source reshaped a bit to play well with a bgsaving thread, still work to do, does...

[redis.git] / src / rdb.c

diff --git a/src/rdb.c b/src/rdb.c
index 3fa284e127be9cd78a1a6bc9c5a7dfaa492189df..62756d3047252c9ac04149ce4384c1a3f1f77985 100644 (file)
--- a/src/rdb.c
+++ b/src/rdb.c
@@ -7,39 +7,51 @@
 #include <sys/resource.h>
 #include <sys/wait.h>
 #include <arpa/inet.h>
 #include <sys/resource.h>
 #include <sys/wait.h>
 #include <arpa/inet.h>

+#include <sys/stat.h>
+
+/* Convenience wrapper around fwrite, that returns the number of bytes written
+ * to the file instead of the number of objects (see fwrite(3)) and -1 in the
+ * case of an error. It also supports a NULL *fp to skip writing altogether
+ * instead of writing to /dev/null. */
+static int rdbWriteRaw(FILE *fp, void *p, size_t len) {
+    if (fp != NULL && fwrite(p,len,1,fp) == 0) return -1;
+    return len;
+}

 
 int rdbSaveType(FILE *fp, unsigned char type) {
 
 int rdbSaveType(FILE *fp, unsigned char type) {

-    if (fwrite(&type,1,1,fp) == 0) return -1;
-    return 0;
+    return rdbWriteRaw(fp,&type,1);

 }
 
 int rdbSaveTime(FILE *fp, time_t t) {
     int32_t t32 = (int32_t) t;
 }
 
 int rdbSaveTime(FILE *fp, time_t t) {
     int32_t t32 = (int32_t) t;

-    if (fwrite(&t32,4,1,fp) == 0) return -1;
-    return 0;
+    return rdbWriteRaw(fp,&t32,4);

 }
 
 /* check rdbLoadLen() comments for more info */
 int rdbSaveLen(FILE *fp, uint32_t len) {
     unsigned char buf[2];
 }
 
 /* check rdbLoadLen() comments for more info */
 int rdbSaveLen(FILE *fp, uint32_t len) {
     unsigned char buf[2];

+    int nwritten;

 
     if (len < (1<<6)) {
         /* Save a 6 bit len */
         buf[0] = (len&0xFF)|(REDIS_RDB_6BITLEN<<6);
 
     if (len < (1<<6)) {
         /* Save a 6 bit len */
         buf[0] = (len&0xFF)|(REDIS_RDB_6BITLEN<<6);

-        if (fwrite(buf,1,1,fp) == 0) return -1;
+        if (rdbWriteRaw(fp,buf,1) == -1) return -1;
+        nwritten = 1;

     } else if (len < (1<<14)) {
         /* Save a 14 bit len */
         buf[0] = ((len>>8)&0xFF)|(REDIS_RDB_14BITLEN<<6);
         buf[1] = len&0xFF;
     } else if (len < (1<<14)) {
         /* Save a 14 bit len */
         buf[0] = ((len>>8)&0xFF)|(REDIS_RDB_14BITLEN<<6);
         buf[1] = len&0xFF;

-        if (fwrite(buf,2,1,fp) == 0) return -1;
+        if (rdbWriteRaw(fp,buf,2) == -1) return -1;
+        nwritten = 2;

     } else {
         /* Save a 32 bit len */
         buf[0] = (REDIS_RDB_32BITLEN<<6);
     } else {
         /* Save a 32 bit len */
         buf[0] = (REDIS_RDB_32BITLEN<<6);

-        if (fwrite(buf,1,1,fp) == 0) return -1;
+        if (rdbWriteRaw(fp,buf,1) == -1) return -1;

         len = htonl(len);
         len = htonl(len);

-        if (fwrite(&len,4,1,fp) == 0) return -1;
+        if (rdbWriteRaw(fp,&len,4) == -1) return -1;
+        nwritten = 1+4;

     }
     }

-    return 0;
+    return nwritten;

 }
 
 /* Encode 'value' as an integer if possible (if integer will fit the
 }
 
 /* Encode 'value' as an integer if possible (if integer will fit the


@@ -92,6 +104,7 @@ int rdbTryIntegerEncoding(char *s, size_t len, unsigned char *enc) {
 int rdbSaveLzfStringObject(FILE *fp, unsigned char *s, size_t len) {
     size_t comprlen, outlen;
     unsigned char byte;
 int rdbSaveLzfStringObject(FILE *fp, unsigned char *s, size_t len) {
     size_t comprlen, outlen;
     unsigned char byte;

+    int n, nwritten = 0;

     void *out;
 
     /* We require at least four bytes compression for this to be worth it */
     void *out;
 
