handled DEL command as a special optimized case for disk store

[redis.git] / src / redis.h

diff --git a/src/redis.h b/src/redis.h
index 0b78320c117db7ad0d304afcfb075ff26b650a86..15c192cfa02c6a2a72b2c098451047a6ef8b2146 100644 (file)
--- a/src/redis.h
+++ b/src/redis.h
@@ -17,6 +17,7 @@
 #include <errno.h>
 #include <inttypes.h>
 #include <pthread.h>
 #include <errno.h>
 #include <inttypes.h>
 #include <pthread.h>

+#include <syslog.h>

 
 #include "ae.h"     /* Event driven programming library */
 #include "sds.h"    /* Dynamic safe strings */
 
 #include "ae.h"     /* Event driven programming library */
 #include "sds.h"    /* Dynamic safe strings */


@@ -41,12 +42,13 @@
 #define REDIS_STATIC_ARGS       8
 #define REDIS_DEFAULT_DBNUM     16
 #define REDIS_CONFIGLINE_MAX    1024
 #define REDIS_STATIC_ARGS       8
 #define REDIS_DEFAULT_DBNUM     16
 #define REDIS_CONFIGLINE_MAX    1024

-#define REDIS_OBJFREELIST_MAX   1000000 /* Max number of objects to cache */

 #define REDIS_MAX_SYNC_TIME     60      /* Slave can't take more to sync */
 #define REDIS_EXPIRELOOKUPS_PER_CRON    10 /* lookup 10 expires per loop */
 #define REDIS_MAX_WRITE_PER_EVENT (1024*64)
 #define REDIS_REQUEST_MAX_SIZE (1024*1024*256) /* max bytes in inline command */
 #define REDIS_SHARED_INTEGERS 10000
 #define REDIS_MAX_SYNC_TIME     60      /* Slave can't take more to sync */
 #define REDIS_EXPIRELOOKUPS_PER_CRON    10 /* lookup 10 expires per loop */
 #define REDIS_MAX_WRITE_PER_EVENT (1024*64)
 #define REDIS_REQUEST_MAX_SIZE (1024*1024*256) /* max bytes in inline command */
 #define REDIS_SHARED_INTEGERS 10000

+#define REDIS_REPLY_CHUNK_BYTES (5*1500) /* 5 TCP packets with default MTU */
+#define REDIS_MAX_LOGMSG_LEN    1024 /* Default maximum length of syslog messages */

 
 /* If more then REDIS_WRITEV_THRESHOLD write packets are pending use writev */
 #define REDIS_WRITEV_THRESHOLD      3
 
 /* If more then REDIS_WRITEV_THRESHOLD write packets are pending use writev */
 #define REDIS_WRITEV_THRESHOLD      3


@@ -56,15 +58,15 @@
 /* Hash table parameters */
 #define REDIS_HT_MINFILL        10      /* Minimal hash table fill 10% */
 
 /* Hash table parameters */
 #define REDIS_HT_MINFILL        10      /* Minimal hash table fill 10% */
 

-/* Command flags */
-#define REDIS_CMD_BULK          1       /* Bulk write command */
-#define REDIS_CMD_INLINE        2       /* Inline command */
-/* REDIS_CMD_DENYOOM reserves a longer comment: all the commands marked with
-   this flags will return an error when the 'maxmemory' option is set in the
-   config file and the server is using more than maxmemory bytes of memory.
-   In short this commands are denied on low memory conditions. */
-#define REDIS_CMD_DENYOOM       4
-#define REDIS_CMD_FORCE_REPLICATION 8 /* Force replication even if dirty is 0 */
+/* Command flags:
+ *   REDIS_CMD_DENYOOM:
+ *     Commands marked with this flag will return an error when 'maxmemory' is
+ *     set and the server is using more than 'maxmemory' bytes of memory.
+ *     In short: commands with this flag are denied on low memory conditions.
+ *   REDIS_CMD_FORCE_REPLICATION:
+ *     Force replication even if dirty is 0. */
+#define REDIS_CMD_DENYOOM 4
+#define REDIS_CMD_FORCE_REPLICATION 8

 
 /* Object types */
 #define REDIS_STRING 0
 
 /* Object types */
 #define REDIS_STRING 0


@@ -117,23 +119,13 @@
 #define REDIS_RDB_ENC_INT32 2       /* 32 bit signed integer */
 #define REDIS_RDB_ENC_LZF 3         /* string compressed with FASTLZ */
 
 #define REDIS_RDB_ENC_INT32 2       /* 32 bit signed integer */
 #define REDIS_RDB_ENC_LZF 3         /* string compressed with FASTLZ */
 

-/* Virtual memory object->where field. */
-#define REDIS_VM_MEMORY 0       /* The object is on memory */
-#define REDIS_VM_SWAPPED 1      /* The object is on disk */
-#define REDIS_VM_SWAPPING 2     /* Redis is swapping this object on disk */
-#define REDIS_VM_LOADING 3      /* Redis is loading this object from disk */
-
-/* Virtual memory static configuration stuff.
- * Check vmFindContiguousPages() to know more about this magic numbers. */
-#define REDIS_VM_MAX_NEAR_PAGES 65536
-#define REDIS_VM_MAX_RANDOM_JUMP 4096
-#define REDIS_VM_MAX_THREADS 32
-#define REDIS_THREAD_STACK_SIZE (1024*1024*4)
-/* The following is the *percentage* of completed I/O jobs to process when the
- * handelr is called. While Virtual Memory I/O operations are performed by
- * threads, this operations must be processed by the main thread when completed
- * in order to take effect. */
+/* Disk store cache object->storage values */
+#define REDIS_DS_MEMORY 0       /* The object is on memory */
+#define REDIS_DS_DIRTY 1        /* The object was modified */
+#define REDIS_DS_SAVING 2       /* There is an IO Job created for this obj. */
+

