redis-trib create, first version is know working

[redis.git] / src / redis.h

diff --git a/src/redis.h b/src/redis.h
index 08a377624d41f22a4f6ef9b3e0277e65634862f9..7e3a8c8f051bc5e5a2bc8bfd74ef522dabb543fb 100644 (file)
--- a/src/redis.h
+++ b/src/redis.h
@@ -17,6 +17,8 @@
 #include <errno.h>
 #include <inttypes.h>
 #include <pthread.h>
 #include <errno.h>
 #include <inttypes.h>
 #include <pthread.h>

+#include <syslog.h>
+#include <netinet/in.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 */


@@ -47,11 +49,7 @@
 #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_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 */

-
-/* 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
+#define REDIS_MAX_LOGMSG_LEN    1024 /* Default maximum length of syslog messages */

 
 /* 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% */


@@ -73,6 +71,10 @@
 #define REDIS_ZSET 3
 #define REDIS_HASH 4
 #define REDIS_VMPOINTER 8
 #define REDIS_ZSET 3
 #define REDIS_HASH 4
 #define REDIS_VMPOINTER 8

+/* Object types only used for persistence in .rdb files */
+#define REDIS_HASH_ZIPMAP 9
+#define REDIS_LIST_ZIPLIST 10
+#define REDIS_SET_INTSET 11

 
 /* Objects encoding. Some kind of objects like Strings and Hashes can be
  * internally represented in multiple ways. The 'encoding' field of the object
 
 /* Objects encoding. Some kind of objects like Strings and Hashes can be
  * internally represented in multiple ways. The 'encoding' field of the object


@@ -84,6 +86,7 @@
 #define REDIS_ENCODING_LINKEDLIST 4 /* Encoded as regular linked list */
 #define REDIS_ENCODING_ZIPLIST 5 /* Encoded as ziplist */
 #define REDIS_ENCODING_INTSET 6  /* Encoded as intset */
 #define REDIS_ENCODING_LINKEDLIST 4 /* Encoded as regular linked list */
 #define REDIS_ENCODING_ZIPLIST 5 /* Encoded as ziplist */
 #define REDIS_ENCODING_INTSET 6  /* Encoded as intset */

+#define REDIS_ENCODING_SKIPLIST 7  /* Encoded as skiplist */

 
 /* Object types only used for dumping to disk */
 #define REDIS_EXPIRETIME 253
 
 /* Object types only used for dumping to disk */
 #define REDIS_EXPIRETIME 253


@@ -117,23 +120,18 @@
 #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. */
+/* Scheduled IO opeations flags. */
+#define REDIS_IO_LOAD 1
+#define REDIS_IO_SAVE 2
+#define REDIS_IO_LOADINPROG 4
+#define REDIS_IO_SAVEINPROG 8
+
+/* Generic IO flags */
+#define REDIS_IO_ONLYLOADS 1
+#define REDIS_IO_ASAP 2
+

 #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 */


@@ -144,6 +142,8 @@
 #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. */
 #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. */

+#define REDIS_UNBLOCKED 256 /* This client was unblocked and is stored in
+                               server.unblocked_clients */

 
 /* Client request types */
 #define REDIS_REQ_INLINE 1
 
 /* Client request types */
 #define REDIS_REQ_INLINE 1


@@ -179,6 +179,7 @@
 #define REDIS_VERBOSE 1
 #define REDIS_NOTICE 2
 #define REDIS_WARNING 3
 #define REDIS_VERBOSE 1
 #define REDIS_NOTICE 2
 #define REDIS_WARNING 3

+#define REDIS_LOG_RAW (1<<10) /* Modifier to log without timestamp */

 
 /* Anti-warning macro... */
 #define REDIS_NOTUSED(V) ((void) V)
 
 /* Anti-warning macro... */
 #define REDIS_NOTUSED(V) ((void) V)


@@ -192,11 +193,11 @@
 #define APPENDFSYNC_EVERYSEC 2
 
 /* Zip structure related defaults */
 #define APPENDFSYNC_EVERYSEC 2
 
 /* 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_HASH_MAX_ZIPMAP_ENTRIES 512
+#define REDIS_HASH_MAX_ZIPMAP_VALUE 64
+#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
 
