* POSSIBILITY OF SUCH DAMAGE.
*/
-#define REDIS_VERSION "1.3.10"
+#define REDIS_VERSION "2.1.1"
#include "fmacros.h"
#include "config.h"
#include <sys/resource.h>
#include <sys/uio.h>
#include <limits.h>
+#include <float.h>
#include <math.h>
#include <pthread.h>
#include "zmalloc.h" /* total memory usage aware version of malloc/free */
#include "lzf.h" /* LZF compression library */
#include "pqsort.h" /* Partial qsort for SORT+LIMIT */
-#include "zipmap.h"
+#include "zipmap.h" /* Compact dictionary-alike data structure */
+#include "ziplist.h" /* Compact list data structure */
+#include "sha1.h" /* SHA1 is used for DEBUG DIGEST */
/* Error codes */
#define REDIS_OK 0
#define REDIS_SET 2
#define REDIS_ZSET 3
#define REDIS_HASH 4
+#define REDIS_VMPOINTER 8
/* Objects encoding. Some kind of objects like Strings and Hashes can be
* internally represented in multiple ways. The 'encoding' field of the object
* is set to one of this fields for this object. */
-#define REDIS_ENCODING_RAW 0 /* Raw representation */
-#define REDIS_ENCODING_INT 1 /* Encoded as integer */
-#define REDIS_ENCODING_ZIPMAP 2 /* Encoded as zipmap */
-#define REDIS_ENCODING_HT 3 /* Encoded as an hash table */
+#define REDIS_ENCODING_RAW 0 /* Raw representation */
+#define REDIS_ENCODING_INT 1 /* Encoded as integer */
+#define REDIS_ENCODING_HT 2 /* Encoded as hash table */
+#define REDIS_ENCODING_ZIPMAP 3 /* Encoded as zipmap */
+#define REDIS_ENCODING_LINKEDLIST 4 /* Encoded as regular linked list */
+#define REDIS_ENCODING_ZIPLIST 5 /* Encoded as ziplist */
static char* strencoding[] = {
- "raw", "int", "zipmap", "hashtable"
+ "raw", "int", "hashtable", "zipmap", "linkedlist", "ziplist"
};
/* Object types only used for dumping to disk */
#define REDIS_MULTI 8 /* This client is in a MULTI context */
#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. */
/* Slave replication state - slave side */
#define REDIS_REPL_NONE 0 /* No active replication */
#define APPENDFSYNC_ALWAYS 1
#define APPENDFSYNC_EVERYSEC 2
-/* Hashes related defaults */
+/* 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
/* We can print the stacktrace, so our assert is defined this way: */
#define redisAssert(_e) ((_e)?(void)0 : (_redisAssert(#_e,__FILE__,__LINE__),_exit(1)))
/* A redis object, that is a type able to hold a string / list / set */
-/* The VM object structure */
-struct redisObjectVM {
- off_t page; /* the page at witch the object is stored on disk */
- off_t usedpages; /* number of pages used on disk */
- time_t atime; /* Last access time */
-} vm;
-
/* The actual Redis Object */
typedef struct redisObject {
- void *ptr;
- unsigned char type;
- unsigned char encoding;
- unsigned char storage; /* If this object is a key, where is the value?
- * REDIS_VM_MEMORY, REDIS_VM_SWAPPED, ... */
- unsigned char vtype; /* If this object is a key, and value is swapped out,
- * this is the type of the swapped out object. */
+ unsigned type:4;
+ unsigned storage:2; /* REDIS_VM_MEMORY or REDIS_VM_SWAPPING */
+ unsigned encoding:4;
+ unsigned lru:22; /* lru time (relative to server.lruclock) */
int refcount;
- /* VM fields, this are only allocated if VM is active, otherwise the
+ void *ptr;
+ /* VM fields are only allocated if VM is active, otherwise the
* object allocation function will just allocate
* sizeof(redisObjct) minus sizeof(redisObjectVM), so using
* Redis without VM active will not have any overhead. */
- struct redisObjectVM vm;
} robj;
+/* The VM pointer structure - identifies an object in the swap file.
+ *
+ * This object is stored in place of the value
+ * object in the main key->value hash table representing a database.
+ * Note that the first fields (type, storage) are the same as the redisObject
+ * structure so that vmPointer strucuters can be accessed even when casted
+ * as redisObject structures.
+ *
+ * This is useful as we don't know if a value object is or not on disk, but we
+ * are always able to read obj->storage to check this. For vmPointer
+ * structures "type" is set to REDIS_VMPOINTER (even if without this field
+ * is still possible to check the kind of object from the value of 'storage').*/
+typedef struct vmPointer {
+ unsigned type:4;
+ unsigned storage:2; /* REDIS_VM_SWAPPED or REDIS_VM_LOADING */
+ unsigned notused:26;
+ unsigned int vtype; /* type of the object stored in the swap file */
+ off_t page; /* the page at witch the object is stored on disk */
+ off_t usedpages; /* number of pages used on disk */
+} vmpointer;
+
/* Macro used to initalize a Redis object allocated on the stack.