     /* We require at least four bytes compression for this to be worth it */


@@ -105,12 +118,20 @@ int rdbSaveLzfStringObject(FILE *fp, unsigned char *s, size_t len) {
     }
     /* Data compressed! Let's save it on disk */
     byte = (REDIS_RDB_ENCVAL<<6)|REDIS_RDB_ENC_LZF;
     }
     /* Data compressed! Let's save it on disk */
     byte = (REDIS_RDB_ENCVAL<<6)|REDIS_RDB_ENC_LZF;

-    if (fwrite(&byte,1,1,fp) == 0) goto writeerr;
-    if (rdbSaveLen(fp,comprlen) == -1) goto writeerr;
-    if (rdbSaveLen(fp,len) == -1) goto writeerr;
-    if (fwrite(out,comprlen,1,fp) == 0) goto writeerr;
+    if ((n = rdbWriteRaw(fp,&byte,1)) == -1) goto writeerr;
+    nwritten += n;
+
+    if ((n = rdbSaveLen(fp,comprlen)) == -1) goto writeerr;
+    nwritten += n;
+
+    if ((n = rdbSaveLen(fp,len)) == -1) goto writeerr;
+    nwritten += n;
+
+    if ((n = rdbWriteRaw(fp,out,comprlen)) == -1) goto writeerr;
+    nwritten += n;
+

     zfree(out);
     zfree(out);

-    return comprlen;
+    return nwritten;

 
 writeerr:
     zfree(out);
 
 writeerr:
     zfree(out);


@@ -121,47 +142,53 @@ writeerr:
  * representation of an integer value we try to safe it in a special form */
 int rdbSaveRawString(FILE *fp, unsigned char *s, size_t len) {
     int enclen;
  * representation of an integer value we try to safe it in a special form */
 int rdbSaveRawString(FILE *fp, unsigned char *s, size_t len) {
     int enclen;

+    int n, nwritten = 0;

 
     /* Try integer encoding */
     if (len <= 11) {
         unsigned char buf[5];
         if ((enclen = rdbTryIntegerEncoding((char*)s,len,buf)) > 0) {
 
     /* Try integer encoding */
     if (len <= 11) {
         unsigned char buf[5];
         if ((enclen = rdbTryIntegerEncoding((char*)s,len,buf)) > 0) {

-            if (fwrite(buf,enclen,1,fp) == 0) return -1;
-            return 0;
+            if (rdbWriteRaw(fp,buf,enclen) == -1) return -1;
+            return enclen;

         }
     }
 
     /* Try LZF compression - under 20 bytes it's unable to compress even
      * aaaaaaaaaaaaaaaaaa so skip it */
     if (server.rdbcompression && len > 20) {
         }
     }
 
     /* Try LZF compression - under 20 bytes it's unable to compress even
      * aaaaaaaaaaaaaaaaaa so skip it */
     if (server.rdbcompression && len > 20) {

-        int retval;
-
-        retval = rdbSaveLzfStringObject(fp,s,len);
-        if (retval == -1) return -1;
-        if (retval > 0) return 0;
-        /* retval == 0 means data can't be compressed, save the old way */
+        n = rdbSaveLzfStringObject(fp,s,len);
+        if (n == -1) return -1;
+        if (n > 0) return n;
+        /* Return value of 0 means data can't be compressed, save the old way */

     }
 
     /* Store verbatim */
     }
 
     /* Store verbatim */

-    if (rdbSaveLen(fp,len) == -1) return -1;
-    if (len && fwrite(s,len,1,fp) == 0) return -1;
-    return 0;
+    if ((n = rdbSaveLen(fp,len)) == -1) return -1;
+    nwritten += n;
+    if (len > 0) {
+        if (rdbWriteRaw(fp,s,len) == -1) return -1;
+        nwritten += len;
+    }
+    return nwritten;

 }
 
 /* Save a long long value as either an encoded string or a string. */
 int rdbSaveLongLongAsStringObject(FILE *fp, long long value) {
     unsigned char buf[32];
 }
 
 /* Save a long long value as either an encoded string or a string. */
 int rdbSaveLongLongAsStringObject(FILE *fp, long long value) {
     unsigned char buf[32];

+    int n, nwritten = 0;

     int enclen = rdbEncodeInteger(value,buf);
     if (enclen > 0) {
     int enclen = rdbEncodeInteger(value,buf);
     if (enclen > 0) {

-        if (fwrite(buf,enclen,1,fp) == 0) return -1;
+        return rdbWriteRaw(fp,buf,enclen);

     } else {
         /* Encode as string */
         enclen = ll2string((char*)buf,32,value);
         redisAssert(enclen < 32);
     } else {
         /* Encode as string */
         enclen = ll2string((char*)buf,32,value);
         redisAssert(enclen < 32);

