ {"zincrby",zincrbyCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
{"zrem",zremCommand,3,REDIS_CMD_BULK},
{"zremrangebyscore",zremrangebyscoreCommand,4,REDIS_CMD_INLINE},
{"zrange",zrangeCommand,4,REDIS_CMD_INLINE},
- {"zrangebyscore",zrangebyscoreCommand,4,REDIS_CMD_INLINE},
+ {"zrangebyscore",zrangebyscoreCommand,-4,REDIS_CMD_INLINE},
{"zrevrange",zrevrangeCommand,4,REDIS_CMD_INLINE},
{"zcard",zcardCommand,2,REDIS_CMD_INLINE},
{"zscore",zscoreCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
{"echo",echoCommand,2,REDIS_CMD_BULK},
{"save",saveCommand,1,REDIS_CMD_INLINE},
{"bgsave",bgsaveCommand,1,REDIS_CMD_INLINE},
+ {"bgrewriteaof",bgrewriteaofCommand,1,REDIS_CMD_INLINE},
{"shutdown",shutdownCommand,1,REDIS_CMD_INLINE},
{"lastsave",lastsaveCommand,1,REDIS_CMD_INLINE},
{"type",typeCommand,2,REDIS_CMD_INLINE},
static int dictEncObjKeyCompare(void *privdata, const void *key1,
const void *key2)
{
- const robj *o1 = key1, *o2 = key2;
-
- if (o1->encoding == REDIS_ENCODING_RAW &&
- o2->encoding == REDIS_ENCODING_RAW)
- return sdsDictKeyCompare(privdata,o1->ptr,o2->ptr);
- else {
- robj *dec1, *dec2;
- int cmp;
+ robj *o1 = (robj*) key1, *o2 = (robj*) key2;
+ int cmp;
- dec1 = o1->encoding != REDIS_ENCODING_RAW ?
- getDecodedObject(o1) : (robj*)o1;
- dec2 = o2->encoding != REDIS_ENCODING_RAW ?
- getDecodedObject(o2) : (robj*)o2;
- cmp = sdsDictKeyCompare(privdata,dec1->ptr,dec2->ptr);
- if (dec1 != o1) decrRefCount(dec1);
- if (dec2 != o2) decrRefCount(dec2);
- return cmp;
- }
+ o1 = getDecodedObject(o1);
+ o2 = getDecodedObject(o2);
+ cmp = sdsDictKeyCompare(privdata,o1->ptr,o2->ptr);
+ decrRefCount(o1);
+ decrRefCount(o2);
+ return cmp;
}
static unsigned int dictEncObjHash(const void *key) {
- const robj *o = key;
+ robj *o = (robj*) key;
- if (o->encoding == REDIS_ENCODING_RAW)
- return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
- else {
- robj *dec = getDecodedObject(o);
- unsigned int hash = dictGenHashFunction(dec->ptr, sdslen((sds)dec->ptr));
- decrRefCount(dec);
- return hash;
- }
+ o = getDecodedObject(o);
+ unsigned int hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
+ decrRefCount(o);
+ return hash;
}
static dictType setDictType = {
}
}
+/* A background saving child (BGSAVE) terminated its work. Handle this. */
+void backgroundSaveDoneHandler(int statloc) {
+ int exitcode = WEXITSTATUS(statloc);
+ int bysignal = WIFSIGNALED(statloc);
+
+ if (!bysignal && exitcode == 0) {
+ redisLog(REDIS_NOTICE,
+ "Background saving terminated with success");
+ server.dirty = 0;
+ server.lastsave = time(NULL);
+ } else if (!bysignal && exitcode != 0) {
+ redisLog(REDIS_WARNING, "Background saving error");
+ } else {
+ redisLog(REDIS_WARNING,
+ "Background saving terminated by signal");
+ rdbRemoveTempFile(server.bgsavechildpid);
+ }
+ server.bgsavechildpid = -1;
+ /* Possibly there are slaves waiting for a BGSAVE in order to be served
+ * (the first stage of SYNC is a bulk transfer of dump.rdb) */
+ updateSlavesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR);
+}
+
+/* A background append only file rewriting (BGREWRITEAOF) terminated its work.
+ * Handle this. */
+void backgroundRewriteDoneHandler(int statloc) {
+ int exitcode = WEXITSTATUS(statloc);
+ int bysignal = WIFSIGNALED(statloc);
+
+ if (!bysignal && exitcode == 0) {
+ int fd;
+ char tmpfile[256];
+
+ redisLog(REDIS_NOTICE,
+ "Background append only file rewriting terminated with success");
+ /* Now it's time to flush the differences accumulated by the parent */
+ snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) server.bgrewritechildpid);
+ fd = open(tmpfile,O_WRONLY|O_APPEND);
+ if (fd == -1) {
+ redisLog(REDIS_WARNING, "Not able to open the temp append only file produced by the child: %s", strerror(errno));
+ goto cleanup;
+ }
+ /* Flush our data... */
+ if (write(fd,server.bgrewritebuf,sdslen(server.bgrewritebuf)) !=
+ (signed) sdslen(server.bgrewritebuf)) {
+ redisLog(REDIS_WARNING, "Error or short write trying to flush the parent diff of the append log file in the child temp file: %s", strerror(errno));
+ close(fd);
+ goto cleanup;
+ }
+ redisLog(REDIS_WARNING,"Parent diff flushed into the new append log file with success");
+ /* Now our work is to rename the temp file into the stable file. And
+ * switch the file descriptor used by the server for append only. */
+ if (rename(tmpfile,server.appendfilename) == -1) {
+ redisLog(REDIS_WARNING,"Can't rename the temp append only file into the stable one: %s", strerror(errno));
+ close(fd);
+ goto cleanup;
+ }
+ /* Mission completed... almost */
+ redisLog(REDIS_NOTICE,"Append only file successfully rewritten.");
+ if (server.appendfd != -1) {
+ /* If append only is actually enabled... */
+ close(server.appendfd);
+ server.appendfd = fd;
+ fsync(fd);
+ server.appendseldb = -1; /* Make sure it will issue SELECT */
+ redisLog(REDIS_NOTICE,"The new append only file was selected for future appends.");
+ } else {
+ /* If append only is disabled we just generate a dump in this
+ * format. Why not? */
+ close(fd);
+ }
+ } else if (!bysignal && exitcode != 0) {
+ redisLog(REDIS_WARNING, "Background append only file rewriting error");
+ } else {
+ redisLog(REDIS_WARNING,
+ "Background append only file rewriting terminated by signal");
+ }
+cleanup:
+ sdsfree(server.bgrewritebuf);
+ server.bgrewritebuf = sdsempty();
+ aofRemoveTempFile(server.bgrewritechildpid);
+ server.bgrewritechildpid = -1;
+}
+
