A reimplementation of blocking operation internals.

[redis.git] / src / rdb.c

diff --git a/src/rdb.c b/src/rdb.c
index 519b645d157281e86b6e2213ddfdfb648e36069a..9190ea514cb65ed12af35795003fabade5fd6102 100644 (file)
--- a/src/rdb.c
+++ b/src/rdb.c
@@ -1,6 +1,7 @@
 #include "redis.h"
 #include "lzf.h"    /* LZF compression library */
 #include "zipmap.h"
 #include "redis.h"
 #include "lzf.h"    /* LZF compression library */
 #include "zipmap.h"

+#include "endianconv.h"

 
 #include <math.h>
 #include <sys/types.h>
 
 #include <math.h>
 #include <sys/types.h>


@@ -20,17 +21,15 @@ int rdbSaveType(rio *rdb, unsigned char type) {
     return rdbWriteRaw(rdb,&type,1);
 }
 
     return rdbWriteRaw(rdb,&type,1);
 }
 

+/* Load a "type" in RDB format, that is a one byte unsigned integer.
+ * This function is not only used to load object types, but also special
+ * "types" like the end-of-file type, the EXPIRE type, and so forth. */

 int rdbLoadType(rio *rdb) {
     unsigned char type;
     if (rioRead(rdb,&type,1) == 0) return -1;
     return type;
 }
 
 int rdbLoadType(rio *rdb) {
     unsigned char type;
     if (rioRead(rdb,&type,1) == 0) return -1;
     return type;
 }
 

-int rdbSaveTime(rio *rdb, time_t t) {
-    int32_t t32 = (int32_t) t;
-    return rdbWriteRaw(rdb,&t32,4);
-}
-

 time_t rdbLoadTime(rio *rdb) {
     int32_t t32;
     if (rioRead(rdb,&t32,4) == 0) return -1;
 time_t rdbLoadTime(rio *rdb) {
     int32_t t32;
     if (rioRead(rdb,&t32,4) == 0) return -1;


@@ -437,7 +436,8 @@ int rdbSaveObjectType(rio *rdb, robj *o) {
     return -1; /* avoid warning */
 }
 
     return -1; /* avoid warning */
 }
 

-/* Load object type. Return -1 when the byte doesn't contain an object type. */
+/* Use rdbLoadType() to load a TYPE in RDB format, but returns -1 if the
+ * type is not specifically a valid Object Type. */

 int rdbLoadObjectType(rio *rdb) {
     int type;
     if ((type = rdbLoadType(rdb)) == -1) return -1;
 int rdbLoadObjectType(rio *rdb) {
     int type;
     if ((type = rdbLoadType(rdb)) == -1) return -1;


@@ -602,10 +602,12 @@ int rdbSave(char *filename) {
     dictIterator *di = NULL;
     dictEntry *de;
     char tmpfile[256];
     dictIterator *di = NULL;
     dictEntry *de;
     char tmpfile[256];

+    char magic[10];

     int j;
     long long now = mstime();
     FILE *fp;
     rio rdb;
     int j;
     long long now = mstime();
     FILE *fp;
     rio rdb;

+    uint64_t cksum;

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


@@ -616,7 +618,10 @@ int rdbSave(char *filename) {
     }
 
     rioInitWithFile(&rdb,fp);
     }
 
     rioInitWithFile(&rdb,fp);

-    if (rdbWriteRaw(&rdb,"REDIS0004",9) == -1) goto werr;
+    if (server.rdb_checksum)
+        rdb.update_cksum = rioGenericUpdateChecksum;
+    snprintf(magic,sizeof(magic),"REDIS%04d",REDIS_RDB_VERSION);
+    if (rdbWriteRaw(&rdb,magic,9) == -1) goto werr;

 
     for (j = 0; j < server.dbnum; j++) {
         redisDb *db = server.db+j;
 
     for (j = 0; j < server.dbnum; j++) {
         redisDb *db = server.db+j;


@@ -644,9 +649,17 @@ int rdbSave(char *filename) {
         }
         dictReleaseIterator(di);
     }
         }
         dictReleaseIterator(di);
     }