 /* Sets operations codes */
 #define REDIS_OP_UNION 0


@@ -211,6 +212,12 @@
 #define REDIS_MAXMEMORY_ALLKEYS_RANDOM 4
 #define REDIS_MAXMEMORY_NO_EVICTION 5
 
 #define REDIS_MAXMEMORY_ALLKEYS_RANDOM 4
 #define REDIS_MAXMEMORY_NO_EVICTION 5
 

+/* Diskstore background saving thread states */
+#define REDIS_BGSAVE_THREAD_UNACTIVE 0
+#define REDIS_BGSAVE_THREAD_ACTIVE 1
+#define REDIS_BGSAVE_THREAD_DONE_OK 2
+#define REDIS_BGSAVE_THREAD_DONE_ERR 3
+

 /* 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)


@@ -228,7 +235,7 @@ void _redisPanic(char *msg, char *file, int line);
 #define REDIS_LRU_CLOCK_RESOLUTION 10 /* LRU clock resolution in seconds */
 typedef struct redisObject {
     unsigned type:4;
 #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 */
+    unsigned notused:2;     /* Not used */

     unsigned encoding:4;
     unsigned lru:22;        /* lru time (relative to server.lruclock) */
     int refcount;
     unsigned encoding:4;
     unsigned lru:22;        /* lru time (relative to server.lruclock) */
     int refcount;


@@ -269,14 +276,15 @@ 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; \

 } while(0);
 
 typedef struct redisDb {
     dict *dict;                 /* The keyspace for this DB */
     dict *expires;              /* Timeout of keys with a timeout set */
     dict *blocking_keys;        /* Keys with clients waiting for data (BLPOP) */
 } while(0);
 
 typedef struct redisDb {
     dict *dict;                 /* The keyspace for this DB */
     dict *expires;              /* Timeout of keys with a timeout set */
     dict *blocking_keys;        /* Keys with clients waiting for data (BLPOP) */

-    dict *io_keys;              /* Keys with clients waiting for VM I/O */
+    dict *io_keys;              /* Keys with clients waiting for DS I/O */
+    dict *io_negcache;          /* Negative caching for disk store */
+    dict *io_queued;            /* Queued IO operations hash table */

     dict *watched_keys;         /* WATCHED keys for MULTI/EXEC CAS */
     int id;
 } redisDb;
     dict *watched_keys;         /* WATCHED keys for MULTI/EXEC CAS */
     int id;
 } redisDb;


@@ -293,6 +301,16 @@ 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 {


@@ -316,11 +334,7 @@ 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 */
     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 */


@@ -349,19 +363,140 @@ struct sharedObjectsStruct {
     *integers[REDIS_SHARED_INTEGERS];
 };
 
     *integers[REDIS_SHARED_INTEGERS];
 };
 