-        if (rdbSaveLen(fp,enclen) == -1) return -1;
-        if (fwrite(buf,enclen,1,fp) == 0) return -1;
+        if ((n = rdbSaveLen(fp,enclen)) == -1) return -1;
+        nwritten += n;
+        if ((n = rdbWriteRaw(fp,buf,enclen)) == -1) return -1;
+        nwritten += n;

     }
     }

-    return 0;
+    return nwritten;

 }
 
 /* Like rdbSaveStringObjectRaw() but handle encoded objects */
 }
 
 /* Like rdbSaveStringObjectRaw() but handle encoded objects */


@@ -215,15 +242,17 @@ int rdbSaveDoubleValue(FILE *fp, double val) {
         buf[0] = strlen((char*)buf+1);
         len = buf[0]+1;
     }
         buf[0] = strlen((char*)buf+1);
         len = buf[0]+1;
     }

-    if (fwrite(buf,len,1,fp) == 0) return -1;
-    return 0;
+    return rdbWriteRaw(fp,buf,len);

 }
 
 /* Save a Redis object. */
 int rdbSaveObject(FILE *fp, robj *o) {
 }
 
 /* Save a Redis object. */
 int rdbSaveObject(FILE *fp, robj *o) {

+    int n, nwritten = 0;
+

     if (o->type == REDIS_STRING) {
         /* Save a string value */
     if (o->type == REDIS_STRING) {
         /* Save a string value */

-        if (rdbSaveStringObject(fp,o) == -1) return -1;
+        if ((n = rdbSaveStringObject(fp,o)) == -1) return -1;
+        nwritten += n;

     } else if (o->type == REDIS_LIST) {
         /* Save a list value */
         if (o->encoding == REDIS_ENCODING_ZIPLIST) {
     } else if (o->type == REDIS_LIST) {
         /* Save a list value */
         if (o->encoding == REDIS_ENCODING_ZIPLIST) {


@@ -232,15 +261,19 @@ int rdbSaveObject(FILE *fp, robj *o) {
             unsigned int vlen;
             long long vlong;
 
             unsigned int vlen;
             long long vlong;
 

-            if (rdbSaveLen(fp,ziplistLen(o->ptr)) == -1) return -1;
+            if ((n = rdbSaveLen(fp,ziplistLen(o->ptr))) == -1) return -1;
+            nwritten += n;
+

             p = ziplistIndex(o->ptr,0);
             while(ziplistGet(p,&vstr,&vlen,&vlong)) {
                 if (vstr) {
             p = ziplistIndex(o->ptr,0);
             while(ziplistGet(p,&vstr,&vlen,&vlong)) {
                 if (vstr) {

-                    if (rdbSaveRawString(fp,vstr,vlen) == -1)
+                    if ((n = rdbSaveRawString(fp,vstr,vlen)) == -1)

                         return -1;
                         return -1;

+                    nwritten += n;

                 } else {
                 } else {

-                    if (rdbSaveLongLongAsStringObject(fp,vlong) == -1)
+                    if ((n = rdbSaveLongLongAsStringObject(fp,vlong)) == -1)

                         return -1;
                         return -1;

+                    nwritten += n;

                 }
                 p = ziplistNext(o->ptr,p);
             }
                 }
                 p = ziplistNext(o->ptr,p);
             }


@@ -249,41 +282,66 @@ int rdbSaveObject(FILE *fp, robj *o) {
             listIter li;
             listNode *ln;
 
             listIter li;
             listNode *ln;
 

-            if (rdbSaveLen(fp,listLength(list)) == -1) return -1;
+            if ((n = rdbSaveLen(fp,listLength(list))) == -1) return -1;
+            nwritten += n;
+

             listRewind(list,&li);
             while((ln = listNext(&li))) {
                 robj *eleobj = listNodeValue(ln);
             listRewind(list,&li);
             while((ln = listNext(&li))) {
                 robj *eleobj = listNodeValue(ln);

-                if (rdbSaveStringObject(fp,eleobj) == -1) return -1;
+                if ((n = rdbSaveStringObject(fp,eleobj)) == -1) return -1;
+                nwritten += n;

             }
         } else {
             redisPanic("Unknown list encoding");
         }
     } else if (o->type == REDIS_SET) {
         /* Save a set value */
             }
         } else {
             redisPanic("Unknown list encoding");
         }
     } else if (o->type == REDIS_SET) {
         /* Save a set value */

-        dict *set = o->ptr;
-        dictIterator *di = dictGetIterator(set);
-        dictEntry *de;
+        if (o->encoding == REDIS_ENCODING_HT) {
+            dict *set = o->ptr;
+            dictIterator *di = dictGetIterator(set);
+            dictEntry *de;