 #define REDIS_MAX_COMPLETED_JOBS_PROCESSED 1
 #define REDIS_MAX_COMPLETED_JOBS_PROCESSED 1

+#define REDIS_THREAD_STACK_SIZE (1024*1024*4)

 
 /* Client flags */
 #define REDIS_SLAVE 1       /* This client is a slave server */
 
 /* Client flags */
 #define REDIS_SLAVE 1       /* This client is a slave server */


@@ -143,11 +135,17 @@
 #define REDIS_BLOCKED 16    /* The client is waiting in a blocking operation */
 #define REDIS_IO_WAIT 32    /* The client is waiting for Virtual Memory I/O */
 #define REDIS_DIRTY_CAS 64  /* Watched keys modified. EXEC will fail. */
 #define REDIS_BLOCKED 16    /* The client is waiting in a blocking operation */
 #define REDIS_IO_WAIT 32    /* The client is waiting for Virtual Memory I/O */
 #define REDIS_DIRTY_CAS 64  /* Watched keys modified. EXEC will fail. */

+#define REDIS_CLOSE_AFTER_REPLY 128 /* Close after writing entire reply. */
+
+/* Client request types */
+#define REDIS_REQ_INLINE 1
+#define REDIS_REQ_MULTIBULK 2

 
 /* Slave replication state - slave side */
 #define REDIS_REPL_NONE 0   /* No active replication */
 #define REDIS_REPL_CONNECT 1    /* Must connect to master */
 
 /* Slave replication state - slave side */
 #define REDIS_REPL_NONE 0   /* No active replication */
 #define REDIS_REPL_CONNECT 1    /* Must connect to master */

-#define REDIS_REPL_CONNECTED 2  /* Connected to master */
+#define REDIS_REPL_TRANSFER 2    /* Receiving .rdb from master */
+#define REDIS_REPL_CONNECTED 3  /* Connected to master */

 
 /* Slave replication state - from the point of view of master
  * Note that in SEND_BULK and ONLINE state the slave receives new updates
 
 /* Slave replication state - from the point of view of master
  * Note that in SEND_BULK and ONLINE state the slave receives new updates


@@ -188,15 +186,23 @@
 /* Zip structure related defaults */
 #define REDIS_HASH_MAX_ZIPMAP_ENTRIES 64
 #define REDIS_HASH_MAX_ZIPMAP_VALUE 512
 /* Zip structure related defaults */
 #define REDIS_HASH_MAX_ZIPMAP_ENTRIES 64
 #define REDIS_HASH_MAX_ZIPMAP_VALUE 512

-#define REDIS_LIST_MAX_ZIPLIST_ENTRIES 1024
-#define REDIS_LIST_MAX_ZIPLIST_VALUE 32
-#define REDIS_SET_MAX_INTSET_ENTRIES 4096
+#define REDIS_LIST_MAX_ZIPLIST_ENTRIES 512
+#define REDIS_LIST_MAX_ZIPLIST_VALUE 64
+#define REDIS_SET_MAX_INTSET_ENTRIES 512

 
 /* Sets operations codes */
 #define REDIS_OP_UNION 0
 #define REDIS_OP_DIFF 1
 #define REDIS_OP_INTER 2
 
 
 /* Sets operations codes */
 #define REDIS_OP_UNION 0
 #define REDIS_OP_DIFF 1
 #define REDIS_OP_INTER 2
 

+/* Redis maxmemory strategies */
+#define REDIS_MAXMEMORY_VOLATILE_LRU 0
+#define REDIS_MAXMEMORY_VOLATILE_TTL 1
+#define REDIS_MAXMEMORY_VOLATILE_RANDOM 2
+#define REDIS_MAXMEMORY_ALLKEYS_LRU 3
+#define REDIS_MAXMEMORY_ALLKEYS_RANDOM 4
+#define REDIS_MAXMEMORY_NO_EVICTION 5
+

 /* We can print the stacktrace, so our assert is defined this way: */
 #define redisAssert(_e) ((_e)?(void)0 : (_redisAssert(#_e,__FILE__,__LINE__),_exit(1)))
 #define redisPanic(_e) _redisPanic(#_e,__FILE__,__LINE__),_exit(1)
 /* We can print the stacktrace, so our assert is defined this way: */
 #define redisAssert(_e) ((_e)?(void)0 : (_redisAssert(#_e,__FILE__,__LINE__),_exit(1)))
 #define redisPanic(_e) _redisPanic(#_e,__FILE__,__LINE__),_exit(1)


@@ -210,6 +216,8 @@ void _redisPanic(char *msg, char *file, int line);
 /* A redis object, that is a type able to hold a string / list / set */
 
 /* The actual Redis Object */
 /* A redis object, that is a type able to hold a string / list / set */
 
 /* The actual Redis Object */

+#define REDIS_LRU_CLOCK_MAX ((1<<21)-1) /* Max value of obj->lru */
+#define REDIS_LRU_CLOCK_RESOLUTION 10 /* LRU clock resolution in seconds */

 typedef struct redisObject {
     unsigned type:4;
     unsigned storage:2;     /* REDIS_VM_MEMORY or REDIS_VM_SWAPPING */
 typedef struct redisObject {
     unsigned type:4;
     unsigned storage:2;     /* REDIS_VM_MEMORY or REDIS_VM_SWAPPING */


@@ -253,7 +261,7 @@ typedef struct vmPointer {
     _var.type = REDIS_STRING; \
     _var.encoding = REDIS_ENCODING_RAW; \
     _var.ptr = _ptr; \
     _var.type = REDIS_STRING; \
     _var.encoding = REDIS_ENCODING_RAW; \
     _var.ptr = _ptr; \