* Note that this macro is taken near the structure definition to make sure
* we'll update it when the structure is changed, to avoid bugs like
_var.type = REDIS_STRING; \
_var.encoding = REDIS_ENCODING_RAW; \
_var.ptr = _ptr; \
- if (server.vm_enabled) _var.storage = REDIS_VM_MEMORY; \
+ _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 *blockingkeys; /* Keys with clients waiting for data (BLPOP) */
+ dict *blocking_keys; /* Keys with clients waiting for data (BLPOP) */
dict *io_keys; /* Keys with clients waiting for VM I/O */
+ dict *watched_keys; /* WATCHED keys for MULTI/EXEC CAS */
int id;
} redisDb;
long repldboff; /* replication DB file offset */
off_t repldbsize; /* replication DB file size */
multiState mstate; /* MULTI/EXEC state */
- robj **blockingkeys; /* The key we are waiting to terminate a blocking
+ robj **blocking_keys; /* The key we are waiting to terminate a blocking
* operation such as BLPOP. Otherwise NULL. */
- int blockingkeysnum; /* Number of blocking keys */
+ 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. */
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) */
} redisClient;
int daemonize;
int appendonly;
int appendfsync;
+ int no_appendfsync_on_rewrite;
+ int shutdown_asap;
time_t lastfsync;
int appendfd;
int appendseldb;
pid_t bgsavechildpid;
pid_t bgrewritechildpid;
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;
off_t vm_page_size;
off_t vm_pages;
unsigned long long vm_max_memory;
- /* Hashes config */
+ /* 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;
/* Virtual memory state */
FILE *vm_fp;
int vm_fd;
list *pubsub_patterns; /* A list of pubsub_patterns */
/* Misc */
FILE *devnull;
+ unsigned lruclock:22; /* clock incrementing every minute, for LRU */
+ unsigned lruclock_padding:10;
};
typedef struct pubsubPattern {
} pubsubPattern;
typedef void redisCommandProc(redisClient *c);
+typedef void redisVmPreloadProc(redisClient *c, struct redisCommand *cmd, int argc, robj **argv);
struct redisCommand {
char *name;
redisCommandProc *proc;
/* 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 */
- redisCommandProc *vm_preload_proc;
+ redisVmPreloadProc *vm_preload_proc;
/* 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 */
#define REDIS_SHARED_INTEGERS 10000
struct sharedObjectsStruct {
- robj *crlf, *ok, *err, *emptybulk, *czero, *cone, *pong, *space,
+ robj *crlf, *ok, *err, *emptybulk, *czero, *cone, *cnegone, *pong, *space,
*colon, *nullbulk, *nullmultibulk, *queued,
*emptymultibulk, *wrongtypeerr, *nokeyerr, *syntaxerr, *sameobjecterr,
*outofrangeerr, *plus,
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 */
} iojob;
/*================================ Prototypes =============================== */
+char *redisGitSHA1(void);
+char *redisGitDirty(void);
static void freeStringObject(robj *o);
static void freeListObject(robj *o);
static robj *dupStringObject(robj *o);
static void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc);
static void replicationFeedMonitors(list *monitors, int dictid, robj **argv, int argc);
+static void flushAppendOnlyFile(void);
static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc);
static int syncWithMaster(void);
static robj *tryObjectEncoding(robj *o);
static int removeExpire(redisDb *db, robj *key);
static int expireIfNeeded(redisDb *db, robj *key);
static int deleteIfVolatile(redisDb *db, robj *key);
-static int deleteIfSwapped(redisDb *db, robj *key);
-static int deleteKey(redisDb *db, robj *key);
+static int dbDelete(redisDb *db, robj *key);
static time_t getExpire(redisDb *db, robj *key);
static int setExpire(redisDb *db, robj *key, time_t when);
static void updateSlavesWaitingBgsave(int bgsaveerr);
static int handleClientsWaitingListPush(redisClient *c, robj *key, robj *ele);
static void vmInit(void);
static void vmMarkPagesFree(off_t page, off_t count);
-static robj *vmLoadObject(robj *key);
-static robj *vmPreviewObject(robj *key);
+static robj *vmLoadObject(robj *o);
+static robj *vmPreviewObject(robj *o);
static int vmSwapOneObjectBlocking(void);
static int vmSwapOneObjectThreaded(void);
static int vmCanSwapOut(void);
static void waitEmptyIOJobsQueue(void);
static void vmReopenSwapFile(void);
static int vmFreePage(off_t page);
-static void zunionInterBlockClientOnSwappedKeys(redisClient *c);
-static int blockClientOnSwappedKeys(struct redisCommand *cmd, redisClient *c);
+static void zunionInterBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv);
+static void execBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv);
+static int blockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd);
static int dontWaitForSwappedKey(redisClient *c, robj *key);
static void handleClientsBlockedOnSwappedKey(redisDb *db, robj *key);
static void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask);
static void call(redisClient *c, struct redisCommand *cmd);
static void resetClient(redisClient *c);
static void convertToRealHash(robj *o);
+static void listTypeConvert(robj *o, int enc);
static int pubsubUnsubscribeAllChannels(redisClient *c, int notify);