 
 

-        if (rdbSaveLen(fp,dictSize(set)) == -1) return -1;
-        while((de = dictNext(di)) != NULL) {
-            robj *eleobj = dictGetEntryKey(de);
+            if ((n = rdbSaveLen(fp,dictSize(set))) == -1) return -1;
+            nwritten += n;

 
 

-            if (rdbSaveStringObject(fp,eleobj) == -1) return -1;
+            while((de = dictNext(di)) != NULL) {
+                robj *eleobj = dictGetEntryKey(de);
+                if ((n = rdbSaveStringObject(fp,eleobj)) == -1) return -1;
+                nwritten += n;
+            }
+            dictReleaseIterator(di);
+        } else if (o->encoding == REDIS_ENCODING_INTSET) {
+            intset *is = o->ptr;
+            int64_t llval;
+            int i = 0;
+
+            if ((n = rdbSaveLen(fp,intsetLen(is))) == -1) return -1;
+            nwritten += n;
+
+            while(intsetGet(is,i++,&llval)) {
+                if ((n = rdbSaveLongLongAsStringObject(fp,llval)) == -1) return -1;
+                nwritten += n;
+            }
+        } else {
+            redisPanic("Unknown set encoding");

         }
         }

-        dictReleaseIterator(di);

     } else if (o->type == REDIS_ZSET) {
         /* Save a set value */
         zset *zs = o->ptr;
         dictIterator *di = dictGetIterator(zs->dict);
         dictEntry *de;
 
     } else if (o->type == REDIS_ZSET) {
         /* Save a set value */
         zset *zs = o->ptr;
         dictIterator *di = dictGetIterator(zs->dict);
         dictEntry *de;
 

-        if (rdbSaveLen(fp,dictSize(zs->dict)) == -1) return -1;
+        if ((n = rdbSaveLen(fp,dictSize(zs->dict))) == -1) return -1;
+        nwritten += n;
+

         while((de = dictNext(di)) != NULL) {
             robj *eleobj = dictGetEntryKey(de);
             double *score = dictGetEntryVal(de);
 
         while((de = dictNext(di)) != NULL) {
             robj *eleobj = dictGetEntryKey(de);
             double *score = dictGetEntryVal(de);
 

-            if (rdbSaveStringObject(fp,eleobj) == -1) return -1;
-            if (rdbSaveDoubleValue(fp,*score) == -1) return -1;
+            if ((n = rdbSaveStringObject(fp,eleobj)) == -1) return -1;
+            nwritten += n;
+            if ((n = rdbSaveDoubleValue(fp,*score)) == -1) return -1;
+            nwritten += n;

         }
         dictReleaseIterator(di);
     } else if (o->type == REDIS_HASH) {
         }
         dictReleaseIterator(di);
     } else if (o->type == REDIS_HASH) {


@@ -294,47 +352,72 @@ int rdbSaveObject(FILE *fp, robj *o) {
             unsigned char *key, *val;
             unsigned int klen, vlen;
 
             unsigned char *key, *val;
             unsigned int klen, vlen;
 

-            if (rdbSaveLen(fp,count) == -1) return -1;
+            if ((n = rdbSaveLen(fp,count)) == -1) return -1;
+            nwritten += n;
+

             while((p = zipmapNext(p,&key,&klen,&val,&vlen)) != NULL) {
             while((p = zipmapNext(p,&key,&klen,&val,&vlen)) != NULL) {

-                if (rdbSaveRawString(fp,key,klen) == -1) return -1;
-                if (rdbSaveRawString(fp,val,vlen) == -1) return -1;
+                if ((n = rdbSaveRawString(fp,key,klen)) == -1) return -1;
+                nwritten += n;
+                if ((n = rdbSaveRawString(fp,val,vlen)) == -1) return -1;
+                nwritten += n;

             }
         } else {
             dictIterator *di = dictGetIterator(o->ptr);
             dictEntry *de;
 
             }
         } else {
             dictIterator *di = dictGetIterator(o->ptr);
             dictEntry *de;
 

-            if (rdbSaveLen(fp,dictSize((dict*)o->ptr)) == -1) return -1;
+            if ((n = rdbSaveLen(fp,dictSize((dict*)o->ptr))) == -1) return -1;
+            nwritten += n;
+

             while((de = dictNext(di)) != NULL) {
                 robj *key = dictGetEntryKey(de);
                 robj *val = dictGetEntryVal(de);
 
             while((de = dictNext(di)) != NULL) {
                 robj *key = dictGetEntryKey(de);
                 robj *val = dictGetEntryVal(de);
 