-    _var.storage = REDIS_VM_MEMORY; \
+    _var.storage = REDIS_DS_MEMORY; \

 } while(0);
 
 typedef struct redisDb {
 } while(0);
 
 typedef struct redisDb {


@@ -277,6 +285,16 @@ typedef struct multiState {
     int count;              /* Total number of MULTI commands */
 } multiState;
 
     int count;              /* Total number of MULTI commands */
 } multiState;
 

+typedef struct blockingState {
+    robj **keys;            /* The key we are waiting to terminate a blocking
+                             * operation such as BLPOP. Otherwise NULL. */
+    int count;              /* Number of blocking keys */
+    time_t timeout;         /* Blocking operation timeout. If UNIX current time
+                             * is >= timeout then the operation timed out. */
+    robj *target;           /* The key that should receive the element,
+                             * for BRPOPLPUSH. */
+} blockingState;
+

 /* With multiplexing we need to take per-clinet state.
  * Clients are taken in a liked list. */
 typedef struct redisClient {
 /* With multiplexing we need to take per-clinet state.
  * Clients are taken in a liked list. */
 typedef struct redisClient {


@@ -284,10 +302,11 @@ typedef struct redisClient {
     redisDb *db;
     int dictid;
     sds querybuf;
     redisDb *db;
     int dictid;
     sds querybuf;

-    robj **argv, **mbargv;
-    int argc, mbargc;
-    int bulklen;            /* bulk read len. -1 if not in bulk read mode */
-    int multibulk;          /* multi bulk command format active */
+    int argc;
+    robj **argv;
+    int reqtype;
+    int multibulklen;       /* number of multi bulk arguments left to read */
+    long bulklen;           /* length of bulk argument in multi bulk request */

     list *reply;
     int sentlen;
     time_t lastinteraction; /* time of the last interaction, used for timeout */
     list *reply;
     int sentlen;
     time_t lastinteraction; /* time of the last interaction, used for timeout */


@@ -299,16 +318,16 @@ typedef struct redisClient {
     long repldboff;         /* replication DB file offset */
     off_t repldbsize;       /* replication DB file size */
     multiState mstate;      /* MULTI/EXEC state */
     long repldboff;         /* replication DB file offset */
     off_t repldbsize;       /* replication DB file size */
     multiState mstate;      /* MULTI/EXEC state */

-    robj **blocking_keys;   /* The key we are waiting to terminate a blocking
-                             * operation such as BLPOP. Otherwise NULL. */
-    int blocking_keys_num;  /* Number of blocking keys */
-    time_t blockingto;      /* Blocking operation timeout. If UNIX current time
-                             * is >= blockingto then the operation timed out. */
+    blockingState bpop;   /* blocking state */

     list *io_keys;          /* Keys this client is waiting to be loaded from the
                              * swap file in order to continue. */
     list *watched_keys;     /* Keys WATCHED for MULTI/EXEC CAS */
     dict *pubsub_channels;  /* channels a client is interested in (SUBSCRIBE) */
     list *pubsub_patterns;  /* patterns a client is interested in (SUBSCRIBE) */
     list *io_keys;          /* Keys this client is waiting to be loaded from the
                              * swap file in order to continue. */
     list *watched_keys;     /* Keys WATCHED for MULTI/EXEC CAS */
     dict *pubsub_channels;  /* channels a client is interested in (SUBSCRIBE) */
     list *pubsub_patterns;  /* patterns a client is interested in (SUBSCRIBE) */

+
+    /* Response buffer */
+    int bufpos;
+    char buf[REDIS_REPLY_CHUNK_BYTES];

 } redisClient;
 
 struct saveparam {
 } redisClient;
 
 struct saveparam {


@@ -320,7 +339,7 @@ struct sharedObjectsStruct {
     robj *crlf, *ok, *err, *emptybulk, *czero, *cone, *cnegone, *pong, *space,
     *colon, *nullbulk, *nullmultibulk, *queued,
     *emptymultibulk, *wrongtypeerr, *nokeyerr, *syntaxerr, *sameobjecterr,
     robj *crlf, *ok, *err, *emptybulk, *czero, *cone, *cnegone, *pong, *space,
     *colon, *nullbulk, *nullmultibulk, *queued,
     *emptymultibulk, *wrongtypeerr, *nokeyerr, *syntaxerr, *sameobjecterr,

-    *outofrangeerr, *plus,
+    *outofrangeerr, *loadingerr, *plus,

     *select0, *select1, *select2, *select3, *select4,
     *select5, *select6, *select7, *select8, *select9,
     *messagebulk, *pmessagebulk, *subscribebulk, *unsubscribebulk, *mbulk3,
     *select0, *select1, *select2, *select3, *select4,
     *select5, *select6, *select7, *select8, *select9,
     *messagebulk, *pmessagebulk, *subscribebulk, *unsubscribebulk, *mbulk3,


@@ -332,21 +351,35 @@ struct sharedObjectsStruct {
 struct redisServer {
     pthread_t mainthread;
     int port;
 struct redisServer {
     pthread_t mainthread;
     int port;

-    int fd;
+    char *bindaddr;
+    char *unixsocket;
+    int ipfd;
+    int sofd;

     redisDb *db;
     long long dirty;            /* changes to DB from the last save */
     redisDb *db;
     long long dirty;            /* changes to DB from the last save */

+    long long dirty_before_bgsave; /* used to restore dirty on failed BGSAVE */

     list *clients;
     list *clients;