static int pubsubUnsubscribeAllPatterns(redisClient *c, int notify);
static void freePubsubPattern(void *p);
static int listMatchPubsubPattern(void *a, void *b);
static int compareStringObjects(robj *a, robj *b);
+static int equalStringObjects(robj *a, robj *b);
static void usage();
static int rewriteAppendOnlyFileBackground(void);
+static vmpointer *vmSwapObjectBlocking(robj *val);
+static int prepareForShutdown();
+static void touchWatchedKey(redisDb *db, robj *key);
+static void touchWatchedKeysOnFlush(int dbid);
+static void unwatchAllKeys(redisClient *c);
static void authCommand(redisClient *c);
static void pingCommand(redisClient *c);
static void renamenxCommand(redisClient *c);
static void lpushCommand(redisClient *c);
static void rpushCommand(redisClient *c);
+static void lpushxCommand(redisClient *c);
+static void rpushxCommand(redisClient *c);
+static void linsertCommand(redisClient *c);
static void lpopCommand(redisClient *c);
static void rpopCommand(redisClient *c);
static void llenCommand(redisClient *c);
static void hdelCommand(redisClient *c);
static void hlenCommand(redisClient *c);
static void zremrangebyrankCommand(redisClient *c);
-static void zunionCommand(redisClient *c);
-static void zinterCommand(redisClient *c);
+static void zunionstoreCommand(redisClient *c);
+static void zinterstoreCommand(redisClient *c);
static void hkeysCommand(redisClient *c);
static void hvalsCommand(redisClient *c);
static void hgetallCommand(redisClient *c);
static void psubscribeCommand(redisClient *c);
static void punsubscribeCommand(redisClient *c);
static void publishCommand(redisClient *c);
+static void watchCommand(redisClient *c);
+static void unwatchCommand(redisClient *c);
/*================================= Globals ================================= */
/* Global vars */
static struct redisServer server; /* server global state */
-static struct redisCommand cmdTable[] = {
+static struct redisCommand *commandTable;
+static struct redisCommand readonlyCommandTable[] = {
{"get",getCommand,2,REDIS_CMD_INLINE,NULL,1,1,1},
{"set",setCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,0,0,0},
{"setnx",setnxCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,0,0,0},
{"mget",mgetCommand,-2,REDIS_CMD_INLINE,NULL,1,-1,1},
{"rpush",rpushCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1},
{"lpush",lpushCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1},
+ {"rpushx",rpushxCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1},
+ {"lpushx",lpushxCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1},
+ {"linsert",linsertCommand,5,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1},
{"rpop",rpopCommand,2,REDIS_CMD_INLINE,NULL,1,1,1},
{"lpop",lpopCommand,2,REDIS_CMD_INLINE,NULL,1,1,1},
{"brpop",brpopCommand,-3,REDIS_CMD_INLINE,NULL,1,1,1},
{"zrem",zremCommand,3,REDIS_CMD_BULK,NULL,1,1,1},
{"zremrangebyscore",zremrangebyscoreCommand,4,REDIS_CMD_INLINE,NULL,1,1,1},
{"zremrangebyrank",zremrangebyrankCommand,4,REDIS_CMD_INLINE,NULL,1,1,1},
- {"zunion",zunionCommand,-4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0},
- {"zinter",zinterCommand,-4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0},
+ {"zunionstore",zunionstoreCommand,-4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0},
+ {"zinterstore",zinterstoreCommand,-4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0},
{"zrange",zrangeCommand,-4,REDIS_CMD_INLINE,NULL,1,1,1},
{"zrangebyscore",zrangebyscoreCommand,-4,REDIS_CMD_INLINE,NULL,1,1,1},
{"zcount",zcountCommand,4,REDIS_CMD_INLINE,NULL,1,1,1},
{"lastsave",lastsaveCommand,1,REDIS_CMD_INLINE,NULL,0,0,0},
{"type",typeCommand,2,REDIS_CMD_INLINE,NULL,1,1,1},
{"multi",multiCommand,1,REDIS_CMD_INLINE,NULL,0,0,0},
- {"exec",execCommand,1,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,NULL,0,0,0},
+ {"exec",execCommand,1,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,execBlockClientOnSwappedKeys,0,0,0},
{"discard",discardCommand,1,REDIS_CMD_INLINE,NULL,0,0,0},
{"sync",syncCommand,1,REDIS_CMD_INLINE,NULL,0,0,0},
{"flushdb",flushdbCommand,1,REDIS_CMD_INLINE,NULL,0,0,0},
{"psubscribe",psubscribeCommand,-2,REDIS_CMD_INLINE,NULL,0,0,0},
{"punsubscribe",punsubscribeCommand,-1,REDIS_CMD_INLINE,NULL,0,0,0},
{"publish",publishCommand,3,REDIS_CMD_BULK|REDIS_CMD_FORCE_REPLICATION,NULL,0,0,0},
- {NULL,NULL,0,0,NULL,0,0,0}
+ {"watch",watchCommand,-2,REDIS_CMD_INLINE,NULL,0,0,0},
+ {"unwatch",unwatchCommand,1,REDIS_CMD_INLINE,NULL,0,0,0}
};
/*============================ Utility functions ============================ */
return val*mul;
}
+/* Convert a long long into a string. Returns the number of
+ * characters needed to represent the number, that can be shorter if passed
+ * buffer length is not enough to store the whole number. */
+static int ll2string(char *s, size_t len, long long value) {
+ char buf[32], *p;
+ unsigned long long v;
+ size_t l;
+
+ if (len == 0) return 0;
+ v = (value < 0) ? -value : value;
+ p = buf+31; /* point to the last character */
+ do {
+ *p-- = '0'+(v%10);
+ v /= 10;
+ } while(v);