-                if (rdbSaveStringObject(fp,key) == -1) return -1;
-                if (rdbSaveStringObject(fp,val) == -1) return -1;
+                if ((n = rdbSaveStringObject(fp,key)) == -1) return -1;
+                nwritten += n;
+                if ((n = rdbSaveStringObject(fp,val)) == -1) return -1;
+                nwritten += n;

             }
             dictReleaseIterator(di);
         }
     } else {
         redisPanic("Unknown object type");
     }
             }
             dictReleaseIterator(di);
         }
     } else {
         redisPanic("Unknown object type");
     }

-    return 0;
+    return nwritten;

 }
 
 /* Return the length the object will have on disk if saved with
  * the rdbSaveObject() function. Currently we use a trick to get
  * this length with very little changes to the code. In the future
  * we could switch to a faster solution. */
 }
 
 /* Return the length the object will have on disk if saved with
  * the rdbSaveObject() function. Currently we use a trick to get
  * this length with very little changes to the code. In the future
  * we could switch to a faster solution. */

-off_t rdbSavedObjectLen(robj *o, FILE *fp) {
-    if (fp == NULL) fp = server.devnull;
-    rewind(fp);
-    redisAssert(rdbSaveObject(fp,o) != 1);
-    return ftello(fp);
+off_t rdbSavedObjectLen(robj *o) {
+    int len = rdbSaveObject(NULL,o);
+    redisAssert(len != -1);
+    return len;

 }
 
 }
 

-/* Return the number of pages required to save this object in the swap file */
-off_t rdbSavedObjectPages(robj *o, FILE *fp) {
-    off_t bytes = rdbSavedObjectLen(o,fp);
-
-    return (bytes+(server.vm_page_size-1))/server.vm_page_size;
+/* Save a key-value pair, with expire time, type, key, value.
+ * On error -1 is returned.
+ * On success if the key was actaully saved 1 is returned, otherwise 0
+ * is returned (the key was already expired). */
+int rdbSaveKeyValuePair(FILE *fp, redisDb *db, robj *key, robj *val,
+                        time_t now)
+{
+    time_t expiretime;
+    
+    expiretime = getExpire(db,key);
+
+    /* Save the expire time */
+    if (expiretime != -1) {
+        /* If this key is already expired skip it */
+        if (expiretime < now) return 0;
+        if (rdbSaveType(fp,REDIS_EXPIRETIME) == -1) return -1;
+        if (rdbSaveTime(fp,expiretime) == -1) return -1;
+    }
+    /* Save type, key, value */
+    if (rdbSaveType(fp,val->type) == -1) return -1;
+    if (rdbSaveStringObject(fp,key) == -1) return -1;
+    if (rdbSaveObject(fp,val) == -1) return -1;
+    return 1;

 }
 
 /* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success */
 }
 
 /* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success */


@@ -346,11 +429,10 @@ int rdbSave(char *filename) {
     int j;
     time_t now = time(NULL);
 
     int j;
     time_t now = time(NULL);
 

-    /* Wait for I/O therads to terminate, just in case this is a
-     * foreground-saving, to avoid seeking the swap file descriptor at the
-     * same time. */
-    if (server.vm_enabled)
-        waitEmptyIOJobsQueue();
+    if (server.ds_enabled) {
+        cacheForcePointInTime();
+        return dsRdbSave(filename);
+    }

 
     snprintf(tmpfile,256,"temp-%d.rdb", (int) getpid());
     fp = fopen(tmpfile,"w");
 
     snprintf(tmpfile,256,"temp-%d.rdb", (int) getpid());
     fp = fopen(tmpfile,"w");


@@ -377,38 +459,9 @@ int rdbSave(char *filename) {
         while((de = dictNext(di)) != NULL) {
             sds keystr = dictGetEntryKey(de);
             robj key, *o = dictGetEntryVal(de);
         while((de = dictNext(di)) != NULL) {
             sds keystr = dictGetEntryKey(de);
             robj key, *o = dictGetEntryVal(de);

-            time_t expiretime;

             
             initStaticStringObject(key,keystr);
             
             initStaticStringObject(key,keystr);