-/* Global server state structure */
+/*-----------------------------------------------------------------------------
+ * Redis cluster data structures
+ *----------------------------------------------------------------------------*/
+
+#define REDIS_CLUSTER_SLOTS 4096
+#define REDIS_CLUSTER_OK 0          /* Everything looks ok */
+#define REDIS_CLUSTER_FAIL 1        /* The cluster can't work */
+#define REDIS_CLUSTER_NEEDHELP 2    /* The cluster works, but needs some help */
+#define REDIS_CLUSTER_NAMELEN 40    /* sha1 hex length */
+#define REDIS_CLUSTER_PORT_INCR 10000 /* Cluster port = baseport + PORT_INCR */
+
+struct clusterNode;
+
+/* clusterLink encapsulates everything needed to talk with a remote node. */
+typedef struct clusterLink {
+    int fd;                     /* TCP socket file descriptor */
+    sds sndbuf;                 /* Packet send buffer */
+    sds rcvbuf;                 /* Packet reception buffer */
+    struct clusterNode *node;   /* Node related to this link if any, or NULL */
+} clusterLink;
+
+/* Node flags */
+#define REDIS_NODE_MASTER 1     /* The node is a master */
+#define REDIS_NODE_SLAVE 2      /* The node is a slave */
+#define REDIS_NODE_PFAIL 4      /* Failure? Need acknowledge */
+#define REDIS_NODE_FAIL 8       /* The node is believed to be malfunctioning */
+#define REDIS_NODE_MYSELF 16    /* This node is myself */
+#define REDIS_NODE_HANDSHAKE 32 /* We have still to exchange the first ping */
+#define REDIS_NODE_NOADDR   64  /* We don't know the address of this node */
+#define REDIS_NODE_MEET 128     /* Send a MEET message to this node */
+#define REDIS_NODE_NULL_NAME "\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000"
+
+struct clusterNode {
+    char name[REDIS_CLUSTER_NAMELEN]; /* Node name, hex string, sha1-size */
+    int flags;      /* REDIS_NODE_... */
+    unsigned char slots[REDIS_CLUSTER_SLOTS/8]; /* slots handled by this node */
+    int numslaves;  /* Number of slave nodes, if this is a master */
+    struct clusterNode **slaves; /* pointers to slave nodes */
+    struct clusterNode *slaveof; /* pointer to the master node */
+    time_t ping_sent;       /* Unix time we sent latest ping */
+    time_t pong_received;   /* Unix time we received the pong */
+    char *configdigest;         /* Configuration digest of this node */
+    time_t configdigest_ts;     /* Configuration digest timestamp */
+    char ip[16];                /* Latest known IP address of this node */
+    int port;                   /* Latest known port of this node */
+    clusterLink *link;          /* TCP/IP link with this node */
+};
+typedef struct clusterNode clusterNode;
+
+typedef struct {
+    char *configfile;
+    clusterNode *myself;  /* This node */
+    int state;            /* REDIS_CLUSTER_OK, REDIS_CLUSTER_FAIL, ... */
+    int node_timeout;
+    dict *nodes;          /* Hash table of name -> clusterNode structures */
+    clusterNode *migrating_slots_to[REDIS_CLUSTER_SLOTS];
+    clusterNode *importing_slots_from[REDIS_CLUSTER_SLOTS];
+    clusterNode *slots[REDIS_CLUSTER_SLOTS];
+} clusterState;
+
+/* Redis cluster messages header */
+
+/* Note that the PING, PONG and MEET messages are actually the same exact
+ * kind of packet. PONG is the reply to ping, in the extact format as a PING,
+ * while MEET is a special PING that forces the receiver to add the sender
+ * as a node (if it is not already in the list). */
+#define CLUSTERMSG_TYPE_PING 0          /* Ping */
+#define CLUSTERMSG_TYPE_PONG 1          /* Pong (reply to Ping) */
+#define CLUSTERMSG_TYPE_MEET 2          /* Meet "let's join" message */
+#define CLUSTERMSG_TYPE_FAIL 3          /* Mark node xxx as failing */
+
+/* Initially we don't know our "name", but we'll find it once we connect
+ * to the first node, using the getsockname() function. Then we'll use this
+ * address for all the next messages. */
+typedef struct {
+    char nodename[REDIS_CLUSTER_NAMELEN];
+    uint32_t ping_sent;
+    uint32_t pong_received;
+    char ip[16];    /* IP address last time it was seen */
+    uint16_t port;  /* port last time it was seen */
+    uint16_t flags;
+    uint32_t notused; /* for 64 bit alignment */
+} clusterMsgDataGossip;
+
+typedef struct {
+    char nodename[REDIS_CLUSTER_NAMELEN];
+} clusterMsgDataFail;
+
+union clusterMsgData {
+    /* PING, MEET and PONG */
+    struct {
+        /* Array of N clusterMsgDataGossip structures */
+        clusterMsgDataGossip gossip[1];
+    } ping;
+    /* FAIL */
+    struct {
+        clusterMsgDataFail about;
+    } fail;
+};
+
+typedef struct {
+    uint32_t totlen;    /* Total length of this message */
+    uint16_t type;      /* Message type */
+    uint16_t count;     /* Only used for some kind of messages. */
+    char sender[REDIS_CLUSTER_NAMELEN]; /* Name of the sender node */
+    unsigned char myslots[REDIS_CLUSTER_SLOTS/8];
+    char slaveof[REDIS_CLUSTER_NAMELEN];
+    char configdigest[32];
+    uint16_t port;      /* Sender TCP base port */
+    unsigned char state; /* Cluster state from the POV of the sender */
+    unsigned char notused[5]; /* Reserved for future use. For alignment. */
+    union clusterMsgData data;
+} clusterMsg;
+
+/*-----------------------------------------------------------------------------
+ * Global server state
+ *----------------------------------------------------------------------------*/
+