+ if (value < 0) *p-- = '-';
+ p++;
+ l = 32-(p-buf);
+ if (l+1 > len) l = len-1; /* Make sure it fits, including the nul term */
+ memcpy(s,p,l);
+ s[l] = '\0';
+ return l;
+}
+
static void redisLog(int level, const char *fmt, ...) {
va_list ap;
FILE *fp;
listRelease((list*)val);
}
-static int sdsDictKeyCompare(void *privdata, const void *key1,
+static int dictSdsKeyCompare(void *privdata, const void *key1,
const void *key2)
{
int l1,l2;
decrRefCount(val);
}
+static void dictSdsDestructor(void *privdata, void *val)
+{
+ DICT_NOTUSED(privdata);
+
+ sdsfree(val);
+}
+
static int dictObjKeyCompare(void *privdata, const void *key1,
const void *key2)
{
const robj *o1 = key1, *o2 = key2;
- return sdsDictKeyCompare(privdata,o1->ptr,o2->ptr);
+ return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
}
static unsigned int dictObjHash(const void *key) {
return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
}
+static unsigned int dictSdsHash(const void *key) {
+ return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
+}
+
static int dictEncObjKeyCompare(void *privdata, const void *key1,
const void *key2)
{
int cmp;
if (o1->encoding == REDIS_ENCODING_INT &&
- o2->encoding == REDIS_ENCODING_INT &&
- o1->ptr == o2->ptr) return 1;
+ o2->encoding == REDIS_ENCODING_INT)
+ return o1->ptr == o2->ptr;
o1 = getDecodedObject(o1);
o2 = getDecodedObject(o2);
- cmp = sdsDictKeyCompare(privdata,o1->ptr,o2->ptr);
+ cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
decrRefCount(o1);
decrRefCount(o2);
return cmp;
char buf[32];
int len;
- len = snprintf(buf,32,"%ld",(long)o->ptr);
+ len = ll2string(buf,32,(long)o->ptr);
return dictGenHashFunction((unsigned char*)buf, len);
} else {
unsigned int hash;
}
}
-/* Sets type and expires */
+/* Sets type */
static dictType setDictType = {
dictEncObjHash, /* hash function */
NULL, /* key dup */
dictVanillaFree /* val destructor of malloc(sizeof(double)) */
};
-/* Db->dict */
+/* Db->dict, keys are sds strings, vals are Redis objects. */
static dictType dbDictType = {
- dictObjHash, /* hash function */
+ dictSdsHash, /* hash function */
NULL, /* key dup */
NULL, /* val dup */
- dictObjKeyCompare, /* key compare */
- dictRedisObjectDestructor, /* key destructor */
+ dictSdsKeyCompare, /* key compare */
+ dictSdsDestructor, /* key destructor */
dictRedisObjectDestructor /* val destructor */
};
/* Db->expires */
static dictType keyptrDictType = {
- dictObjHash, /* hash function */
+ dictSdsHash, /* hash function */
NULL, /* key dup */
NULL, /* val dup */
- dictObjKeyCompare, /* key compare */
- dictRedisObjectDestructor, /* key destructor */
+ dictSdsKeyCompare, /* key compare */
+ NULL, /* key destructor */
NULL /* val destructor */
};
/* If append only is actually enabled... */
close(server.appendfd);
server.appendfd = fd;
- fsync(fd);
+ if (server.appendfsync != APPENDFSYNC_NO) aof_fsync(fd);
server.appendseldb = -1; /* Make sure it will issue SELECT */
redisLog(REDIS_NOTICE,"The new append only file was selected for future appends.");
} else {
* in objects at every object access, and accuracy is not needed.
* To access a global var is faster than calling time(NULL) */
server.unixtime = time(NULL);
+ /* We have just 21 bits per object for LRU information.
+ * So we use an (eventually wrapping) LRU clock with minutes resolution.
+ *
+ * When we need to select what object to swap, we compute the minimum
+ * time distance between the current lruclock and the object last access
+ * lruclock info. Even if clocks will wrap on overflow, there is
+ * the interesting property that we are sure that at least
+ * ABS(A-B) minutes passed between current time and timestamp B.
+ *
+ * This is not precise but we don't need at all precision, but just
+ * something statistically reasonable.
+ */
+ server.lruclock = (time(NULL)/60)&((1<<21)-1);
+
+ /* We received a SIGTERM, shutting down here in a safe way, as it is
+ * not ok doing so inside the signal handler. */
+ if (server.shutdown_asap) {
+ if (prepareForShutdown() == REDIS_OK) exit(0);
+ redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
+ }
/* Show some info about non-empty databases */
for (j = 0; j < server.dbnum; j++) {
if ((de = dictGetRandomKey(db->expires)) == NULL) break;
t = (time_t) dictGetEntryVal(de);
if (now > t) {
- deleteKey(db,dictGetEntryKey(de));
+ sds key = dictGetEntryKey(de);
+ robj *keyobj = createStringObject(key,sdslen(key));
+
+ dbDelete(db,keyobj);
+ decrRefCount(keyobj);
expired++;
server.stat_expiredkeys++;
}
static void beforeSleep(struct aeEventLoop *eventLoop) {
REDIS_NOTUSED(eventLoop);
+ /* Awake clients that got all the swapped keys they requested */
if (server.vm_enabled && listLength(server.io_ready_clients)) {
listIter li;
listNode *ln;
processInputBuffer(c);
}
}
+ /* Write the AOF buffer on disk */
+ flushAppendOnlyFile();
}
static void createSharedObjects(void) {
shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
+ shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
server.glueoutputbuf = 1;
server.daemonize = 0;
server.appendonly = 0;