+    dict *commands;             /* Command table hahs table */
+    /* RDB / AOF loading information */
+    int loading;
+    off_t loading_total_bytes;
+    off_t loading_loaded_bytes;
+    time_t loading_start_time;
+    /* Fast pointers to often looked up command */
+    struct redisCommand *delCommand, *multiCommand;

     list *slaves, *monitors;
     char neterr[ANET_ERR_LEN];
     aeEventLoop *el;
     int cronloops;              /* number of times the cron function run */
     list *slaves, *monitors;
     char neterr[ANET_ERR_LEN];
     aeEventLoop *el;
     int cronloops;              /* number of times the cron function run */

-    list *objfreelist;          /* A list of freed objects to avoid malloc() */
-    time_t lastsave;            /* Unix time of last save succeeede */
+    time_t lastsave;                /* Unix time of last save succeeede */

     /* Fields used only for stats */
     /* Fields used only for stats */

-    time_t stat_starttime;         /* server start time */
-    long long stat_numcommands;    /* number of processed commands */
-    long long stat_numconnections; /* number of connections received */
-    long long stat_expiredkeys;   /* number of expired keys */
+    time_t stat_starttime;          /* server start time */
+    long long stat_numcommands;     /* number of processed commands */
+    long long stat_numconnections;  /* number of connections received */
+    long long stat_expiredkeys;     /* number of expired keys */
+    long long stat_evictedkeys;     /* number of evicted keys (maxmemory) */
+    long long stat_keyspace_hits;   /* number of successful lookups of keys */
+    long long stat_keyspace_misses; /* number of failed lookups of keys */

     /* Configuration */
     int verbosity;
     int glueoutputbuf;
     /* Configuration */
     int verbosity;
     int glueoutputbuf;


@@ -368,7 +401,9 @@ struct redisServer {
     struct saveparam *saveparams;
     int saveparamslen;
     char *logfile;
     struct saveparam *saveparams;
     int saveparamslen;
     char *logfile;

-    char *bindaddr;
+    int syslog_enabled;
+    char *syslog_ident;
+    int syslog_facility;

     char *dbfilename;
     char *appendfilename;
     char *requirepass;
     char *dbfilename;
     char *appendfilename;
     char *requirepass;


@@ -376,38 +411,44 @@ struct redisServer {
     int activerehashing;
     /* Replication related */
     int isslave;
     int activerehashing;
     /* Replication related */
     int isslave;

+    /* Slave specific fields */

     char *masterauth;
     char *masterhost;
     int masterport;
     redisClient *master;    /* client that is master for this slave */
     char *masterauth;
     char *masterhost;
     int masterport;
     redisClient *master;    /* client that is master for this slave */

-    int replstate;
+    int replstate;          /* replication status if the instance is a slave */
+    off_t repl_transfer_left;  /* bytes left reading .rdb  */
+    int repl_transfer_s;    /* slave -> master SYNC socket */
+    int repl_transfer_fd;   /* slave -> master SYNC temp file descriptor */
+    char *repl_transfer_tmpfile; /* slave-> master SYNC temp file name */
+    time_t repl_transfer_lastio; /* unix time of the latest read, for timeout */
+    int repl_serve_stale_data; /* Serve stale data when link is down? */
+    /* Limits */

     unsigned int maxclients;
     unsigned long long maxmemory;
     unsigned int maxclients;
     unsigned long long maxmemory;

-    unsigned int blpop_blocked_clients;
-    unsigned int vm_blocked_clients;
+    int maxmemory_policy;
+    int maxmemory_samples;
+    /* Blocked clients */
+    unsigned int bpop_blocked_clients;
+    unsigned int cache_blocked_clients;
+    list *unblocked_clients; /* list of clients to unblock before next loop */
+    list *cache_flush_queue; /* keys to flush on disk */
+    int cache_flush_delay;   /* seconds to wait before flushing keys */

     /* Sort parameters - qsort_r() is only available under BSD so we
      * have to take this state global, in order to pass it to sortCompare() */
     int sort_desc;
     int sort_alpha;
     int sort_bypattern;
     /* Virtual memory configuration */
     /* Sort parameters - qsort_r() is only available under BSD so we
      * have to take this state global, in order to pass it to sortCompare() */
     int sort_desc;
     int sort_alpha;
     int sort_bypattern;
     /* Virtual memory configuration */

-    int vm_enabled;
-    char *vm_swap_file;
-    off_t vm_page_size;
-    off_t vm_pages;
-    unsigned long long vm_max_memory;
+    int ds_enabled; /* backend disk in redis.conf */
+    char *ds_path;  /* location of the disk store on disk */
+    unsigned long long cache_max_memory;

     /* Zip structure config */
     size_t hash_max_zipmap_entries;
     size_t hash_max_zipmap_value;
     size_t list_max_ziplist_entries;
     size_t list_max_ziplist_value;
     size_t set_max_intset_entries;
     /* Zip structure config */
     size_t hash_max_zipmap_entries;
     size_t hash_max_zipmap_value;
     size_t list_max_ziplist_entries;
     size_t list_max_ziplist_value;
     size_t set_max_intset_entries;

-    /* Virtual memory state */
-    FILE *vm_fp;
-    int vm_fd;
-    off_t vm_next_page; /* Next probably empty page */
-    off_t vm_near_pages; /* Number of pages allocated sequentially */
-    unsigned char *vm_bitmap; /* Bitmap of free/used pages */

     time_t unixtime;    /* Unix time sampled every second. */
     /* Virtual memory I/O threads stuff */
     /* An I/O thread process an element taken from the io_jobs queue and
     time_t unixtime;    /* Unix time sampled every second. */
     /* Virtual memory I/O threads stuff */
     /* An I/O thread process an element taken from the io_jobs queue and