 struct redisServer {
 struct redisServer {

+    /* General */

     pthread_t mainthread;
     pthread_t mainthread;

+    redisDb *db;
+    dict *commands;             /* Command table hahs table */
+    aeEventLoop *el;
+    /* Networking */

     int port;
     char *bindaddr;
     char *unixsocket;
     int ipfd;
     int sofd;
     int port;
     char *bindaddr;
     char *unixsocket;
     int ipfd;
     int sofd;

-    redisDb *db;
-    long long dirty;            /* changes to DB from the last save */
-    long long dirty_before_bgsave; /* used to restore dirty on failed BGSAVE */
+    int cfd;

     list *clients;
     list *clients;

-    dict *commands;             /* Command table hahs table */
+    list *slaves, *monitors;
+    char neterr[ANET_ERR_LEN];

     /* RDB / AOF loading information */
     int loading;
     off_t loading_total_bytes;
     /* RDB / AOF loading information */
     int loading;
     off_t loading_total_bytes;


@@ -369,9 +504,6 @@ struct redisServer {
     time_t loading_start_time;
     /* Fast pointers to often looked up command */
     struct redisCommand *delCommand, *multiCommand;
     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 */
     time_t lastsave;                /* Unix time of last save succeeede */
     /* Fields used only for stats */
     int cronloops;              /* number of times the cron function run */
     time_t lastsave;                /* Unix time of last save succeeede */
     /* Fields used only for stats */


@@ -379,11 +511,11 @@ struct redisServer {
     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_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;
     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;

     int maxidletime;
     int dbnum;
     int daemonize;
     int maxidletime;
     int dbnum;
     int daemonize;


@@ -391,22 +523,32 @@ struct redisServer {
     int appendfsync;
     int no_appendfsync_on_rewrite;
     int shutdown_asap;
     int appendfsync;
     int no_appendfsync_on_rewrite;
     int shutdown_asap;

+    int activerehashing;
+    char *requirepass;
+    /* Persistence */
+    long long dirty;            /* changes to DB from the last save */
+    long long dirty_before_bgsave; /* used to restore dirty on failed BGSAVE */

     time_t lastfsync;
     int appendfd;
     int appendseldb;
     char *pidfile;
     pid_t bgsavechildpid;
     pid_t bgrewritechildpid;
     time_t lastfsync;
     int appendfd;
     int appendseldb;
     char *pidfile;
     pid_t bgsavechildpid;
     pid_t bgrewritechildpid;

+    int bgsavethread_state;
+    pthread_mutex_t bgsavethread_mutex;
+    pthread_t bgsavethread;

     sds bgrewritebuf; /* buffer taken by parent during oppend only rewrite */
     sds aofbuf;       /* AOF buffer, written before entering the event loop */
     struct saveparam *saveparams;
     int saveparamslen;
     sds bgrewritebuf; /* buffer taken by parent during oppend only rewrite */
     sds aofbuf;       /* AOF buffer, written before entering the event loop */
     struct saveparam *saveparams;
     int saveparamslen;

-    char *logfile;

     char *dbfilename;
     char *dbfilename;

-    char *appendfilename;
-    char *requirepass;

     int rdbcompression;
     int rdbcompression;

-    int activerehashing;
+    char *appendfilename;
+    /* Logging */
+    char *logfile;
+    int syslog_enabled;
+    char *syslog_ident;
+    int syslog_facility;