- server.appendfsync = APPENDFSYNC_ALWAYS;
+ server.appendfsync = APPENDFSYNC_EVERYSEC;
+ server.no_appendfsync_on_rewrite = 0;
server.lastfsync = time(NULL);
server.appendfd = -1;
server.appendseldb = -1; /* Make sure the first time will not match */
server.vm_blocked_clients = 0;
server.hash_max_zipmap_entries = REDIS_HASH_MAX_ZIPMAP_ENTRIES;
server.hash_max_zipmap_value = REDIS_HASH_MAX_ZIPMAP_VALUE;
+ server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
+ server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
+ server.shutdown_asap = 0;
resetServerSaveParams();
for (j = 0; j < server.dbnum; j++) {
server.db[j].dict = dictCreate(&dbDictType,NULL);
server.db[j].expires = dictCreate(&keyptrDictType,NULL);
- server.db[j].blockingkeys = dictCreate(&keylistDictType,NULL);
+ server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
+ server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
if (server.vm_enabled)
server.db[j].io_keys = dictCreate(&keylistDictType,NULL);
server.db[j].id = j;
server.bgsavechildpid = -1;
server.bgrewritechildpid = -1;
server.bgrewritebuf = sdsempty();
+ server.aofbuf = sdsempty();
server.lastsave = time(NULL);
server.dirty = 0;
server.stat_numcommands = 0;
if ((server.appendonly = yesnotoi(argv[1])) == -1) {
err = "argument must be 'yes' or 'no'"; goto loaderr;
}
+ } else if (!strcasecmp(argv[0],"appendfilename") && argc == 2) {
+ zfree(server.appendfilename);
+ server.appendfilename = zstrdup(argv[1]);
+ } else if (!strcasecmp(argv[0],"no-appendfsync-on-rewrite")
+ && argc == 2) {
+ if ((server.no_appendfsync_on_rewrite= yesnotoi(argv[1])) == -1) {
+ err = "argument must be 'yes' or 'no'"; goto loaderr;
+ }
} else if (!strcasecmp(argv[0],"appendfsync") && argc == 2) {
if (!strcasecmp(argv[1],"no")) {
server.appendfsync = APPENDFSYNC_NO;
server.hash_max_zipmap_entries = memtoll(argv[1], NULL);
} else if (!strcasecmp(argv[0],"hash-max-zipmap-value") && argc == 2){
server.hash_max_zipmap_value = memtoll(argv[1], NULL);
+ } else if (!strcasecmp(argv[0],"list-max-ziplist-entries") && argc == 2){
+ server.list_max_ziplist_entries = memtoll(argv[1], NULL);
+ } else if (!strcasecmp(argv[0],"list-max-ziplist-value") && argc == 2){
+ server.list_max_ziplist_value = memtoll(argv[1], NULL);
} else {
err = "Bad directive or wrong number of arguments"; goto loaderr;
}
if (c->flags & REDIS_BLOCKED)
unblockClientWaitingData(c);
+ /* UNWATCH all the keys */
+ unwatchAllKeys(c);
+ listRelease(c->watched_keys);
/* Unsubscribe from all the pubsub channels */
pubsubUnsubscribeAllChannels(c,0);
pubsubUnsubscribeAllPatterns(c,0);
ln = listSearchKey(server.clients,c);
redisAssert(ln != NULL);
listDelNode(server.clients,ln);
- /* Remove from the list of clients waiting for swapped keys */
+ /* Remove from the list of clients that are now ready to be restarted
+ * after waiting for swapped keys */
if (c->flags & REDIS_IO_WAIT && listLength(c->io_keys) == 0) {
ln = listSearchKey(server.io_ready_clients,c);
if (ln) {
server.vm_blocked_clients--;
}
}
+ /* Remove from the list of clients waiting for swapped keys */
while (server.vm_enabled && listLength(c->io_keys)) {
ln = listFirst(c->io_keys);
dontWaitForSwappedKey(c,ln->value);
}
}
+static int qsortRedisCommands(const void *r1, const void *r2) {
+ return strcasecmp(
+ ((struct redisCommand*)r1)->name,
+ ((struct redisCommand*)r2)->name);
+}
+
+static void sortCommandTable() {
+ /* Copy and sort the read-only version of the command table */
+ commandTable = (struct redisCommand*)malloc(sizeof(readonlyCommandTable));
+ memcpy(commandTable,readonlyCommandTable,sizeof(readonlyCommandTable));
+ qsort(commandTable,
+ sizeof(readonlyCommandTable)/sizeof(struct redisCommand),
+ sizeof(struct redisCommand),qsortRedisCommands);
+}
+
static struct redisCommand *lookupCommand(char *name) {
- int j = 0;
- while(cmdTable[j].name != NULL) {
- if (!strcasecmp(name,cmdTable[j].name)) return &cmdTable[j];
- j++;
- }
- return NULL;
+ struct redisCommand tmp = {name,NULL,0,0,NULL,0,0,0};
+ return bsearch(
+ &tmp,
+ commandTable,
+ sizeof(readonlyCommandTable)/sizeof(struct redisCommand),
+ sizeof(struct redisCommand),
+ qsortRedisCommands);
}
/* resetClient prepare the client to process the next command */
}
/* Exec the command */
- if (c->flags & REDIS_MULTI && cmd->proc != execCommand && cmd->proc != discardCommand) {
+ if (c->flags & REDIS_MULTI &&
+ cmd->proc != execCommand && cmd->proc != discardCommand &&
+ cmd->proc != multiCommand && cmd->proc != watchCommand)
+ {
queueMultiCommand(c,cmd);
addReply(c,shared.queued);
} else {
if (server.vm_enabled && server.vm_max_threads > 0 &&
- blockClientOnSwappedKeys(cmd,c)) return 1;
+ blockClientOnSwappedKeys(c,cmd)) return 1;
call(c,cmd);
}
}
static int listMatchObjects(void *a, void *b) {
- return compareStringObjects(a,b) == 0;
+ return equalStringObjects(a,b);
}
static redisClient *createClient(int fd) {
c->reply = listCreate();
listSetFreeMethod(c->reply,decrRefCount);
listSetDupMethod(c->reply,dupClientReplyValue);
- c->blockingkeys = NULL;
- c->blockingkeysnum = 0;