@@ -418,8 +459,7 @@ struct redisServer {
     list *io_processed; /* List of VM I/O jobs already processed */
     list *io_ready_clients; /* Clients ready to be unblocked. All keys loaded */
     pthread_mutex_t io_mutex; /* lock to access io_jobs/io_done/io_thread_job */
     list *io_processed; /* List of VM I/O jobs already processed */
     list *io_ready_clients; /* Clients ready to be unblocked. All keys loaded */
     pthread_mutex_t io_mutex; /* lock to access io_jobs/io_done/io_thread_job */

-    pthread_mutex_t obj_freelist_mutex; /* safe redis objects creation/free */
-    pthread_mutex_t io_swapfile_mutex; /* So we can lseek + write */
+    pthread_cond_t io_condvar; /* I/O threads conditional variable */

     pthread_attr_t io_threads_attr; /* attributes for threads creation */
     int io_active_threads; /* Number of running I/O threads */
     int vm_max_threads; /* Max number of I/O threads running at the same time */
     pthread_attr_t io_threads_attr; /* attributes for threads creation */
     int io_active_threads; /* Number of running I/O threads */
     int vm_max_threads; /* Max number of I/O threads running at the same time */


@@ -438,7 +478,6 @@ struct redisServer {
     dict *pubsub_channels; /* Map channels to list of subscribed clients */
     list *pubsub_patterns; /* A list of pubsub_patterns */
     /* Misc */
     dict *pubsub_channels; /* Map channels to list of subscribed clients */
     list *pubsub_patterns; /* A list of pubsub_patterns */
     /* Misc */

-    FILE *devnull;

     unsigned lruclock:22;        /* clock incrementing every minute, for LRU */
     unsigned lruclock_padding:10;
 };
     unsigned lruclock:22;        /* clock incrementing every minute, for LRU */
     unsigned lruclock_padding:10;
 };


@@ -484,13 +523,14 @@ typedef struct _redisSortOperation {
 } redisSortOperation;
 
 /* ZSETs use a specialized version of Skiplists */
 } redisSortOperation;
 
 /* ZSETs use a specialized version of Skiplists */

-

 typedef struct zskiplistNode {
 typedef struct zskiplistNode {

-    struct zskiplistNode **forward;
-    struct zskiplistNode *backward;
-    unsigned int *span;
-    double score;

     robj *obj;
     robj *obj;

+    double score;
+    struct zskiplistNode *backward;
+    struct zskiplistLevel {
+        struct zskiplistNode *forward;
+        unsigned int span;
+    } level[];

 } zskiplistNode;
 
 typedef struct zskiplist {
 } zskiplistNode;
 
 typedef struct zskiplist {


@@ -505,24 +545,26 @@ typedef struct zset {
 } zset;
 
 /* VM threaded I/O request message */
 } zset;
 
 /* VM threaded I/O request message */

-#define REDIS_IOJOB_LOAD 0          /* Load from disk to memory */
-#define REDIS_IOJOB_PREPARE_SWAP 1  /* Compute needed pages */
-#define REDIS_IOJOB_DO_SWAP 2       /* Swap from memory to disk */
+#define REDIS_IOJOB_LOAD 0
+#define REDIS_IOJOB_SAVE 1
+

 typedef struct iojob {
     int type;   /* Request type, REDIS_IOJOB_* */
     redisDb *db;/* Redis database */
 typedef struct iojob {
     int type;   /* Request type, REDIS_IOJOB_* */
     redisDb *db;/* Redis database */

-    robj *key;  /* This I/O request is about swapping this key */
-    robj *id;   /* Unique identifier of this job:
-                   this is the object to swap for REDIS_IOREQ_*_SWAP, or the
-                   vmpointer objct for REDIS_IOREQ_LOAD. */
-    robj *val;  /* the value to swap for REDIS_IOREQ_*_SWAP, otherwise this
-                 * field is populated by the I/O thread for REDIS_IOREQ_LOAD. */
-    off_t page; /* Swap page where to read/write the object */
-    off_t pages; /* Swap pages needed to save object. PREPARE_SWAP return val */
-    int canceled; /* True if this command was canceled by blocking side of VM */
-    pthread_t thread; /* ID of the thread processing this entry */
+    robj *key;  /* This I/O request is about this key */
+    robj *val;  /* the value to swap for REDIS_IOJOB_SAVE, otherwise this
+                 * field is populated by the I/O thread for REDIS_IOJOB_LOAD. */
+    time_t expire; /* Expire time for this key on REDIS_IOJOB_LOAD */

 } iojob;
 
 } iojob;
 

+/* When diskstore is enabled and a flush operation is requested we push
+ * one of this structures into server.cache_flush_queue. */
+typedef struct dirtykey {
+    redisDb *db;
+    robj *key;
+    time_t ctime; /* This is the creation time of the entry. */
+} dirtykey;
+

 /* Structure to hold list iteration abstraction. */
 typedef struct {
     robj *subject;
 /* Structure to hold list iteration abstraction. */
 typedef struct {
     robj *subject;


@@ -587,20 +629,37 @@ void resetClient(redisClient *c);
 void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask);
 void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReply(redisClient *c, robj *obj);
 void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask);
 void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReply(redisClient *c, robj *obj);

+void *addDeferredMultiBulkLength(redisClient *c);
+void setDeferredMultiBulkLength(redisClient *c, void *node, long length);

 void addReplySds(redisClient *c, sds s);
 void processInputBuffer(redisClient *c);
 void addReplySds(redisClient *c, sds s);
 void processInputBuffer(redisClient *c);