     /* Replication related */
     int isslave;
     /* Slave specific fields */
     /* Replication related */
     int isslave;
     /* Slave specific fields */


@@ -427,31 +569,26 @@ struct redisServer {
     int maxmemory_policy;
     int maxmemory_samples;
     /* Blocked clients */
     int maxmemory_policy;
     int maxmemory_samples;
     /* Blocked clients */

-    unsigned int blpop_blocked_clients;
-    unsigned int vm_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_io_queue;    /* IO operations queue */
+    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


@@ -462,7 +599,7 @@ 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 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 */


@@ -483,6 +620,8 @@ struct redisServer {
     /* Misc */
     unsigned lruclock:22;        /* clock incrementing every minute, for LRU */
     unsigned lruclock_padding:10;
     /* Misc */
     unsigned lruclock:22;        /* clock incrementing every minute, for LRU */
     unsigned lruclock_padding:10;

+    int cluster_enabled;
+    clusterState cluster;

 };
 
 typedef struct pubsubPattern {
 };
 
 typedef struct pubsubPattern {


@@ -491,20 +630,20 @@ typedef struct pubsubPattern {
 } pubsubPattern;
 
 typedef void redisCommandProc(redisClient *c);
 } pubsubPattern;
 
 typedef void redisCommandProc(redisClient *c);

-typedef void redisVmPreloadProc(redisClient *c, struct redisCommand *cmd, int argc, robj **argv);
+typedef int *redisGetKeysProc(struct redisCommand *cmd, robj **argv, int argc, int *numkeys, int flags);

 struct redisCommand {
     char *name;
     redisCommandProc *proc;
     int arity;
     int flags;
 struct redisCommand {
     char *name;
     redisCommandProc *proc;
     int arity;
     int flags;

-    /* Use a function to determine which keys need to be loaded
-     * in the background prior to executing this command. Takes precedence
-     * over vm_firstkey and others, ignored when NULL */
-    redisVmPreloadProc *vm_preload_proc;
+    /* Use a function to determine keys arguments in a command line.
+     * Used both for diskstore preloading and Redis Cluster. */
+    redisGetKeysProc *getkeys_proc;

     /* What keys should be loaded in background when calling this command? */
     /* What keys should be loaded in background when calling this command? */

-    int vm_firstkey; /* The first argument that's a key (0 = no keys) */
-    int vm_lastkey;  /* THe last argument that's a key */
-    int vm_keystep;  /* The step between first and last key */
+    int firstkey; /* The first argument that's a key (0 = no keys) */
+    int lastkey;  /* THe last argument that's a key */
+    int keystep;  /* The step between first and last key */
+    long long microseconds, calls;

 };
 
 struct redisFunctionSym {
 };
 
 struct redisFunctionSym {


@@ -547,25 +686,27 @@ typedef struct zset {
     zskiplist *zsl;
 } zset;
 
     zskiplist *zsl;
 } 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 */
+/* DIsk store threaded I/O request message */
+#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;
 

+/* IO operations scheduled -- check dscache.c for more info */
+typedef struct ioop {
+    int type;
+    redisDb *db;
+    robj *key;
+    time_t ctime; /* This is the creation time of the entry. */
+} ioop;
+

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


@@ -615,6 +756,7 @@ extern struct redisServer server;
 extern struct sharedObjectsStruct shared;
 extern dictType setDictType;
 extern dictType zsetDictType;
 extern struct sharedObjectsStruct shared;
 extern dictType setDictType;
 extern dictType zsetDictType;

+extern dictType clusterNodesDictType;

 extern double R_Zero, R_PosInf, R_NegInf, R_Nan;
 dictType hashDictType;
 
 extern double R_Zero, R_PosInf, R_NegInf, R_Nan;
 dictType hashDictType;
 


@@ -622,13 +764,15 @@ dictType hashDictType;
  * Functions prototypes
  *----------------------------------------------------------------------------*/
 
  * Functions prototypes
  *----------------------------------------------------------------------------*/
 