+    di = NULL; /* So that we don't release it again on error. */
+

     /* EOF opcode */
     if (rdbSaveType(&rdb,REDIS_RDB_OPCODE_EOF) == -1) goto werr;
 
     /* EOF opcode */
     if (rdbSaveType(&rdb,REDIS_RDB_OPCODE_EOF) == -1) goto werr;
 

+    /* CRC64 checksum. It will be zero if checksum computation is disabled, the
+     * loading code skips the check in this case. */
+    cksum = rdb.cksum;
+    memrev64ifbe(&cksum);
+    rioWrite(&rdb,&cksum,8);
+

     /* Make sure data will not remain on the OS's output buffers */
     fflush(fp);
     fsync(fileno(fp));
     /* Make sure data will not remain on the OS's output buffers */
     fflush(fp);
     fsync(fileno(fp));


@@ -689,7 +702,7 @@ int rdbSaveBackground(char *filename) {
         if (server.ipfd > 0) close(server.ipfd);
         if (server.sofd > 0) close(server.sofd);
         retval = rdbSave(filename);
         if (server.ipfd > 0) close(server.ipfd);
         if (server.sofd > 0) close(server.sofd);
         retval = rdbSave(filename);

-        _exit((retval == REDIS_OK) ? 0 : 1);
+        exitFromChild((retval == REDIS_OK) ? 0 : 1);

     } else {
         /* Parent */
         server.stat_fork_time = ustime()-start;
     } else {
         /* Parent */
         server.stat_fork_time = ustime()-start;


@@ -699,6 +712,7 @@ int rdbSaveBackground(char *filename) {
             return REDIS_ERR;
         }
         redisLog(REDIS_NOTICE,"Background saving started by pid %d",childpid);
             return REDIS_ERR;
         }
         redisLog(REDIS_NOTICE,"Background saving started by pid %d",childpid);

+        server.rdb_save_time_start = time(NULL);

         server.rdb_child_pid = childpid;
         updateDictResizePolicy();
         return REDIS_OK;
         server.rdb_child_pid = childpid;
         updateDictResizePolicy();
         return REDIS_OK;


@@ -720,7 +734,7 @@ robj *rdbLoadObject(int rdbtype, rio *rdb) {
     size_t len;
     unsigned int i;
 
     size_t len;
     unsigned int i;
 

-    redisLog(REDIS_DEBUG,"LOADING OBJECT %d (at %d)\n",rdbtype,rdb->tell(rdb));
+    redisLog(REDIS_DEBUG,"LOADING OBJECT %d (at %d)\n",rdbtype,rioTell(rdb));

     if (rdbtype == REDIS_RDB_TYPE_STRING) {
         /* Read string value */
         if ((o = rdbLoadEncodedStringObject(rdb)) == NULL) return NULL;
     if (rdbtype == REDIS_RDB_TYPE_STRING) {
         /* Read string value */
         if ((o = rdbLoadEncodedStringObject(rdb)) == NULL) return NULL;


@@ -789,7 +803,7 @@ robj *rdbLoadObject(int rdbtype, rio *rdb) {
             }
 
             /* This will also be called when the set was just converted
             }
 
             /* This will also be called when the set was just converted

-             * to regular hashtable encoded set */
+             * to a regular hash table encoded set */

             if (o->encoding == REDIS_ENCODING_HT) {
                 dictAdd((dict*)o->ptr,ele,NULL);
             } else {
             if (o->encoding == REDIS_ENCODING_HT) {
                 dictAdd((dict*)o->ptr,ele,NULL);
             } else {


@@ -859,14 +873,17 @@ robj *rdbLoadObject(int rdbtype, rio *rdb) {
             /* Add pair to ziplist */
             o->ptr = ziplistPush(o->ptr, field->ptr, sdslen(field->ptr), ZIPLIST_TAIL);
             o->ptr = ziplistPush(o->ptr, value->ptr, sdslen(value->ptr), ZIPLIST_TAIL);
             /* Add pair to ziplist */
             o->ptr = ziplistPush(o->ptr, field->ptr, sdslen(field->ptr), ZIPLIST_TAIL);
             o->ptr = ziplistPush(o->ptr, value->ptr, sdslen(value->ptr), ZIPLIST_TAIL);