-void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask);
+void acceptTcpHandler(aeEventLoop *el, int fd, void *privdata, int mask);
+void acceptUnixHandler(aeEventLoop *el, int fd, void *privdata, int mask);

 void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReplyBulk(redisClient *c, robj *obj);
 void addReplyBulkCString(redisClient *c, char *s);
 void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReplyBulk(redisClient *c, robj *obj);
 void addReplyBulkCString(redisClient *c, char *s);

+void addReplyBulkCBuffer(redisClient *c, void *p, size_t len);
+void addReplyBulkLongLong(redisClient *c, long long ll);

 void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReply(redisClient *c, robj *obj);
 void addReplySds(redisClient *c, sds s);
 void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReply(redisClient *c, robj *obj);
 void addReplySds(redisClient *c, sds s);

+void addReplyError(redisClient *c, char *err);
+void addReplyStatus(redisClient *c, char *status);

 void addReplyDouble(redisClient *c, double d);
 void addReplyLongLong(redisClient *c, long long ll);
 void addReplyDouble(redisClient *c, double d);
 void addReplyLongLong(redisClient *c, long long ll);

-void addReplyUlong(redisClient *c, unsigned long ul);
+void addReplyMultiBulkLen(redisClient *c, long length);

 void *dupClientReplyValue(void *o);
 
 void *dupClientReplyValue(void *o);
 

+#ifdef __GNUC__
+void addReplyErrorFormat(redisClient *c, const char *fmt, ...)
+    __attribute__((format(printf, 2, 3)));
+void addReplyStatusFormat(redisClient *c, const char *fmt, ...)
+    __attribute__((format(printf, 2, 3)));
+#else
+void addReplyErrorFormat(redisClient *c, const char *fmt, ...);
+void addReplyStatusFormat(redisClient *c, const char *fmt, ...);
+#endif
+

 /* List data type */
 void listTypeTryConversion(robj *subject, robj *value);
 void listTypePush(robj *subject, robj *value, int where);
 /* List data type */
 void listTypeTryConversion(robj *subject, robj *value);
 void listTypePush(robj *subject, robj *value, int where);


@@ -640,7 +699,6 @@ robj *dupStringObject(robj *o);
 robj *tryObjectEncoding(robj *o);
 robj *getDecodedObject(robj *o);
 size_t stringObjectLen(robj *o);
 robj *tryObjectEncoding(robj *o);
 robj *getDecodedObject(robj *o);
 size_t stringObjectLen(robj *o);

-int tryFreeOneObjectFromFreelist(void);

 robj *createStringObjectFromLongLong(long long value);
 robj *createListObject(void);
 robj *createZiplistObject(void);
 robj *createStringObjectFromLongLong(long long value);
 robj *createListObject(void);
 robj *createZiplistObject(void);


@@ -656,12 +714,28 @@ int getLongLongFromObject(robj *o, long long *target);
 char *strEncoding(int encoding);
 int compareStringObjects(robj *a, robj *b);
 int equalStringObjects(robj *a, robj *b);
 char *strEncoding(int encoding);
 int compareStringObjects(robj *a, robj *b);
 int equalStringObjects(robj *a, robj *b);

+unsigned long estimateObjectIdleTime(robj *o);
+
+/* Synchronous I/O with timeout */
+int syncWrite(int fd, char *ptr, ssize_t size, int timeout);
+int syncRead(int fd, char *ptr, ssize_t size, int timeout);
+int syncReadLine(int fd, char *ptr, ssize_t size, int timeout);
+int fwriteBulkString(FILE *fp, char *s, unsigned long len);
+int fwriteBulkDouble(FILE *fp, double d);
+int fwriteBulkLongLong(FILE *fp, long long l);
+int fwriteBulkObject(FILE *fp, robj *obj);

 
 /* Replication */
 void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc);
 void replicationFeedMonitors(list *monitors, int dictid, robj **argv, int argc);
 int syncWithMaster(void);
 void updateSlavesWaitingBgsave(int bgsaveerr);
 
 /* Replication */
 void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc);
 void replicationFeedMonitors(list *monitors, int dictid, robj **argv, int argc);
 int syncWithMaster(void);
 void updateSlavesWaitingBgsave(int bgsaveerr);

+void replicationCron(void);
+
+/* Generic persistence functions */
+void startLoading(FILE *fp);
+void loadingProgress(off_t pos);
+void stopLoading(void);

 
 /* RDB persistence */
 int rdbLoad(char *filename);
 
 /* RDB persistence */
 int rdbLoad(char *filename);


@@ -669,10 +743,14 @@ int rdbSaveBackground(char *filename);
 void rdbRemoveTempFile(pid_t childpid);
 int rdbSave(char *filename);
 int rdbSaveObject(FILE *fp, robj *o);
 void rdbRemoveTempFile(pid_t childpid);
 int rdbSave(char *filename);
 int rdbSaveObject(FILE *fp, robj *o);

-off_t rdbSavedObjectPages(robj *o, FILE *fp);
-off_t rdbSavedObjectLen(robj *o, FILE *fp);
+off_t rdbSavedObjectLen(robj *o);
+off_t rdbSavedObjectPages(robj *o);

 robj *rdbLoadObject(int type, FILE *fp);
 void backgroundSaveDoneHandler(int statloc);
 robj *rdbLoadObject(int type, FILE *fp);
 void backgroundSaveDoneHandler(int statloc);

+int rdbSaveKeyValuePair(FILE *fp, redisDb *db, robj *key, robj *val, time_t now);
+int rdbLoadType(FILE *fp);
+time_t rdbLoadTime(FILE *fp);
+robj *rdbLoadStringObject(FILE *fp);

 
 /* AOF persistence */
 void flushAppendOnlyFile(void);
 
 /* AOF persistence */
 void flushAppendOnlyFile(void);