+/* Utils */
+long long ustime(void);
+

 /* networking.c -- Networking and Client related operations */
 redisClient *createClient(int fd);
 void closeTimedoutClients(void);
 void freeClient(redisClient *c);
 void resetClient(redisClient *c);
 void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask);
 /* networking.c -- Networking and Client related operations */
 redisClient *createClient(int fd);
 void closeTimedoutClients(void);
 void freeClient(redisClient *c);
 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 *addDeferredMultiBulkLength(redisClient *c);
 void setDeferredMultiBulkLength(redisClient *c, void *node, long length);
 void addReply(redisClient *c, robj *obj);
 void *addDeferredMultiBulkLength(redisClient *c);
 void setDeferredMultiBulkLength(redisClient *c, void *node, long length);


@@ -650,6 +794,8 @@ void addReplyDouble(redisClient *c, double d);
 void addReplyLongLong(redisClient *c, long long ll);
 void addReplyMultiBulkLen(redisClient *c, long length);
 void *dupClientReplyValue(void *o);
 void addReplyLongLong(redisClient *c, long long ll);
 void addReplyMultiBulkLen(redisClient *c, long length);
 void *dupClientReplyValue(void *o);

+void getClientsMaxBuffers(unsigned long *longest_output_list,
+                          unsigned long *biggest_input_buffer);

 
 #ifdef __GNUC__
 void addReplyErrorFormat(redisClient *c, const char *fmt, ...)
 
 #ifdef __GNUC__
 void addReplyErrorFormat(redisClient *c, const char *fmt, ...)


@@ -725,6 +871,7 @@ 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);
 int fwriteBulkDouble(FILE *fp, double d);
 int fwriteBulkLongLong(FILE *fp, long long l);
 int fwriteBulkObject(FILE *fp, robj *obj);

+int fwriteBulkCount(FILE *fp, char prefix, int count);

 
 /* Replication */
 void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc);
 
 /* Replication */
 void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc);


@@ -747,7 +894,13 @@ int rdbSaveObject(FILE *fp, robj *o);
 off_t rdbSavedObjectLen(robj *o);
 off_t rdbSavedObjectPages(robj *o);
 robj *rdbLoadObject(int type, FILE *fp);
 off_t rdbSavedObjectLen(robj *o);
 off_t rdbSavedObjectPages(robj *o);
 robj *rdbLoadObject(int type, FILE *fp);

-void backgroundSaveDoneHandler(int statloc);
+void backgroundSaveDoneHandler(int exitcode, int bysignal);
+int rdbSaveKeyValuePair(FILE *fp, robj *key, robj *val, time_t expireitme, time_t now);
+int rdbLoadType(FILE *fp);
+time_t rdbLoadTime(FILE *fp);
+robj *rdbLoadStringObject(FILE *fp);
+int rdbSaveType(FILE *fp, unsigned char type);
+int rdbSaveLen(FILE *fp, uint32_t len);

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


@@ -757,7 +910,7 @@ int rewriteAppendOnlyFileBackground(void);
 int loadAppendOnlyFile(char *filename);
 void stopAppendOnly(void);
 int startAppendOnly(void);
 int loadAppendOnlyFile(char *filename);
 void stopAppendOnly(void);
 int startAppendOnly(void);

-void backgroundRewriteDoneHandler(int statloc);
+void backgroundRewriteDoneHandler(int exitcode, int bysignal);

 
 /* Sorted sets data type */
 zskiplist *zslCreate(void);
 
 /* Sorted sets data type */
 zskiplist *zslCreate(void);


@@ -767,7 +920,7 @@ zskiplistNode *zslInsert(zskiplist *zsl, double score, robj *obj);
 /* Core functions */
 void freeMemoryIfNeeded(void);
 int processCommand(redisClient *c);
 /* Core functions */
 void freeMemoryIfNeeded(void);
 int processCommand(redisClient *c);

-void setupSigSegvAction(void);
+void setupSignalHandlers(void);

 struct redisCommand *lookupCommand(sds name);
 struct redisCommand *lookupCommandByCString(char *s);
 void call(redisClient *c, struct redisCommand *cmd);
 struct redisCommand *lookupCommand(sds name);
 struct redisCommand *lookupCommandByCString(char *s);
 void call(redisClient *c, struct redisCommand *cmd);