static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
int j, loops = server.cronloops++;
REDIS_NOTUSED(eventLoop);
* if we resize the HT while there is the saving child at work actually
* a lot of memory movements in the parent will cause a lot of pages
* copied. */
- if (!server.bgsaveinprogress) tryResizeHashTables();
+ if (server.bgsavechildpid == -1) tryResizeHashTables();
/* Show information about connected clients */
if (!(loops % 5)) {
if (server.maxidletime && !(loops % 10))
closeTimedoutClients();
- /* Check if a background saving in progress terminated */
- if (server.bgsaveinprogress) {
+ /* Check if a background saving or AOF rewrite in progress terminated */
+ if (server.bgsavechildpid != -1 || server.bgrewritechildpid != -1) {
int statloc;
- if (wait4(-1,&statloc,WNOHANG,NULL)) {
- int exitcode = WEXITSTATUS(statloc);
- int bysignal = WIFSIGNALED(statloc);
-
- if (!bysignal && exitcode == 0) {
- redisLog(REDIS_NOTICE,
- "Background saving terminated with success");
- server.dirty = 0;
- server.lastsave = time(NULL);
- } else if (!bysignal && exitcode != 0) {
- redisLog(REDIS_WARNING, "Background saving error");
+ pid_t pid;
+
+ if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
+ if (pid == server.bgsavechildpid) {
+ backgroundSaveDoneHandler(statloc);
} else {
- redisLog(REDIS_WARNING,
- "Background saving terminated by signal");
- rdbRemoveTempFile(server.bgsavechildpid);
+ backgroundRewriteDoneHandler(statloc);
}
- server.bgsaveinprogress = 0;
- server.bgsavechildpid = -1;
- updateSlavesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR);
}
} else {
/* If there is not a background saving in progress check if
server.appendseldb = -1; /* Make sure the first time will not match */
server.pidfile = "/var/run/redis.pid";
server.dbfilename = "dump.rdb";
- server.appendfilename = "appendonly.log";
+ server.appendfilename = "appendonly.aof";
server.requirepass = NULL;
server.shareobjects = 0;
server.sharingpoolsize = 1024;
server.db[j].id = j;
}
server.cronloops = 0;
- server.bgsaveinprogress = 0;
server.bgsavechildpid = -1;
+ server.bgrewritechildpid = -1;
+ server.bgrewritebuf = sdsempty();
server.lastsave = time(NULL);
server.dirty = 0;
server.usedmemory = 0;
zfree(c);
}
+#define GLUEREPLY_UP_TO (1024)
static void glueReplyBuffersIfNeeded(redisClient *c) {
- int totlen = 0;
+ int copylen = 0;
+ char buf[GLUEREPLY_UP_TO];
listNode *ln;
robj *o;
listRewind(c->reply);
while((ln = listYield(c->reply))) {
+ int objlen;
+
o = ln->value;
- totlen += sdslen(o->ptr);
- /* This optimization makes more sense if we don't have to copy
- * too much data */
- if (totlen > 1024) return;
- }
- if (totlen > 0) {
- char buf[1024];
- int copylen = 0;
-
- listRewind(c->reply);
- while((ln = listYield(c->reply))) {
- o = ln->value;
- memcpy(buf+copylen,o->ptr,sdslen(o->ptr));
- copylen += sdslen(o->ptr);
+ objlen = sdslen(o->ptr);
+ if (copylen + objlen <= GLUEREPLY_UP_TO) {
+ memcpy(buf+copylen,o->ptr,objlen);
+ copylen += objlen;
listDelNode(c->reply,ln);
+ } else {
+ if (copylen == 0) return;
+ break;
}
- /* Now the output buffer is empty, add the new single element */
- o = createObject(REDIS_STRING,sdsnewlen(buf,totlen));
- listAddNodeTail(c->reply,o);
}
+ /* Now the output buffer is empty, add the new single element */
+ o = createObject(REDIS_STRING,sdsnewlen(buf,copylen));
+ listAddNodeHead(c->reply,o);
}
static void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) {
REDIS_NOTUSED(el);
REDIS_NOTUSED(mask);
- if (server.glueoutputbuf && listLength(c->reply) > 1)
- glueReplyBuffersIfNeeded(c);
+ /* Use writev() if we have enough buffers to send */
+ if (!server.glueoutputbuf &&
+ listLength(c->reply) > REDIS_WRITEV_THRESHOLD &&
+ !(c->flags & REDIS_MASTER))
+ {
+ sendReplyToClientWritev(el, fd, privdata, mask);
+ return;
+ }
+
while(listLength(c->reply)) {
+ if (server.glueoutputbuf && listLength(c->reply) > 1)
+ glueReplyBuffersIfNeeded(c);
+
o = listNodeValue(listFirst(c->reply));
objlen = sdslen(o->ptr);
}
}
+static void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int mask)
+{
+ redisClient *c = privdata;
+ int nwritten = 0, totwritten = 0, objlen, willwrite;
+ robj *o;
+ struct iovec iov[REDIS_WRITEV_IOVEC_COUNT];
+ int offset, ion = 0;
+ REDIS_NOTUSED(el);
+ REDIS_NOTUSED(mask);
+
+ listNode *node;
+ while (listLength(c->reply)) {
+ offset = c->sentlen;
+ ion = 0;
+ willwrite = 0;
+
+ /* fill-in the iov[] array */
+ for(node = listFirst(c->reply); node; node = listNextNode(node)) {
+ o = listNodeValue(node);
+ objlen = sdslen(o->ptr);
+
+ if (totwritten + objlen - offset > REDIS_MAX_WRITE_PER_EVENT)
+ break;
+
+ if(ion == REDIS_WRITEV_IOVEC_COUNT)
+ break; /* no more iovecs */
+
+ iov[ion].iov_base = ((char*)o->ptr) + offset;
+ iov[ion].iov_len = objlen - offset;
+ willwrite += objlen - offset;
+ offset = 0; /* just for the first item */
+ ion++;
+ }
+
+ if(willwrite == 0)
+ break;
+
+ /* write all collected blocks at once */
+ if((nwritten = writev(fd, iov, ion)) < 0) {
+ if (errno != EAGAIN) {
+ redisLog(REDIS_DEBUG,
+ "Error writing to client: %s", strerror(errno));
+ freeClient(c);
+ return;
+ }
+ break;
+ }
+
+ totwritten += nwritten;
+ offset = c->sentlen;
+
+ /* remove written robjs from c->reply */
+ while (nwritten && listLength(c->reply)) {
+ o = listNodeValue(listFirst(c->reply));
+ objlen = sdslen(o->ptr);
+
+ if(nwritten >= objlen - offset) {
+ listDelNode(c->reply, listFirst(c->reply));