@@ -687,13 +765,14 @@ void backgroundRewriteDoneHandler(int statloc);
 /* Sorted sets data type */
 zskiplist *zslCreate(void);
 void zslFree(zskiplist *zsl);
 /* Sorted sets data type */
 zskiplist *zslCreate(void);
 void zslFree(zskiplist *zsl);

-void zslInsert(zskiplist *zsl, double score, robj *obj);
+zskiplistNode *zslInsert(zskiplist *zsl, double score, robj *obj);

 
 /* Core functions */
 void freeMemoryIfNeeded(void);
 int processCommand(redisClient *c);
 void setupSigSegvAction(void);
 
 /* Core functions */
 void freeMemoryIfNeeded(void);
 int processCommand(redisClient *c);
 void setupSigSegvAction(void);

-struct redisCommand *lookupCommand(char *name);
+struct redisCommand *lookupCommand(sds name);
+struct redisCommand *lookupCommandByCString(char *s);

 void call(redisClient *c, struct redisCommand *cmd);
 int prepareForShutdown();
 void redisLog(int level, const char *fmt, ...);
 void call(redisClient *c, struct redisCommand *cmd);
 int prepareForShutdown();
 void redisLog(int level, const char *fmt, ...);


@@ -701,33 +780,35 @@ void usage();
 void updateDictResizePolicy(void);
 int htNeedsResize(dict *dict);
 void oom(const char *msg);
 void updateDictResizePolicy(void);
 int htNeedsResize(dict *dict);
 void oom(const char *msg);

-
-/* Virtual Memory */
-void vmInit(void);
-void vmMarkPagesFree(off_t page, off_t count);
-robj *vmLoadObject(robj *o);
-robj *vmPreviewObject(robj *o);
-int vmSwapOneObjectBlocking(void);
-int vmSwapOneObjectThreaded(void);
-int vmCanSwapOut(void);
+void populateCommandTable(void);
+
+/* Disk store */
+int dsOpen(void);
+int dsClose(void);
+int dsSet(redisDb *db, robj *key, robj *val);
+robj *dsGet(redisDb *db, robj *key, time_t *expire);
+int dsDel(redisDb *db, robj *key);
+int dsExists(redisDb *db, robj *key);
+int dsFlushDb(int dbid);
+
+/* Disk Store Cache */
+void dsInit(void);

 void vmThreadedIOCompletedJob(aeEventLoop *el, int fd, void *privdata, int mask);
 void vmThreadedIOCompletedJob(aeEventLoop *el, int fd, void *privdata, int mask);

-void vmCancelThreadedIOJob(robj *o);

 void lockThreadedIO(void);
 void unlockThreadedIO(void);
 void lockThreadedIO(void);
 void unlockThreadedIO(void);

-int vmSwapObjectThreaded(robj *key, robj *val, redisDb *db);

 void freeIOJob(iojob *j);
 void queueIOJob(iojob *j);
 void freeIOJob(iojob *j);
 void queueIOJob(iojob *j);

-int vmWriteObjectOnSwap(robj *o, off_t page);
-robj *vmReadObjectFromSwap(off_t page, int type);

 void waitEmptyIOJobsQueue(void);
 void waitEmptyIOJobsQueue(void);

-void vmReopenSwapFile(void);
-int vmFreePage(off_t page);
+void processAllPendingIOJobs(void);

 void zunionInterBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv);
 void execBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv);
 int blockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd);
 int dontWaitForSwappedKey(redisClient *c, robj *key);
 void handleClientsBlockedOnSwappedKey(redisDb *db, robj *key);
 void zunionInterBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv);
 void execBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv);
 int blockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd);
 int dontWaitForSwappedKey(redisClient *c, robj *key);
 void handleClientsBlockedOnSwappedKey(redisDb *db, robj *key);

-vmpointer *vmSwapObjectBlocking(robj *val);
+int cacheFreeOneEntry(void);
+void cacheScheduleForFlush(redisDb *db, robj *key);
+void cacheCron(void);
+int cacheKeyMayExist(redisDb *db, robj *key);

 
 /* Set data type */
 robj *setTypeCreate(robj *value);
 
 /* Set data type */
 robj *setTypeCreate(robj *value);


@@ -736,8 +817,9 @@ int setTypeRemove(robj *subject, robj *value);
 int setTypeIsMember(robj *subject, robj *value);
 setTypeIterator *setTypeInitIterator(robj *subject);
 void setTypeReleaseIterator(setTypeIterator *si);
 int setTypeIsMember(robj *subject, robj *value);
 setTypeIterator *setTypeInitIterator(robj *subject);
 void setTypeReleaseIterator(setTypeIterator *si);

-robj *setTypeNext(setTypeIterator *si);
-robj *setTypeRandomElement(robj *subject);
+int setTypeNext(setTypeIterator *si, robj **objele, int64_t *llele);
+robj *setTypeNextObject(setTypeIterator *si);
+int setTypeRandomElement(robj *setobj, robj **objele, int64_t *llele);

 unsigned long setTypeSize(robj *subject);
 void setTypeConvert(robj *subject, int enc);
 
 unsigned long setTypeSize(robj *subject);
 void setTypeConvert(robj *subject, int enc);
 


@@ -745,7 +827,8 @@ void setTypeConvert(robj *subject, int enc);
 void convertToRealHash(robj *o);
 void hashTypeTryConversion(robj *subject, robj **argv, int start, int end);
 void hashTypeTryObjectEncoding(robj *subject, robj **o1, robj **o2);
 void convertToRealHash(robj *o);
 void hashTypeTryConversion(robj *subject, robj **argv, int start, int end);
 void hashTypeTryObjectEncoding(robj *subject, robj **o1, robj **o2);