@@ -778,33 +931,41 @@ void updateDictResizePolicy(void);
 int htNeedsResize(dict *dict);
 void oom(const char *msg);
 void populateCommandTable(void);
 int htNeedsResize(dict *dict);
 void oom(const char *msg);
 void populateCommandTable(void);

-
-/* 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 resetCommandTableStats(void);
+
+/* Disk store */
+int dsOpen(void);
+int dsClose(void);
+int dsSet(redisDb *db, robj *key, robj *val, time_t expire);
+robj *dsGet(redisDb *db, robj *key, time_t *expire);
+int dsDel(redisDb *db, robj *key);
+int dsExists(redisDb *db, robj *key);
+void dsFlushDb(int dbid);
+int dsRdbSaveBackground(char *filename);
+int dsRdbSave(char *filename);
+
+/* 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 zunionInterBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv);
-void execBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv);
+void processAllPendingIOJobs(void);

 int blockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd);
 int dontWaitForSwappedKey(redisClient *c, robj *key);
 void handleClientsBlockedOnSwappedKey(redisDb *db, robj *key);
 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 cacheScheduleIOAddFlag(redisDb *db, robj *key, long flag);
+void cacheScheduleIODelFlag(redisDb *db, robj *key, long flag);
+int cacheScheduleIOGetFlags(redisDb *db, robj *key);
+void cacheScheduleIO(redisDb *db, robj *key, int type);
+void cacheCron(void);
+int cacheKeyMayExist(redisDb *db, robj *key);
+void cacheSetKeyMayExist(redisDb *db, robj *key);
+void cacheSetKeyDoesNotExist(redisDb *db, robj *key);
+void cacheForcePointInTime(void);

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


@@ -832,7 +993,8 @@ 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 */


@@ -874,6 +1036,26 @@ 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);
+
+/* API to get key arguments from commands */
+#define REDIS_GETKEYS_ALL 0
+#define REDIS_GETKEYS_PRELOAD 1
+int *getKeysFromCommand(struct redisCommand *cmd, robj **argv, int argc, int *numkeys, int flags);
+void getKeysFreeResult(int *result);
+int *noPreloadGetKeys(struct redisCommand *cmd,robj **argv, int argc, int *numkeys, int flags);
+int *renameGetKeys(struct redisCommand *cmd,robj **argv, int argc, int *numkeys, int flags);
+int *zunionInterGetKeys(struct redisCommand *cmd,robj **argv, int argc, int *numkeys, int flags);
+
+/* Cluster */
+void clusterInit(void);
+unsigned short crc16(const char *buf, int len);
+unsigned int keyHashSlot(char *key, int keylen);
+clusterNode *createClusterNode(char *nodename, int flags);
+int clusterAddNode(clusterNode *node);
+void clusterCron(void);
+clusterNode *getNodeByQuery(redisClient *c, struct redisCommand *cmd, robj **argv, int argc, int *hashslot);

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


@@ -889,6 +1071,10 @@ 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);


@@ -936,7 +1122,7 @@ 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);


@@ -965,8 +1151,8 @@ 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);


@@ -993,6 +1179,11 @@ void punsubscribeCommand(redisClient *c);
 void publishCommand(redisClient *c);
 void watchCommand(redisClient *c);
 void unwatchCommand(redisClient *c);
 void publishCommand(redisClient *c);
 void watchCommand(redisClient *c);
 void unwatchCommand(redisClient *c);

+void clusterCommand(redisClient *c);
+void restoreCommand(redisClient *c);
+void migrateCommand(redisClient *c);
+void dumpCommand(redisClient *c);
+void objectCommand(redisClient *c);

 
 #if defined(__GNUC__)
 void *calloc(size_t count, size_t size) __attribute__ ((deprecated));
 
 #if defined(__GNUC__)
 void *calloc(size_t count, size_t size) __attribute__ ((deprecated));