+ nwritten -= objlen - offset;
+ c->sentlen = 0;
+ } else {
+ /* partial write */
+ c->sentlen += nwritten;
+ break;
+ }
+ offset = 0;
+ }
+ }
+
+ if (totwritten > 0)
+ c->lastinteraction = time(NULL);
+
+ if (listLength(c->reply) == 0) {
+ c->sentlen = 0;
+ aeDeleteFileEvent(server.el,c->fd,AE_WRITABLE);
+ }
+}
+
static struct redisCommand *lookupCommand(char *name) {
int j = 0;
while(cmdTable[j].name != NULL) {
listSetFreeMethod(c->reply,decrRefCount);
listSetDupMethod(c->reply,dupClientReplyValue);
if (aeCreateFileEvent(server.el, c->fd, AE_READABLE,
- readQueryFromClient, c, NULL) == AE_ERR) {
+ readQueryFromClient, c) == AE_ERR) {
freeClient(c);
return NULL;
}
(c->replstate == REDIS_REPL_NONE ||
c->replstate == REDIS_REPL_ONLINE) &&
aeCreateFileEvent(server.el, c->fd, AE_WRITABLE,
- sendReplyToClient, c, NULL) == AE_ERR) return;
- if (obj->encoding != REDIS_ENCODING_RAW) {
- obj = getDecodedObject(obj);
- } else {
- incrRefCount(obj);
- }
- listAddNodeTail(c->reply,obj);
+ sendReplyToClient, c) == AE_ERR) return;
+ listAddNodeTail(c->reply,getDecodedObject(obj));
}
static void addReplySds(redisClient *c, sds s) {
decrRefCount(o);
}
+static void addReplyDouble(redisClient *c, double d) {
+ char buf[128];
+
+ snprintf(buf,sizeof(buf),"%.17g",d);
+ addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n%s\r\n",
+ strlen(buf),buf));
+}
+
static void addReplyBulkLen(redisClient *c, robj *obj) {
size_t len;
return REDIS_OK;
}
-/* Get a decoded version of an encoded object (returned as a new object) */
-static robj *getDecodedObject(const robj *o) {
+/* Get a decoded version of an encoded object (returned as a new object).
+ * If the object is already raw-encoded just increment the ref count. */
+static robj *getDecodedObject(robj *o) {
robj *dec;
- assert(o->encoding != REDIS_ENCODING_RAW);
+ if (o->encoding == REDIS_ENCODING_RAW) {
+ incrRefCount(o);
+ return o;
+ }
if (o->type == REDIS_STRING && o->encoding == REDIS_ENCODING_INT) {
char buf[32];
/* Compare two string objects via strcmp() or alike.
* Note that the objects may be integer-encoded. In such a case we
* use snprintf() to get a string representation of the numbers on the stack
- * and compare the strings, it's much faster than calling getDecodedObject(). */
+ * and compare the strings, it's much faster than calling getDecodedObject().
+ *
+ * Important note: if objects are not integer encoded, but binary-safe strings,
+ * sdscmp() from sds.c will apply memcmp() so this function ca be considered
+ * binary safe. */
static int compareStringObjects(robj *a, robj *b) {
assert(a->type == REDIS_STRING && b->type == REDIS_STRING);
char bufa[128], bufb[128], *astr, *bstr;
/* Like rdbSaveStringObjectRaw() but handle encoded objects */
static int rdbSaveStringObject(FILE *fp, robj *obj) {
int retval;
- robj *dec;
- if (obj->encoding != REDIS_ENCODING_RAW) {
- dec = getDecodedObject(obj);
- retval = rdbSaveStringObjectRaw(fp,dec);
- decrRefCount(dec);
- return retval;
- } else {
- return rdbSaveStringObjectRaw(fp,obj);
- }
+ obj = getDecodedObject(obj);
+ retval = rdbSaveStringObjectRaw(fp,obj);
+ decrRefCount(obj);
+ return retval;
}
/* Save a double value. Doubles are saved as strings prefixed by an unsigned
buf[0] = (val < 0) ? 255 : 254;
} else {
snprintf((char*)buf+1,sizeof(buf)-1,"%.17g",val);
- buf[0] = strlen((char*)buf);
+ buf[0] = strlen((char*)buf+1);
len = buf[0]+1;
}
if (fwrite(buf,len,1,fp) == 0) return -1;
static int rdbSaveBackground(char *filename) {
pid_t childpid;
- if (server.bgsaveinprogress) return REDIS_ERR;
+ if (server.bgsavechildpid != -1) return REDIS_ERR;
if ((childpid = fork()) == 0) {
/* Child */
close(server.fd);
return REDIS_ERR;
}
redisLog(REDIS_NOTICE,"Background saving started by pid %d",childpid);
- server.bgsaveinprogress = 1;
server.bgsavechildpid = childpid;
return REDIS_OK;
}
}
}
+static void msetGenericCommand(redisClient *c, int nx) {
+ int j;
+
+ if ((c->argc % 2) == 0) {
+ addReplySds(c,sdsnew("-ERR wrong number of arguments\r\n"));
+ return;
+ }
+ /* Handle the NX flag. The MSETNX semantic is to return zero and don't
+ * set nothing at all if at least one already key exists. */
+ if (nx) {
+ for (j = 1; j < c->argc; j += 2) {
+ if (dictFind(c->db->dict,c->argv[j]) != NULL) {
+ addReply(c, shared.czero);
+ return;
+ }
+ }
+ }
+
+ for (j = 1; j < c->argc; j += 2) {
+ int retval;
+
+ tryObjectEncoding(c->argv[j+1]);
+ retval = dictAdd(c->db->dict,c->argv[j],c->argv[j+1]);
+ if (retval == DICT_ERR) {
+ dictReplace(c->db->dict,c->argv[j],c->argv[j+1]);
+ incrRefCount(c->argv[j+1]);
+ } else {
+ incrRefCount(c->argv[j]);
+ incrRefCount(c->argv[j+1]);
+ }
+ removeExpire(c->db,c->argv[j]);
+ }
+ server.dirty += (c->argc-1)/2;
+ addReply(c, nx ? shared.cone : shared.ok);
+}
+
+static void msetCommand(redisClient *c) {
+ msetGenericCommand(c,0);
+}
+
+static void msetnxCommand(redisClient *c) {
+ msetGenericCommand(c,1);
+}
+
static void incrDecrCommand(redisClient *c, long long incr) {
long long value;
int retval;
}
static void saveCommand(redisClient *c) {
- if (server.bgsaveinprogress) {
+ if (server.bgsavechildpid != -1) {
addReplySds(c,sdsnew("-ERR background save in progress\r\n"));
return;
}
}
static void bgsaveCommand(redisClient *c) {
- if (server.bgsaveinprogress) {
+ if (server.bgsavechildpid != -1) {
addReplySds(c,sdsnew("-ERR background save already in progress\r\n"));
return;
}
/* Kill the saving child if there is a background saving in progress.