+ c->blocking_keys = NULL;
+ c->blocking_keys_num = 0;
c->io_keys = listCreate();
+ c->watched_keys = listCreate();
listSetFreeMethod(c->io_keys,decrRefCount);
c->pubsub_channels = dictCreate(&setDictType,NULL);
c->pubsub_patterns = listCreate();
(unsigned long) strlen(buf),buf));
}
-static void addReplyLong(redisClient *c, long l) {
- char buf[128];
- size_t len;
-
- if (l == 0) {
- addReply(c,shared.czero);
- return;
- } else if (l == 1) {
- addReply(c,shared.cone);
- return;
- }
- len = snprintf(buf,sizeof(buf),":%ld\r\n",l);
- addReplySds(c,sdsnewlen(buf,len));
-}
-
static void addReplyLongLong(redisClient *c, long long ll) {
char buf[128];
size_t len;
addReply(c,shared.cone);
return;
}
- len = snprintf(buf,sizeof(buf),":%lld\r\n",ll);
- addReplySds(c,sdsnewlen(buf,len));
+ buf[0] = ':';
+ len = ll2string(buf+1,sizeof(buf)-1,ll);
+ buf[len+1] = '\r';
+ buf[len+2] = '\n';
+ addReplySds(c,sdsnewlen(buf,len+3));
}
static void addReplyUlong(redisClient *c, unsigned long ul) {
}
static void addReplyBulkLen(redisClient *c, robj *obj) {
- size_t len;
+ size_t len, intlen;
+ char buf[128];
if (obj->encoding == REDIS_ENCODING_RAW) {
len = sdslen(obj->ptr);
len++;
}
}
- addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",(unsigned long)len));
+ buf[0] = '$';
+ intlen = ll2string(buf+1,sizeof(buf)-1,(long long)len);
+ buf[intlen+1] = '\r';
+ buf[intlen+2] = '\n';
+ addReplySds(c,sdsnewlen(buf,intlen+3));
}
static void addReplyBulk(redisClient *c, robj *obj) {
addReply(c,shared.crlf);
}
+static void addReplyBulkSds(redisClient *c, sds s) {
+ robj *o = createStringObject(s, sdslen(s));
+ addReplyBulk(c,o);
+ decrRefCount(o);
+}
+
/* In the CONFIG command we need to add vanilla C string as bulk replies */
static void addReplyBulkCString(redisClient *c, char *s) {
if (s == NULL) {
listDelNode(server.objfreelist,head);
if (server.vm_enabled) pthread_mutex_unlock(&server.obj_freelist_mutex);
} else {
- if (server.vm_enabled) {
+ if (server.vm_enabled)
pthread_mutex_unlock(&server.obj_freelist_mutex);
- o = zmalloc(sizeof(*o));
- } else {
- o = zmalloc(sizeof(*o)-sizeof(struct redisObjectVM));
- }
+ o = zmalloc(sizeof(*o));
}
o->type = type;
o->encoding = REDIS_ENCODING_RAW;
o->refcount = 1;
if (server.vm_enabled) {
/* Note that this code may run in the context of an I/O thread
- * and accessing to server.unixtime in theory is an error
+ * and accessing server.lruclock in theory is an error
* (no locks). But in practice this is safe, and even if we read
- * garbage Redis will not fail, as it's just a statistical info */
- o->vm.atime = server.unixtime;
+ * garbage Redis will not fail. */
+ o->lru = server.lruclock;
o->storage = REDIS_VM_MEMORY;
}
return o;
incrRefCount(shared.integers[value]);
o = shared.integers[value];
} else {
- o = createObject(REDIS_STRING, NULL);
if (value >= LONG_MIN && value <= LONG_MAX) {
+ o = createObject(REDIS_STRING, NULL);
o->encoding = REDIS_ENCODING_INT;
o->ptr = (void*)((long)value);
} else {
- o->ptr = sdscatprintf(sdsempty(),"%lld",value);
+ o = createObject(REDIS_STRING,sdsfromlonglong(value));
}
}
return o;
static robj *createListObject(void) {
list *l = listCreate();
-
+ robj *o = createObject(REDIS_LIST,l);
listSetFreeMethod(l,decrRefCount);
- return createObject(REDIS_LIST,l);
+ o->encoding = REDIS_ENCODING_LINKEDLIST;
+ return o;
+}
+
+static robj *createZiplistObject(void) {
+ unsigned char *zl = ziplistNew();
+ robj *o = createObject(REDIS_LIST,zl);
+ o->encoding = REDIS_ENCODING_ZIPLIST;
+ return o;
}
static robj *createSetObject(void) {
}
static void freeListObject(robj *o) {
- listRelease((list*) o->ptr);
+ switch (o->encoding) {
+ case REDIS_ENCODING_LINKEDLIST:
+ listRelease((list*) o->ptr);
+ break;
+ case REDIS_ENCODING_ZIPLIST:
+ zfree(o->ptr);
+ break;
+ default:
+ redisPanic("Unknown list encoding type");
+ }
}
static void freeSetObject(robj *o) {
static void decrRefCount(void *obj) {
robj *o = obj;
- if (o->refcount <= 0) redisPanic("decrRefCount against refcount <= 0");
- /* Object is a key of a swapped out value, or in the process of being
- * loaded. */
+ /* Object is a swapped out value, or in the process of being loaded. */
if (server.vm_enabled &&
(o->storage == REDIS_VM_SWAPPED || o->storage == REDIS_VM_LOADING))
{
- if (o->storage == REDIS_VM_LOADING) vmCancelThreadedIOJob(obj);
- redisAssert(o->type == REDIS_STRING);
- freeStringObject(o);
- vmMarkPagesFree(o->vm.page,o->vm.usedpages);
- pthread_mutex_lock(&server.obj_freelist_mutex);
- if (listLength(server.objfreelist) > REDIS_OBJFREELIST_MAX ||
- !listAddNodeHead(server.objfreelist,o))
- zfree(o);
- pthread_mutex_unlock(&server.obj_freelist_mutex);
+ vmpointer *vp = obj;
+ if (o->storage == REDIS_VM_LOADING) vmCancelThreadedIOJob(o);
+ vmMarkPagesFree(vp->page,vp->usedpages);
server.vm_stats_swapped_objects--;
+ zfree(vp);
return;
}
- /* Object is in memory, or in the process of being swapped out. */
+
+ if (o->refcount <= 0) redisPanic("decrRefCount against refcount <= 0");
+ /* Object is in memory, or in the process of being swapped out.