-            expiretime = getExpire(db,&key);
-
-            /* Save the expire time */
-            if (expiretime != -1) {
-                /* If this key is already expired skip it */
-                if (expiretime < now) continue;
-                if (rdbSaveType(fp,REDIS_EXPIRETIME) == -1) goto werr;
-                if (rdbSaveTime(fp,expiretime) == -1) goto werr;
-            }
-            /* Save the key and associated value. This requires special
-             * handling if the value is swapped out. */
-            if (!server.vm_enabled || o->storage == REDIS_VM_MEMORY ||
-                                      o->storage == REDIS_VM_SWAPPING) {
-                /* Save type, key, value */
-                if (rdbSaveType(fp,o->type) == -1) goto werr;
-                if (rdbSaveStringObject(fp,&key) == -1) goto werr;
-                if (rdbSaveObject(fp,o) == -1) goto werr;
-            } else {
-                /* REDIS_VM_SWAPPED or REDIS_VM_LOADING */
-                robj *po;
-                /* Get a preview of the object in memory */
-                po = vmPreviewObject(o);
-                /* Save type, key, value */
-                if (rdbSaveType(fp,po->type) == -1) goto werr;
-                if (rdbSaveStringObject(fp,&key) == -1) goto werr;
-                if (rdbSaveObject(fp,po) == -1) goto werr;
-                /* Remove the loaded object from memory */
-                decrRefCount(po);
-            }
+            if (rdbSaveKeyValuePair(fp,db,&key,o,now) == -1) goto werr;

         }
         dictReleaseIterator(di);
     }
         }
         dictReleaseIterator(di);
     }


@@ -443,18 +496,24 @@ werr:
 int rdbSaveBackground(char *filename) {
     pid_t childpid;
 
 int rdbSaveBackground(char *filename) {
     pid_t childpid;
 

-    if (server.bgsavechildpid != -1) return REDIS_ERR;
-    if (server.vm_enabled) waitEmptyIOJobsQueue();
+    if (server.bgsavechildpid != -1 ||
+        server.bgsavethread != (pthread_t) -1) return REDIS_ERR;
+
+    server.dirty_before_bgsave = server.dirty;
+
+    if (server.ds_enabled) {
+        cacheForcePointInTime();
+        return dsRdbSave(filename);
+    }
+

     if ((childpid = fork()) == 0) {
     if ((childpid = fork()) == 0) {

+        int retval;
+

         /* Child */
         /* Child */

-        if (server.vm_enabled) vmReopenSwapFile();

         if (server.ipfd > 0) close(server.ipfd);
         if (server.sofd > 0) close(server.sofd);
         if (server.ipfd > 0) close(server.ipfd);
         if (server.sofd > 0) close(server.sofd);

-        if (rdbSave(filename) == REDIS_OK) {
-            _exit(0);
-        } else {
-            _exit(1);
-        }
+        retval = rdbSave(filename);
+        _exit((retval == REDIS_OK) ? 0 : 1);

     } else {
         /* Parent */
         if (childpid == -1) {
     } else {
         /* Parent */
         if (childpid == -1) {


@@ -629,6 +688,7 @@ int rdbLoadDoubleValue(FILE *fp, double *val) {
 robj *rdbLoadObject(int type, FILE *fp) {
     robj *o, *ele, *dec;
     size_t len;
 robj *rdbLoadObject(int type, FILE *fp) {
     robj *o, *ele, *dec;
     size_t len;

+    unsigned int i;

 
     redisLog(REDIS_DEBUG,"LOADING OBJECT %d (at %d)\n",type,ftell(fp));
     if (type == REDIS_STRING) {
 
     redisLog(REDIS_DEBUG,"LOADING OBJECT %d (at %d)\n",type,ftell(fp));
     if (type == REDIS_STRING) {


@@ -670,16 +730,41 @@ robj *rdbLoadObject(int type, FILE *fp) {
     } else if (type == REDIS_SET) {
         /* Read list/set value */
         if ((len = rdbLoadLen(fp,NULL)) == REDIS_RDB_LENERR) return NULL;
     } else if (type == REDIS_SET) {
         /* Read list/set value */
         if ((len = rdbLoadLen(fp,NULL)) == REDIS_RDB_LENERR) return NULL;

-        o = createSetObject();
-        /* It's faster to expand the dict to the right size asap in order
-         * to avoid rehashing */
-        if (len > DICT_HT_INITIAL_SIZE)
-            dictExpand(o->ptr,len);
+
+        /* Use a regular set when there are too many entries. */
+        if (len > server.set_max_intset_entries) {
+            o = createSetObject();
+            /* It's faster to expand the dict to the right size asap in order
+             * to avoid rehashing */
+            if (len > DICT_HT_INITIAL_SIZE)
+                dictExpand(o->ptr,len);
+        } else {
+            o = createIntsetObject();
+        }
+

         /* Load every single element of the list/set */
         /* Load every single element of the list/set */

-        while(len--) {
+        for (i = 0; i < len; i++) {
+            long long llval;

             if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
             ele = tryObjectEncoding(ele);
             if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
             ele = tryObjectEncoding(ele);