We want to avoid race conditions, for instance our saving child may
overwrite the synchronous saving did by SHUTDOWN. */
- if (server.bgsaveinprogress) {
+ if (server.bgsavechildpid != -1) {
redisLog(REDIS_WARNING,"There is a live saving child. Killing it!");
kill(server.bgsavechildpid,SIGKILL);
rdbRemoveTempFile(server.bgsavechildpid);
/* The actual Z-commands implementations */
-static void zaddCommand(redisClient *c) {
+/* This generic command implements both ZADD and ZINCRBY.
+ * scoreval is the score if the operation is a ZADD (doincrement == 0) or
+ * the increment if the operation is a ZINCRBY (doincrement == 1). */
+static void zaddGenericCommand(redisClient *c, robj *key, robj *ele, double scoreval, int doincrement) {
robj *zsetobj;
zset *zs;
double *score;
- zsetobj = lookupKeyWrite(c->db,c->argv[1]);
+ zsetobj = lookupKeyWrite(c->db,key);
if (zsetobj == NULL) {
zsetobj = createZsetObject();
- dictAdd(c->db->dict,c->argv[1],zsetobj);
- incrRefCount(c->argv[1]);
+ dictAdd(c->db->dict,key,zsetobj);
+ incrRefCount(key);
} else {
if (zsetobj->type != REDIS_ZSET) {
addReply(c,shared.wrongtypeerr);
return;
}
}
- score = zmalloc(sizeof(double));
- *score = strtod(c->argv[2]->ptr,NULL);
zs = zsetobj->ptr;
- if (dictAdd(zs->dict,c->argv[3],score) == DICT_OK) {
+
+ /* Ok now since we implement both ZADD and ZINCRBY here the code
+ * needs to handle the two different conditions. It's all about setting
+ * '*score', that is, the new score to set, to the right value. */
+ score = zmalloc(sizeof(double));
+ if (doincrement) {
+ dictEntry *de;
+
+ /* Read the old score. If the element was not present starts from 0 */
+ de = dictFind(zs->dict,ele);
+ if (de) {
+ double *oldscore = dictGetEntryVal(de);
+ *score = *oldscore + scoreval;
+ } else {
+ *score = scoreval;
+ }
+ } else {
+ *score = scoreval;
+ }
+
+ /* What follows is a simple remove and re-insert operation that is common
+ * to both ZADD and ZINCRBY... */
+ if (dictAdd(zs->dict,ele,score) == DICT_OK) {
/* case 1: New element */
- incrRefCount(c->argv[3]); /* added to hash */
- zslInsert(zs->zsl,*score,c->argv[3]);
- incrRefCount(c->argv[3]); /* added to skiplist */
+ incrRefCount(ele); /* added to hash */
+ zslInsert(zs->zsl,*score,ele);
+ incrRefCount(ele); /* added to skiplist */
server.dirty++;
- addReply(c,shared.cone);
+ if (doincrement)
+ addReplyDouble(c,*score);
+ else
+ addReply(c,shared.cone);
} else {
dictEntry *de;
double *oldscore;
/* case 2: Score update operation */
- de = dictFind(zs->dict,c->argv[3]);
+ de = dictFind(zs->dict,ele);
assert(de != NULL);
oldscore = dictGetEntryVal(de);
if (*score != *oldscore) {
int deleted;
- deleted = zslDelete(zs->zsl,*oldscore,c->argv[3]);
+ /* Remove and insert the element in the skip list with new score */
+ deleted = zslDelete(zs->zsl,*oldscore,ele);
assert(deleted != 0);
- zslInsert(zs->zsl,*score,c->argv[3]);
- incrRefCount(c->argv[3]);
- dictReplace(zs->dict,c->argv[3],score);
+ zslInsert(zs->zsl,*score,ele);
+ incrRefCount(ele);
+ /* Update the score in the hash table */
+ dictReplace(zs->dict,ele,score);
server.dirty++;
} else {
zfree(score);
}
- addReply(c,shared.czero);
+ if (doincrement)
+ addReplyDouble(c,*score);
+ else
+ addReply(c,shared.czero);
}
}
+static void zaddCommand(redisClient *c) {
+ double scoreval;
+
+ scoreval = strtod(c->argv[2]->ptr,NULL);
+ zaddGenericCommand(c,c->argv[1],c->argv[3],scoreval,0);
+}
+