+ *
+ * If the object is being swapped out, abort the operation on
+ * decrRefCount even if the refcount does not drop to 0: the object
+ * is referenced at least two times, as value of the key AND as
+ * job->val in the iojob. So if we don't invalidate the iojob, when it is
+ * done but the relevant key was removed in the meantime, the
+ * complete jobs handler will not find the key about the job and the
+ * assert will fail. */
+ if (server.vm_enabled && o->storage == REDIS_VM_SWAPPING)
+ vmCancelThreadedIOJob(o);
if (--(o->refcount) == 0) {
- if (server.vm_enabled && o->storage == REDIS_VM_SWAPPING)
- vmCancelThreadedIOJob(obj);
switch(o->type) {
case REDIS_STRING: freeStringObject(o); break;
case REDIS_LIST: freeListObject(o); break;
}
}
-static robj *lookupKey(redisDb *db, robj *key) {
- dictEntry *de = dictFind(db->dict,key);
- if (de) {
- robj *key = dictGetEntryKey(de);
- robj *val = dictGetEntryVal(de);
-
- if (server.vm_enabled) {
- if (key->storage == REDIS_VM_MEMORY ||
- key->storage == REDIS_VM_SWAPPING)
- {
- /* If we were swapping the object out, stop it, this key
- * was requested. */
- if (key->storage == REDIS_VM_SWAPPING)
- vmCancelThreadedIOJob(key);
- /* Update the access time of the key for the aging algorithm. */
- key->vm.atime = server.unixtime;
- } else {
- int notify = (key->storage == REDIS_VM_LOADING);
-
- /* Our value was swapped on disk. Bring it at home. */
- redisAssert(val == NULL);
- val = vmLoadObject(key);
- dictGetEntryVal(de) = val;
-
- /* Clients blocked by the VM subsystem may be waiting for
- * this key... */
- if (notify) handleClientsBlockedOnSwappedKey(db,key);
- }
- }
- return val;
- } else {
- return NULL;
- }
-}
-
-static robj *lookupKeyRead(redisDb *db, robj *key) {
- expireIfNeeded(db,key);
- return lookupKey(db,key);
-}
-
-static robj *lookupKeyWrite(redisDb *db, robj *key) {
- deleteIfVolatile(db,key);
- return lookupKey(db,key);
-}
-
-static robj *lookupKeyReadOrReply(redisClient *c, robj *key, robj *reply) {
- robj *o = lookupKeyRead(c->db, key);
- if (!o) addReply(c,reply);
- return o;
-}
-
-static robj *lookupKeyWriteOrReply(redisClient *c, robj *key, robj *reply) {
- robj *o = lookupKeyWrite(c->db, key);
- if (!o) addReply(c,reply);
- return o;
-}
-
static int checkType(redisClient *c, robj *o, int type) {
if (o->type != type) {
addReply(c,shared.wrongtypeerr);
return 0;
}
-static int deleteKey(redisDb *db, robj *key) {
- int retval;
-
- /* We need to protect key from destruction: after the first dictDelete()
- * it may happen that 'key' is no longer valid if we don't increment
- * it's count. This may happen when we get the object reference directly
- * from the hash table with dictRandomKey() or dict iterators */
- incrRefCount(key);
- if (dictSize(db->expires)) dictDelete(db->expires,key);
- retval = dictDelete(db->dict,key);
- decrRefCount(key);
-
- return retval == DICT_OK;
-}
-
/* Check if the nul-terminated string 's' can be represented by a long
* (that is, is a number that fits into long without any other space or
* character before or after the digits).
value = strtol(s, &endptr, 10);
if (endptr[0] != '\0') return REDIS_ERR;
- slen = snprintf(buf,32,"%ld",value);
+ slen = ll2string(buf,32,value);
/* If the number converted back into a string is not identical
* then it's not possible to encode the string as integer */
if (o->type == REDIS_STRING && o->encoding == REDIS_ENCODING_INT) {
char buf[32];
- snprintf(buf,32,"%ld",(long)o->ptr);
+ ll2string(buf,32,(long)o->ptr);
dec = createStringObject(buf,strlen(buf));
return dec;
} else {
/* Compare two string objects via strcmp() or alike.
* Note that the objects may be integer-encoded. In such a case we
- * use snprintf() to get a string representation of the numbers on the stack
+ * use ll2string() to get a string representation of the numbers on the stack
* and compare the strings, it's much faster than calling getDecodedObject().