-

             /* Convert to hash table if size threshold is exceeded */
             if (sdslen(field->ptr) > server.hash_max_ziplist_value ||
                 sdslen(value->ptr) > server.hash_max_ziplist_value)
             {
             /* Convert to hash table if size threshold is exceeded */
             if (sdslen(field->ptr) > server.hash_max_ziplist_value ||
                 sdslen(value->ptr) > server.hash_max_ziplist_value)
             {

+                decrRefCount(field);
+                decrRefCount(value);

                 hashTypeConvert(o, REDIS_ENCODING_HT);
                 break;
             }
                 hashTypeConvert(o, REDIS_ENCODING_HT);
                 break;
             }

+            decrRefCount(field);
+            decrRefCount(value);

         }
 
         /* Load remaining fields and values into the hash table */
         }
 
         /* Load remaining fields and values into the hash table */


@@ -1016,6 +1033,8 @@ int rdbLoad(char *filename) {
         return REDIS_ERR;
     }
     rioInitWithFile(&rdb,fp);
         return REDIS_ERR;
     }
     rioInitWithFile(&rdb,fp);

+    if (server.rdb_checksum)
+        rdb.update_cksum = rioGenericUpdateChecksum;

     if (rioRead(&rdb,buf,9) == 0) goto eoferr;
     buf[9] = '\0';
     if (memcmp(buf,"REDIS",5) != 0) {
     if (rioRead(&rdb,buf,9) == 0) goto eoferr;
     buf[9] = '\0';
     if (memcmp(buf,"REDIS",5) != 0) {


@@ -1025,7 +1044,7 @@ int rdbLoad(char *filename) {
         return REDIS_ERR;
     }
     rdbver = atoi(buf+5);
         return REDIS_ERR;
     }
     rdbver = atoi(buf+5);

-    if (rdbver < 1 || rdbver > 4) {
+    if (rdbver < 1 || rdbver > REDIS_RDB_VERSION) {

         fclose(fp);
         redisLog(REDIS_WARNING,"Can't handle RDB format version %d",rdbver);
         errno = EINVAL;
         fclose(fp);
         redisLog(REDIS_WARNING,"Can't handle RDB format version %d",rdbver);
         errno = EINVAL;


@@ -1039,7 +1058,7 @@ int rdbLoad(char *filename) {
 
         /* Serve the clients from time to time */
         if (!(loops++ % 1000)) {
 
         /* Serve the clients from time to time */
         if (!(loops++ % 1000)) {

-            loadingProgress(rdb.tell(&rdb));
+            loadingProgress(rioTell(&rdb));

             aeProcessEvents(server.el, AE_FILE_EVENTS|AE_DONT_WAIT);
         }
 
             aeProcessEvents(server.el, AE_FILE_EVENTS|AE_DONT_WAIT);
         }
 


@@ -1096,6 +1115,20 @@ int rdbLoad(char *filename) {
 
         decrRefCount(key);
     }
 
         decrRefCount(key);
     }

+    /* Verify the checksum if RDB version is >= 5 */
+    if (rdbver >= 5 && server.rdb_checksum) {
+        uint64_t cksum, expected = rdb.cksum;
+
+        if (rioRead(&rdb,&cksum,8) == 0) goto eoferr;
+        memrev64ifbe(&cksum);
+        if (cksum == 0) {
+            redisLog(REDIS_WARNING,"RDB file was saved with checksum disabled: no check performed.");
+        } else if (cksum != expected) {
+            redisLog(REDIS_WARNING,"Wrong RDB checksum. Aborting now.");
+            exit(1);
+        }
+    }
+

     fclose(fp);
     stopLoading();
     return REDIS_OK;
     fclose(fp);
     stopLoading();
     return REDIS_OK;


@@ -1124,6 +1157,8 @@ void backgroundSaveDoneHandler(int exitcode, int bysignal) {
         server.lastbgsave_status = REDIS_ERR;
     }
     server.rdb_child_pid = -1;
         server.lastbgsave_status = REDIS_ERR;
     }
     server.rdb_child_pid = -1;

+    server.rdb_save_time_last = time(NULL)-server.rdb_save_time_start;
+    server.rdb_save_time_start = -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);
     /* 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);