+static void zincrbyCommand(redisClient *c) {
+ double scoreval;
+
+ scoreval = strtod(c->argv[2]->ptr,NULL);
+ zaddGenericCommand(c,c->argv[1],c->argv[3],scoreval,1);
+}
+
static void zremCommand(redisClient *c) {
robj *zsetobj;
zset *zs;
robj *o;
double min = strtod(c->argv[2]->ptr,NULL);
double max = strtod(c->argv[3]->ptr,NULL);
+ int offset = 0, limit = -1;
+
+ if (c->argc != 4 && c->argc != 7) {
+ addReplySds(c,sdsnew("-ERR wrong number of arguments\r\n"));
+ return;
+ } else if (c->argc == 7 && strcasecmp(c->argv[4]->ptr,"limit")) {
+ addReply(c,shared.syntaxerr);
+ return;
+ } else if (c->argc == 7) {
+ offset = atoi(c->argv[5]->ptr);
+ limit = atoi(c->argv[6]->ptr);
+ if (offset < 0) offset = 0;
+ }
o = lookupKeyRead(c->db,c->argv[1]);
if (o == NULL) {
* it later */
lenobj = createObject(REDIS_STRING,NULL);
addReply(c,lenobj);
+ decrRefCount(lenobj);
while(ln && ln->score <= max) {
+ if (offset) {
+ offset--;
+ ln = ln->forward[0];
+ continue;
+ }
+ if (limit == 0) break;
ele = ln->obj;
addReplyBulkLen(c,ele);
addReply(c,ele);
addReply(c,shared.crlf);
ln = ln->forward[0];
rangelen++;
+ if (limit > 0) limit--;
}
lenobj->ptr = sdscatprintf(sdsempty(),"*%d\r\n",rangelen);
}
if (!de) {
addReply(c,shared.nullbulk);
} else {
- char buf[128];
double *score = dictGetEntryVal(de);
- snprintf(buf,sizeof(buf),"%.17g",*score);
- addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n%s\r\n",
- strlen(buf),buf));
+ addReplyDouble(c,*score);
}
}
}
char buf[REDIS_SORTKEY_MAX+1];
} keyname;
- if (subst->encoding == REDIS_ENCODING_RAW)
- incrRefCount(subst);
- else {
- subst = getDecodedObject(subst);
+ /* If the pattern is "#" return the substitution object itself in order
+ * to implement the "SORT ... GET #" feature. */
+ spat = pattern->ptr;
+ if (spat[0] == '#' && spat[1] == '\0') {
+ return subst;
}
- spat = pattern->ptr;
+ /* The substitution object may be specially encoded. If so we create
+ * a decoded object on the fly. Otherwise getDecodedObject will just
+ * increment the ref count, that we'll decrement later. */
+ subst = getDecodedObject(subst);
+
ssub = subst->ptr;
if (sdslen(spat)+sdslen(ssub)-1 > REDIS_SORTKEY_MAX) return NULL;
p = strchr(spat,'*');
- if (!p) return NULL;
+ if (!p) {
+ decrRefCount(subst);
+ return NULL;
+ }
prefixlen = p-spat;
sublen = sdslen(ssub);
}
} else {
/* Compare elements directly */
- if (so1->obj->encoding == REDIS_ENCODING_RAW &&
- so2->obj->encoding == REDIS_ENCODING_RAW) {
- cmp = strcoll(so1->obj->ptr,so2->obj->ptr);
- } else {
- robj *dec1, *dec2;
-
- dec1 = so1->obj->encoding == REDIS_ENCODING_RAW ?
- so1->obj : getDecodedObject(so1->obj);
- dec2 = so2->obj->encoding == REDIS_ENCODING_RAW ?
- so2->obj : getDecodedObject(so2->obj);
- cmp = strcoll(dec1->ptr,dec2->ptr);
- if (dec1 != so1->obj) decrRefCount(dec1);
- if (dec2 != so2->obj) decrRefCount(dec2);
- }
+ robj *dec1, *dec2;
+
+ dec1 = getDecodedObject(so1->obj);
+ dec2 = getDecodedObject(so2->obj);
+ cmp = strcoll(dec1->ptr,dec2->ptr);
+ decrRefCount(dec1);
+ decrRefCount(dec2);
}
}
return server.sort_desc ? -cmp : cmp;
addReply(c,shared.nokeyerr);
return;
}
- if (sortval->type != REDIS_SET && sortval->type != REDIS_LIST) {
+ if (sortval->type != REDIS_SET && sortval->type != REDIS_LIST &&
+ sortval->type != REDIS_ZSET)
+ {
addReply(c,shared.wrongtypeerr);
return;
}
}
/* Load the sorting vector with all the objects to sort */
- vectorlen = (sortval->type == REDIS_LIST) ?