-            dictAdd((dict*)o->ptr,ele,NULL);
+
+            if (o->encoding == REDIS_ENCODING_INTSET) {
+                /* Fetch integer value from element */
+                if (isObjectRepresentableAsLongLong(ele,&llval) == REDIS_OK) {
+                    o->ptr = intsetAdd(o->ptr,llval,NULL);
+                } else {
+                    setTypeConvert(o,REDIS_ENCODING_HT);
+                    dictExpand(o->ptr,len);
+                }
+            }
+
+            /* This will also be called when the set was just converted
+             * to regular hashtable encoded set */
+            if (o->encoding == REDIS_ENCODING_HT) {
+                dictAdd((dict*)o->ptr,ele,NULL);
+            } else {
+                decrRefCount(ele);
+            }

         }
     } else if (type == REDIS_ZSET) {
         /* Read list/set value */
         }
     } else if (type == REDIS_ZSET) {
         /* Read list/set value */


@@ -692,13 +777,14 @@ robj *rdbLoadObject(int type, FILE *fp) {
         /* Load every single element of the list/set */
         while(zsetlen--) {
             robj *ele;
         /* Load every single element of the list/set */
         while(zsetlen--) {
             robj *ele;

-            double *score = zmalloc(sizeof(double));
+            double score;
+            zskiplistNode *znode;

 
             if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
             ele = tryObjectEncoding(ele);
 
             if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
             ele = tryObjectEncoding(ele);

-            if (rdbLoadDoubleValue(fp,score) == -1) return NULL;
-            dictAdd(zs->dict,ele,score);
-            zslInsert(zs->zsl,*score,ele);
+            if (rdbLoadDoubleValue(fp,&score) == -1) return NULL;
+            znode = zslInsert(zs->zsl,score,ele);
+            dictAdd(zs->dict,ele,&znode->score);

             incrRefCount(ele); /* added to skiplist */
         }
     } else if (type == REDIS_HASH) {
             incrRefCount(ele); /* added to skiplist */
         }
     } else if (type == REDIS_HASH) {


@@ -753,14 +839,39 @@ robj *rdbLoadObject(int type, FILE *fp) {
     return o;
 }
 
     return o;
 }
 

+/* Mark that we are loading in the global state and setup the fields
+ * needed to provide loading stats. */
+void startLoading(FILE *fp) {
+    struct stat sb;
+
+    /* Load the DB */
+    server.loading = 1;
+    server.loading_start_time = time(NULL);
+    if (fstat(fileno(fp), &sb) == -1) {
+        server.loading_total_bytes = 1; /* just to avoid division by zero */
+    } else {
+        server.loading_total_bytes = sb.st_size;
+    }
+}
+
+/* Refresh the loading progress info */
+void loadingProgress(off_t pos) {
+    server.loading_loaded_bytes = pos;
+}
+
+/* Loading finished */
+void stopLoading(void) {
+    server.loading = 0;
+}
+

 int rdbLoad(char *filename) {
     FILE *fp;
     uint32_t dbid;
     int type, retval, rdbver;
 int rdbLoad(char *filename) {
     FILE *fp;
     uint32_t dbid;
     int type, retval, rdbver;

-    int swap_all_values = 0;

     redisDb *db = server.db+0;
     char buf[1024];
     time_t expiretime, now = time(NULL);
     redisDb *db = server.db+0;
     char buf[1024];
     time_t expiretime, now = time(NULL);

+    long loops = 0;

 
     fp = fopen(filename,"r");
     if (!fp) return REDIS_ERR;
 
     fp = fopen(filename,"r");
     if (!fp) return REDIS_ERR;


@@ -777,11 +888,18 @@ int rdbLoad(char *filename) {
         redisLog(REDIS_WARNING,"Can't handle RDB format version %d",rdbver);
         return REDIS_ERR;
     }
         redisLog(REDIS_WARNING,"Can't handle RDB format version %d",rdbver);
         return REDIS_ERR;
     }

+
+    startLoading(fp);

     while(1) {
         robj *key, *val;
     while(1) {
         robj *key, *val;

-        int force_swapout;
-

         expiretime = -1;
         expiretime = -1;

+
+        /* Serve the clients from time to time */
+        if (!(loops++ % 1000)) {
+            loadingProgress(ftello(fp));
+            aeProcessEvents(server.el, AE_FILE_EVENTS|AE_DONT_WAIT);
+        }
+

         /* Read type. */
         if ((type = rdbLoadType(fp)) == -1) goto eoferr;
         if (type == REDIS_EXPIRETIME) {
         /* Read type. */
         if ((type = rdbLoadType(fp)) == -1) goto eoferr;
         if (type == REDIS_EXPIRETIME) {


@@ -820,46 +938,10 @@ int rdbLoad(char *filename) {
         /* Set the expire time if needed */
         if (expiretime != -1) setExpire(db,key,expiretime);
 