-robj *hashTypeGet(robj *o, robj *key);
+int hashTypeGet(robj *o, robj *key, robj **objval, unsigned char **v, unsigned int *vlen);
+robj *hashTypeGetObject(robj *o, robj *key);

 int hashTypeExists(robj *o, robj *key);
 int hashTypeSet(robj *o, robj *key, robj *value);
 int hashTypeDelete(robj *o, robj *key);
 int hashTypeExists(robj *o, robj *key);
 int hashTypeSet(robj *o, robj *key, robj *value);
 int hashTypeDelete(robj *o, robj *key);


@@ -753,7 +836,8 @@ unsigned long hashTypeLength(robj *o);
 hashTypeIterator *hashTypeInitIterator(robj *subject);
 void hashTypeReleaseIterator(hashTypeIterator *hi);
 int hashTypeNext(hashTypeIterator *hi);
 hashTypeIterator *hashTypeInitIterator(robj *subject);
 void hashTypeReleaseIterator(hashTypeIterator *hi);
 int hashTypeNext(hashTypeIterator *hi);

-robj *hashTypeCurrent(hashTypeIterator *hi, int what);
+int hashTypeCurrent(hashTypeIterator *hi, int what, robj **objval, unsigned char **v, unsigned int *vlen);
+robj *hashTypeCurrentObject(hashTypeIterator *hi, int what);

 robj *hashTypeLookupWriteOrCreate(redisClient *c, robj *key);
 
 /* Pub / Sub */
 robj *hashTypeLookupWriteOrCreate(redisClient *c, robj *key);
 
 /* Pub / Sub */


@@ -769,6 +853,8 @@ int stringmatch(const char *pattern, const char *string, int nocase);
 long long memtoll(const char *p, int *err);
 int ll2string(char *s, size_t len, long long value);
 int isStringRepresentableAsLong(sds s, long *longval);
 long long memtoll(const char *p, int *err);
 int ll2string(char *s, size_t len, long long value);
 int isStringRepresentableAsLong(sds s, long *longval);

+int isStringRepresentableAsLongLong(sds s, long long *longval);
+int isObjectRepresentableAsLongLong(robj *o, long long *llongval);

 
 /* Configuration */
 void loadServerConfig(char *filename);
 
 /* Configuration */
 void loadServerConfig(char *filename);


@@ -793,6 +879,8 @@ robj *dbRandomKey(redisDb *db);
 int dbDelete(redisDb *db, robj *key);
 long long emptyDb();
 int selectDb(redisClient *c, int id);
 int dbDelete(redisDb *db, robj *key);
 long long emptyDb();
 int selectDb(redisClient *c, int id);

+void signalModifiedKey(redisDb *db, robj *key);
+void signalFlushedDb(int dbid);

 
 /* Git SHA1 */
 char *redisGitSHA1(void);
 
 /* Git SHA1 */
 char *redisGitSHA1(void);


@@ -808,6 +896,10 @@ void setexCommand(redisClient *c);
 void getCommand(redisClient *c);
 void delCommand(redisClient *c);
 void existsCommand(redisClient *c);
 void getCommand(redisClient *c);
 void delCommand(redisClient *c);
 void existsCommand(redisClient *c);

+void setbitCommand(redisClient *c);
+void getbitCommand(redisClient *c);
+void setrangeCommand(redisClient *c);
+void getrangeCommand(redisClient *c);

 void incrCommand(redisClient *c);
 void decrCommand(redisClient *c);
 void incrbyCommand(redisClient *c);
 void incrCommand(redisClient *c);
 void decrCommand(redisClient *c);
 void incrbyCommand(redisClient *c);


@@ -855,7 +947,7 @@ void flushdbCommand(redisClient *c);
 void flushallCommand(redisClient *c);
 void sortCommand(redisClient *c);
 void lremCommand(redisClient *c);
 void flushallCommand(redisClient *c);
 void sortCommand(redisClient *c);
 void lremCommand(redisClient *c);

-void rpoplpushcommand(redisClient *c);
+void rpoplpushCommand(redisClient *c);

 void infoCommand(redisClient *c);
 void mgetCommand(redisClient *c);
 void monitorCommand(redisClient *c);
 void infoCommand(redisClient *c);
 void mgetCommand(redisClient *c);
 void monitorCommand(redisClient *c);


@@ -872,6 +964,7 @@ void zaddCommand(redisClient *c);
 void zincrbyCommand(redisClient *c);
 void zrangeCommand(redisClient *c);
 void zrangebyscoreCommand(redisClient *c);
 void zincrbyCommand(redisClient *c);
 void zrangeCommand(redisClient *c);
 void zrangebyscoreCommand(redisClient *c);

+void zrevrangebyscoreCommand(redisClient *c);

 void zcountCommand(redisClient *c);
 void zrevrangeCommand(redisClient *c);
 void zcardCommand(redisClient *c);
 void zcountCommand(redisClient *c);
 void zrevrangeCommand(redisClient *c);
 void zcardCommand(redisClient *c);


@@ -883,8 +976,8 @@ void execCommand(redisClient *c);
 void discardCommand(redisClient *c);
 void blpopCommand(redisClient *c);
 void brpopCommand(redisClient *c);
 void discardCommand(redisClient *c);
 void blpopCommand(redisClient *c);
 void brpopCommand(redisClient *c);

+void brpoplpushCommand(redisClient *c);

 void appendCommand(redisClient *c);
 void appendCommand(redisClient *c);

-void substrCommand(redisClient *c);

 void strlenCommand(redisClient *c);
 void zrankCommand(redisClient *c);
 void zrevrankCommand(redisClient *c);
 void strlenCommand(redisClient *c);
 void zrankCommand(redisClient *c);
 void zrevrankCommand(redisClient *c);