- listLength((list*)sortval->ptr) :
- dictSize((dict*)sortval->ptr);
+ switch(sortval->type) {
+ case REDIS_LIST: vectorlen = listLength((list*)sortval->ptr); break;
+ case REDIS_SET: vectorlen = dictSize((dict*)sortval->ptr); break;
+ case REDIS_ZSET: vectorlen = dictSize(((zset*)sortval->ptr)->dict); break;
+ default: vectorlen = 0; assert(0); /* Avoid GCC warning */
+ }
vector = zmalloc(sizeof(redisSortObject)*vectorlen);
j = 0;
+
if (sortval->type == REDIS_LIST) {
list *list = sortval->ptr;
listNode *ln;
j++;
}
} else {
- dict *set = sortval->ptr;
+ dict *set;
dictIterator *di;
dictEntry *setele;
+ if (sortval->type == REDIS_SET) {
+ set = sortval->ptr;
+ } else {
+ zset *zs = sortval->ptr;
+ set = zs->dict;
+ }
+
di = dictGetIterator(set);
while((setele = dictNext(di)) != NULL) {
vector[j].obj = dictGetEntryKey(setele);
byval = lookupKeyByPattern(c->db,sortby,vector[j].obj);
if (!byval || byval->type != REDIS_STRING) continue;
if (alpha) {
- if (byval->encoding == REDIS_ENCODING_RAW) {
- vector[j].u.cmpobj = byval;
- incrRefCount(byval);
- } else {
- vector[j].u.cmpobj = getDecodedObject(byval);
- }
+ vector[j].u.cmpobj = getDecodedObject(byval);
} else {
if (byval->encoding == REDIS_ENCODING_RAW) {
vector[j].u.score = strtod(byval->ptr,NULL);
} else {
+ /* Don't need to decode the object if it's
+ * integer-encoded (the only encoding supported) so
+ * far. We can just cast it */
if (byval->encoding == REDIS_ENCODING_INT) {
vector[j].u.score = (long)byval->ptr;
} else
zfree(vector);
}
-static void infoCommand(redisClient *c) {
+/* Create the string returned by the INFO command. This is decoupled
+ * by the INFO command itself as we need to report the same information
+ * on memory corruption problems. */
+static sds genRedisInfoString(void) {
sds info;
time_t uptime = time(NULL)-server.stat_starttime;
int j;
info = sdscatprintf(sdsempty(),
"redis_version:%s\r\n"
"arch_bits:%s\r\n"
+ "multiplexing_api:%s\r\n"
"uptime_in_seconds:%d\r\n"
"uptime_in_days:%d\r\n"
"connected_clients:%d\r\n"
"role:%s\r\n"
,REDIS_VERSION,
(sizeof(long) == 8) ? "64" : "32",
+ aeGetApiName(),
uptime,
uptime/(3600*24),
listLength(server.clients)-listLength(server.slaves),
listLength(server.slaves),
server.usedmemory,
server.dirty,
- server.bgsaveinprogress,
+ server.bgsavechildpid != -1,
server.lastsave,
server.stat_numconnections,
server.stat_numcommands,
keys = dictSize(server.db[j].dict);
vkeys = dictSize(server.db[j].expires);
if (keys || vkeys) {
- info = sdscatprintf(info, "db%d: keys=%lld,expires=%lld\r\n",
+ info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
j, keys, vkeys);
}
}
+ return info;
+}
+
+static void infoCommand(redisClient *c) {
+ sds info = genRedisInfoString();
addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",sdslen(info)));
addReplySds(c,info);
addReply(c,shared.crlf);
addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",ttl));
}
-static void msetGenericCommand(redisClient *c, int nx) {
- int j;
-
- if ((c->argc % 2) == 0) {
- addReplySds(c,sdsnew("-ERR wrong number of arguments\r\n"));
- return;
- }
- /* Handle the NX flag. The MSETNX semantic is to return zero and don't
- * set nothing at all if at least one already key exists. */
- if (nx) {
- for (j = 1; j < c->argc; j += 2) {
- if (dictFind(c->db->dict,c->argv[j]) != NULL) {
- addReply(c, shared.czero);
- return;
- }
- }
- }
-
- for (j = 1; j < c->argc; j += 2) {
- int retval;
-
- retval = dictAdd(c->db->dict,c->argv[j],c->argv[j+1]);
- if (retval == DICT_ERR) {
- dictReplace(c->db->dict,c->argv[j],c->argv[j+1]);
- incrRefCount(c->argv[j+1]);
- } else {
- incrRefCount(c->argv[j]);
- incrRefCount(c->argv[j+1]);
- }
- removeExpire(c->db,c->argv[j]);
- }
- server.dirty += (c->argc-1)/2;
- addReply(c, nx ? shared.cone : shared.ok);
-}
-
-static void msetCommand(redisClient *c) {
- msetGenericCommand(c,0);
-}
-
-static void msetnxCommand(redisClient *c) {
- msetGenericCommand(c,1);
-}
-
/* =============================== Replication ============================= */
static int syncWrite(int fd, char *ptr, ssize_t size, int timeout) {
redisLog(REDIS_NOTICE,"Slave ask for synchronization");
/* Here we need to check if there is a background saving operation
* in progress, or if it is required to start one */
- if (server.bgsaveinprogress) {
+ if (server.bgsavechildpid != -1) {
/* Ok a background save is in progress. Let's check if it is a good
* one for replication, i.e. if there is another slave that is
* registering differences since the server forked to save */
aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE);
slave->replstate = REDIS_REPL_ONLINE;
if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE,
- sendReplyToClient, slave, NULL) == AE_ERR) {
+ sendReplyToClient, slave) == AE_ERR) {
freeClient(slave);
return;
}
slave->repldbsize = buf.st_size;
slave->replstate = REDIS_REPL_SEND_BULK;
aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE);
- if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, sendBulkToSlave, slave, NULL) == AE_ERR) {
+ if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, sendBulkToSlave, slave) == AE_ERR) {
freeClient(slave);
continue;
}
for (j = 0; j < argc; j++) {
robj *o = argv[j];
- if (o->encoding != REDIS_ENCODING_RAW)
- o = getDecodedObject(o);
+ o = getDecodedObject(o);
buf = sdscatprintf(buf,"$%d\r\n",sdslen(o->ptr));
buf = sdscatlen(buf,o->ptr,sdslen(o->ptr));
buf = sdscatlen(buf,"\r\n",2);
- if (o != argv[j])
- decrRefCount(o);
+ decrRefCount(o);
}
/* Free the objects from the modified argv for EXPIREAT */
}
exit(1);
}
+ /* If a background append only file rewriting is in progress we want to
+ * accumulate the differences between the child DB and the current one
+ * in a buffer, so that when the child process will do its work we
+ * can append the differences to the new append only file. */
+ if (server.bgrewritechildpid != -1)
+ server.bgrewritebuf = sdscatlen(server.bgrewritebuf,buf,sdslen(buf));