         /* Set the expire time if needed */
         if (expiretime != -1) setExpire(db,key,expiretime);
 

-        /* Handle swapping while loading big datasets when VM is on */
-
-        /* If we detecter we are hopeless about fitting something in memory
-         * we just swap every new key on disk. Directly...
-         * Note that's important to check for this condition before resorting
-         * to random sampling, otherwise we may try to swap already
-         * swapped keys. */
-        if (swap_all_values) {
-            dictEntry *de = dictFind(db->dict,key->ptr);
-
-            /* de may be NULL since the key already expired */
-            if (de) {
-                vmpointer *vp;
-                val = dictGetEntryVal(de);
-
-                if (val->refcount == 1 &&
-                    (vp = vmSwapObjectBlocking(val)) != NULL)
-                    dictGetEntryVal(de) = vp;
-            }
-            decrRefCount(key);
-            continue;
-        }

         decrRefCount(key);
         decrRefCount(key);

-
-        /* Flush data on disk once 32 MB of additional RAM are used... */
-        force_swapout = 0;
-        if ((zmalloc_used_memory() - server.vm_max_memory) > 1024*1024*32)
-            force_swapout = 1;
-
-        /* If we have still some hope of having some value fitting memory
-         * then we try random sampling. */
-        if (!swap_all_values && server.vm_enabled && force_swapout) {
-            while (zmalloc_used_memory() > server.vm_max_memory) {
-                if (vmSwapOneObjectBlocking() == REDIS_ERR) break;
-            }
-            if (zmalloc_used_memory() > server.vm_max_memory)
-                swap_all_values = 1; /* We are already using too much mem */
-        }

     }
     fclose(fp);
     }
     fclose(fp);

+    stopLoading();

     return REDIS_OK;
 
 eoferr: /* unexpected end of file is handled here with a fatal exit */
     return REDIS_OK;
 
 eoferr: /* unexpected end of file is handled here with a fatal exit */


@@ -869,24 +951,47 @@ eoferr: /* unexpected end of file is handled here with a fatal exit */
 }
 
 /* A background saving child (BGSAVE) terminated its work. Handle this. */
 }
 
 /* A background saving child (BGSAVE) terminated its work. Handle this. */

-void backgroundSaveDoneHandler(int statloc) {
-    int exitcode = WEXITSTATUS(statloc);
-    int bysignal = WIFSIGNALED(statloc);
-
+void backgroundSaveDoneHandler(int exitcode, int bysignal) {

     if (!bysignal && exitcode == 0) {
         redisLog(REDIS_NOTICE,
             "Background saving terminated with success");
     if (!bysignal && exitcode == 0) {
         redisLog(REDIS_NOTICE,
             "Background saving terminated with success");

-        server.dirty = 0;
+        server.dirty = server.dirty - server.dirty_before_bgsave;

         server.lastsave = time(NULL);
     } else if (!bysignal && exitcode != 0) {
         redisLog(REDIS_WARNING, "Background saving error");
     } else {
         redisLog(REDIS_WARNING,
         server.lastsave = time(NULL);
     } else if (!bysignal && exitcode != 0) {
         redisLog(REDIS_WARNING, "Background saving error");
     } else {
         redisLog(REDIS_WARNING,

-            "Background saving terminated by signal %d", WTERMSIG(statloc));
+            "Background saving terminated by signal %d", bysignal);

         rdbRemoveTempFile(server.bgsavechildpid);
     }
     server.bgsavechildpid = -1;
         rdbRemoveTempFile(server.bgsavechildpid);
     }
     server.bgsavechildpid = -1;

+    server.bgsavethread = (pthread_t) -1;
+    server.bgsavethread_state = REDIS_BGSAVE_THREAD_UNACTIVE;

     /* 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);
 }
     /* 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);
 }

+
+void saveCommand(redisClient *c) {
+    if (server.bgsavechildpid != -1 || server.bgsavethread != (pthread-t)-1) {
+        addReplyError(c,"Background save already in progress");
+        return;
+    }
+    if (rdbSave(server.dbfilename) == REDIS_OK) {
+        addReply(c,shared.ok);
+    } else {
+        addReply(c,shared.err);
+    }
+}
+
+void bgsaveCommand(redisClient *c) {
+    if (server.bgsavechildpid != -1 || server.bgsavethread != (pthread-t)-1) {
+        addReplyError(c,"Background save already in progress");
+        return;
+    }
+    if (rdbSaveBackground(server.dbfilename) == REDIS_OK) {
+        addReplyStatus(c,"Background saving started");
+    } else {
+        addReply(c,shared.err);
+    }
+}