+
+ sdsfree(buf);
now = time(NULL);
if (server.appendfsync == APPENDFSYNC_ALWAYS ||
(server.appendfsync == APPENDFSYNC_EVERYSEC &&
if (buf[0] != '$') goto fmterr;
len = strtol(buf+1,NULL,10);
argsds = sdsnewlen(NULL,len);
- if (fread(argsds,len,1,fp) == 0) goto fmterr;
+ if (len && fread(argsds,len,1,fp) == 0) goto fmterr;
argv[j] = createObject(REDIS_STRING,argsds);
if (fread(buf,2,1,fp) == 0) goto fmterr; /* discard CRLF */
}
exit(1);
}
+/* Write an object into a file in the bulk format $<count>\r\n<payload>\r\n */
+static int fwriteBulk(FILE *fp, robj *obj) {
+ char buf[128];
+ obj = getDecodedObject(obj);
+ snprintf(buf,sizeof(buf),"$%ld\r\n",(long)sdslen(obj->ptr));
+ if (fwrite(buf,strlen(buf),1,fp) == 0) goto err;
+ if (fwrite(obj->ptr,sdslen(obj->ptr),1,fp) == 0) goto err;
+ if (fwrite("\r\n",2,1,fp) == 0) goto err;
+ decrRefCount(obj);
+ return 1;
+err:
+ decrRefCount(obj);
+ return 0;
+}
+
+/* Write a double value in bulk format $<count>\r\n<payload>\r\n */
+static int fwriteBulkDouble(FILE *fp, double d) {
+ char buf[128], dbuf[128];
+
+ snprintf(dbuf,sizeof(dbuf),"%.17g\r\n",d);
+ snprintf(buf,sizeof(buf),"$%lu\r\n",(unsigned long)strlen(dbuf)-2);
+ if (fwrite(buf,strlen(buf),1,fp) == 0) return 0;
+ if (fwrite(dbuf,strlen(dbuf),1,fp) == 0) return 0;
+ return 1;
+}
+
+/* Write a long value in bulk format $<count>\r\n<payload>\r\n */
+static int fwriteBulkLong(FILE *fp, long l) {
+ char buf[128], lbuf[128];
+
+ snprintf(lbuf,sizeof(lbuf),"%ld\r\n",l);
+ snprintf(buf,sizeof(buf),"$%lu\r\n",(unsigned long)strlen(lbuf)-2);
+ if (fwrite(buf,strlen(buf),1,fp) == 0) return 0;
+ if (fwrite(lbuf,strlen(lbuf),1,fp) == 0) return 0;
+ return 1;
+}
+
+/* Write a sequence of commands able to fully rebuild the dataset into
+ * "filename". Used both by REWRITEAOF and BGREWRITEAOF. */
+static int rewriteAppendOnlyFile(char *filename) {
+ dictIterator *di = NULL;
+ dictEntry *de;
+ FILE *fp;
+ char tmpfile[256];
+ int j;
+ time_t now = time(NULL);
+
+ /* Note that we have to use a different temp name here compared to the
+ * one used by rewriteAppendOnlyFileBackground() function. */
+ snprintf(tmpfile,256,"temp-rewriteaof-%d.aof", (int) getpid());
+ fp = fopen(tmpfile,"w");
+ if (!fp) {
+ redisLog(REDIS_WARNING, "Failed rewriting the append only file: %s", strerror(errno));
+ return REDIS_ERR;
+ }
+ for (j = 0; j < server.dbnum; j++) {
+ char selectcmd[] = "*2\r\n$6\r\nSELECT\r\n";
+ redisDb *db = server.db+j;
+ dict *d = db->dict;
+ if (dictSize(d) == 0) continue;
+ di = dictGetIterator(d);
+ if (!di) {
+ fclose(fp);
+ return REDIS_ERR;
+ }
+
+ /* SELECT the new DB */
+ if (fwrite(selectcmd,sizeof(selectcmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulkLong(fp,j) == 0) goto werr;
+
+ /* Iterate this DB writing every entry */
+ while((de = dictNext(di)) != NULL) {
+ robj *key = dictGetEntryKey(de);
+ robj *o = dictGetEntryVal(de);
+ time_t expiretime = getExpire(db,key);
+
+ /* Save the key and associated value */
+ if (o->type == REDIS_STRING) {
+ /* Emit a SET command */
+ char cmd[]="*3\r\n$3\r\nSET\r\n";
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ /* Key and value */
+ if (fwriteBulk(fp,key) == 0) goto werr;
+ if (fwriteBulk(fp,o) == 0) goto werr;
+ } else if (o->type == REDIS_LIST) {
+ /* Emit the RPUSHes needed to rebuild the list */
+ list *list = o->ptr;
+ listNode *ln;
+
+ listRewind(list);
+ while((ln = listYield(list))) {
+ char cmd[]="*3\r\n$5\r\nRPUSH\r\n";
+ robj *eleobj = listNodeValue(ln);
+
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulk(fp,key) == 0) goto werr;
+ if (fwriteBulk(fp,eleobj) == 0) goto werr;
+ }
+ } else if (o->type == REDIS_SET) {
+ /* Emit the SADDs needed to rebuild the set */
+ dict *set = o->ptr;
+ dictIterator *di = dictGetIterator(set);
+ dictEntry *de;
+
+ while((de = dictNext(di)) != NULL) {
+ char cmd[]="*3\r\n$4\r\nSADD\r\n";
+ robj *eleobj = dictGetEntryKey(de);
+
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulk(fp,key) == 0) goto werr;
+ if (fwriteBulk(fp,eleobj) == 0) goto werr;
+ }
+ dictReleaseIterator(di);
+ } else if (o->type == REDIS_ZSET) {
+ /* Emit the ZADDs needed to rebuild the sorted set */
+ zset *zs = o->ptr;
+ dictIterator *di = dictGetIterator(zs->dict);
+ dictEntry *de;
+
+ while((de = dictNext(di)) != NULL) {
+ char cmd[]="*4\r\n$4\r\nZADD\r\n";
+ robj *eleobj = dictGetEntryKey(de);
+ double *score = dictGetEntryVal(de);
+
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulk(fp,key) == 0) goto werr;
+ if (fwriteBulkDouble(fp,*score) == 0) goto werr;
+ if (fwriteBulk(fp,eleobj) == 0) goto werr;
+ }
+ dictReleaseIterator(di);
+ } else {
+ assert(0 != 0);
+ }
+ /* Save the expire time */
+ if (expiretime != -1) {
+ char cmd[]="*3\r\n$6\r\nEXPIRE\r\n";
+ /* If this key is already expired skip it */
+ if (expiretime < now) continue;
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulk(fp,key) == 0) goto werr;
+ if (fwriteBulkLong(fp,expiretime) == 0) goto werr;
+ }
+ }
+ dictReleaseIterator(di);
+ }
+
+ /* Make sure data will not remain on the OS's output buffers */
+ fflush(fp);
+ fsync(fileno(fp));
+ fclose(fp);
+
+ /* Use RENAME to make sure the DB file is changed atomically only
+ * if the generate DB file is ok. */
+ if (rename(tmpfile,filename) == -1) {
+ redisLog(REDIS_WARNING,"Error moving temp append only file on the final destination: %s", strerror(errno));
+ unlink(tmpfile);
+ return REDIS_ERR;
+ }
+ redisLog(REDIS_NOTICE,"SYNC append only file rewrite performed");
+ return REDIS_OK;
+
+werr:
+ fclose(fp);
+ unlink(tmpfile);
+ redisLog(REDIS_WARNING,"Write error writing append only fileon disk: %s", strerror(errno));
+ if (di) dictReleaseIterator(di);
+ return REDIS_ERR;
+}
+
+/* This is how rewriting of the append only file in background works:
+ *
+ * 1) The user calls BGREWRITEAOF
+ * 2) Redis calls this function, that forks():
+ * 2a) the child rewrite the append only file in a temp file.
+ * 2b) the parent accumulates differences in server.bgrewritebuf.
+ * 3) When the child finished '2a' exists.
+ * 4) The parent will trap the exit code, if it's OK, will append the
+ * data accumulated into server.bgrewritebuf into the temp file, and
+ * finally will rename(2) the temp file in the actual file name.
+ * The the new file is reopened as the new append only file. Profit!
+ */
+static int rewriteAppendOnlyFileBackground(void) {
+ pid_t childpid;
+
+ if (server.bgrewritechildpid != -1) return REDIS_ERR;
+ if ((childpid = fork()) == 0) {
+ /* Child */
+ char tmpfile[256];
+ close(server.fd);
+
+ snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) getpid());
+ if (rewriteAppendOnlyFile(tmpfile) == REDIS_OK) {
+ exit(0);
+ } else {
+ exit(1);
+ }
+ } else {
+ /* Parent */
+ if (childpid == -1) {
+ redisLog(REDIS_WARNING,
+ "Can't rewrite append only file in background: fork: %s",
+ strerror(errno));
+ return REDIS_ERR;
+ }
+ redisLog(REDIS_NOTICE,
+ "Background append only file rewriting started by pid %d",childpid);
+ server.bgrewritechildpid = childpid;
+ /* We set appendseldb to -1 in order to force the next call to the
+ * feedAppendOnlyFile() to issue a SELECT command, so the differences
+ * accumulated by the parent into server.bgrewritebuf will start
+ * with a SELECT statement and it will be safe to merge. */
+ server.appendseldb = -1;
+ return REDIS_OK;
+ }
+ return REDIS_OK; /* unreached */
+}
+
+static void bgrewriteaofCommand(redisClient *c) {
+ if (server.bgrewritechildpid != -1) {
+ addReplySds(c,sdsnew("-ERR background append only file rewriting already in progress\r\n"));
+ return;
+ }
+ if (rewriteAppendOnlyFileBackground() == REDIS_OK) {
+ addReply(c,shared.ok);
+ } else {
+ addReply(c,shared.err);
+ }
+}
+
+static void aofRemoveTempFile(pid_t childpid) {
+ char tmpfile[256];
+
+ snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) childpid);
+ unlink(tmpfile);
+}
+
/* ================================= Debugging ============================== */
static void debugCommand(redisClient *c) {
if (!strcasecmp(c->argv[1]->ptr,"segfault")) {
*((char*)-1) = 'x';
+ } else if (!strcasecmp(c->argv[1]->ptr,"reload")) {
+ if (rdbSave(server.dbfilename) != REDIS_OK) {
+ addReply(c,shared.err);
+ return;
+ }
+ emptyDb();
+ if (rdbLoad(server.dbfilename) != REDIS_OK) {
+ addReply(c,shared.err);
+ return;
+ }
+ redisLog(REDIS_WARNING,"DB reloaded by DEBUG RELOAD");
+ addReply(c,shared.ok);
} else if (!strcasecmp(c->argv[1]->ptr,"object") && c->argc == 3) {
dictEntry *de = dictFind(c->db->dict,c->argv[2]);
robj *key, *val;
key, key->refcount, val, val->refcount, val->encoding));
} else {
addReplySds(c,sdsnew(
- "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT <key>]\r\n"));
+ "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT <key>|RELOAD]\r\n"));
}
}
redisLog(REDIS_NOTICE,"DB loaded from disk");
}
if (aeCreateFileEvent(server.el, server.fd, AE_READABLE,
- acceptHandler, NULL, NULL) == AE_ERR) oom("creating file event");
+ acceptHandler, NULL) == AE_ERR) oom("creating file event");
redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
aeMain(server.el);
aeDeleteEventLoop(server.el);
char **messages = NULL;
int i, trace_size = 0;
unsigned long offset=0;
- time_t uptime = time(NULL)-server.stat_starttime;
ucontext_t *uc = (ucontext_t*) secret;
+ sds infostring;
REDIS_NOTUSED(info);
redisLog(REDIS_WARNING,
"======= Ooops! Redis %s got signal: -%d- =======", REDIS_VERSION, sig);
- redisLog(REDIS_WARNING, "%s", sdscatprintf(sdsempty(),
- "redis_version:%s; "
- "uptime_in_seconds:%d; "
- "connected_clients:%d; "
- "connected_slaves:%d; "
- "used_memory:%zu; "
- "changes_since_last_save:%lld; "
- "bgsave_in_progress:%d; "
- "last_save_time:%d; "
- "total_connections_received:%lld; "
- "total_commands_processed:%lld; "
- "role:%s;"
- ,REDIS_VERSION,
- uptime,
- listLength(server.clients)-listLength(server.slaves),
- listLength(server.slaves),
- server.usedmemory,
- server.dirty,
- server.bgsaveinprogress,
- server.lastsave,
- server.stat_numconnections,
- server.stat_numcommands,
- server.masterhost == NULL ? "master" : "slave"
- ));
+ infostring = genRedisInfoString();
+ redisLog(REDIS_WARNING, "%s",infostring);
+ /* It's not safe to sdsfree() the returned string under memory
+ * corruption conditions. Let it leak as we are going to abort */
trace_size = backtrace(trace, 100);
/* overwrite sigaction with caller's address */
redisLog(REDIS_WARNING,"%d redis-server %p %s + %d", i, trace[i], fn, (unsigned int)offset);
}
}
- free(messages);
+ // free(messages); Don't call free() with possibly corrupted memory.
exit(0);
}
projects / redis.git / blobdiff