*
* Important note: if objects are not integer encoded, but binary-safe strings,
if (a == b) return 0;
if (a->encoding != REDIS_ENCODING_RAW) {
- snprintf(bufa,sizeof(bufa),"%ld",(long) a->ptr);
+ ll2string(bufa,sizeof(bufa),(long) a->ptr);
astr = bufa;
bothsds = 0;
} else {
astr = a->ptr;
}
if (b->encoding != REDIS_ENCODING_RAW) {
- snprintf(bufb,sizeof(bufb),"%ld",(long) b->ptr);
+ ll2string(bufb,sizeof(bufb),(long) b->ptr);
bstr = bufb;
bothsds = 0;
} else {
return bothsds ? sdscmp(astr,bstr) : strcmp(astr,bstr);
}
+/* Equal string objects return 1 if the two objects are the same from the
+ * point of view of a string comparison, otherwise 0 is returned. Note that
+ * this function is faster then checking for (compareStringObject(a,b) == 0)
+ * because it can perform some more optimization. */
+static int equalStringObjects(robj *a, robj *b) {
+ if (a->encoding != REDIS_ENCODING_RAW && b->encoding != REDIS_ENCODING_RAW){
+ return a->ptr == b->ptr;
+ } else {
+ return compareStringObjects(a,b) == 0;
+ }
+}
+
static size_t stringObjectLen(robj *o) {
redisAssert(o->type == REDIS_STRING);
if (o->encoding == REDIS_ENCODING_RAW) {
} else {
char buf[32];
- return snprintf(buf,32,"%ld",(long)o->ptr);
+ return ll2string(buf,32,(long)o->ptr);
}
}
return REDIS_OK;
}
+/* =========================== Keyspace access API ========================== */
+
+static robj *lookupKey(redisDb *db, robj *key) {
+ dictEntry *de = dictFind(db->dict,key->ptr);
+ if (de) {
+ robj *val = dictGetEntryVal(de);
+
+ if (server.vm_enabled) {
+ if (val->storage == REDIS_VM_MEMORY ||
+ val->storage == REDIS_VM_SWAPPING)
+ {
+ /* If we were swapping the object out, cancel the operation */
+ if (val->storage == REDIS_VM_SWAPPING)
+ vmCancelThreadedIOJob(val);
+ /* Update the access time for the aging algorithm. */
+ val->lru = server.lruclock;
+ } else {
+ int notify = (val->storage == REDIS_VM_LOADING);
+
+ /* Our value was swapped on disk. Bring it at home. */
+ redisAssert(val->type == REDIS_VMPOINTER);
+ val = vmLoadObject(val);
+ dictGetEntryVal(de) = val;
+
+ /* Clients blocked by the VM subsystem may be waiting for
+ * this key... */
+ if (notify) handleClientsBlockedOnSwappedKey(db,key);
+ }
+ }
+ return val;
+ } else {
+ return NULL;
+ }
+}
+
+static robj *lookupKeyRead(redisDb *db, robj *key) {
+ expireIfNeeded(db,key);
+ return lookupKey(db,key);
+}
+
+static robj *lookupKeyWrite(redisDb *db, robj *key) {
+ deleteIfVolatile(db,key);
+ touchWatchedKey(db,key);
+ return lookupKey(db,key);
+}
+
+static robj *lookupKeyReadOrReply(redisClient *c, robj *key, robj *reply) {
+ robj *o = lookupKeyRead(c->db, key);
+ if (!o) addReply(c,reply);
+ return o;
+}
+
+static robj *lookupKeyWriteOrReply(redisClient *c, robj *key, robj *reply) {
+ robj *o = lookupKeyWrite(c->db, key);
+ if (!o) addReply(c,reply);
+ return o;
+}
+
+/* Add the key to the DB. If the key already exists REDIS_ERR is returned,
+ * otherwise REDIS_OK is returned, and the caller should increment the
+ * refcount of 'val'. */
+static int dbAdd(redisDb *db, robj *key, robj *val) {
+ /* Perform a lookup before adding the key, as we need to copy the
+ * key value. */
+ if (dictFind(db->dict, key->ptr) != NULL) {
+ return REDIS_ERR;
+ } else {
+ sds copy = sdsdup(key->ptr);
+ dictAdd(db->dict, copy, val);
+ return REDIS_OK;
+ }
+}
+
+/* If the key does not exist, this is just like dbAdd(). Otherwise
+ * the value associated to the key is replaced with the new one.
+ *
+ * On update (key already existed) 0 is returned. Otherwise 1. */
+static int dbReplace(redisDb *db, robj *key, robj *val) {
+ if (dictFind(db->dict,key->ptr) == NULL) {
+ sds copy = sdsdup(key->ptr);
+ dictAdd(db->dict, copy, val);
+ return 1;
+ } else {
+ dictReplace(db->dict, key->ptr, val);
+ return 0;
+ }
+}
+
+static int dbExists(redisDb *db, robj *key) {
+ return dictFind(db->dict,key->ptr) != NULL;
+}
+
+/* Return a random key, in form of a Redis object.
+ * If there are no keys, NULL is returned.
+ *
+ * The function makes sure to return keys not already expired. */
+static robj *dbRandomKey(redisDb *db) {
+ struct dictEntry *de;
+
+ while(1) {
+ sds key;
+ robj *keyobj;
+
+ de = dictGetRandomKey(db->dict);
+ if (de == NULL) return NULL;
+
+ key = dictGetEntryKey(de);
+ keyobj = createStringObject(key,sdslen(key));
+ if (dictFind(db->expires,key)) {
+ if (expireIfNeeded(db,keyobj)) {
+ decrRefCount(keyobj);
+ continue; /* search for another key. This expired. */
+ }
+ }
+ return keyobj;
+ }
+}
+
+/* Delete a key, value, and associated expiration entry if any, from the DB */
+static int dbDelete(redisDb *db, robj *key) {
+ /* Deleting an entry from the expires dict will not free the sds of
+ * the key, because it is shared with the main dictionary. */
+ if (dictSize(db->expires) > 0) dictDelete(db->expires,key->ptr);
+ return dictDelete(db->dict,key->ptr) == DICT_OK;
+}
+
/*============================ RDB saving/loading =========================== */
static int rdbSaveType(FILE *fp, unsigned char type) {
return 0;
}
-/* String objects in the form "2391" "-100" without any space and with a
- * range of values that can fit in an 8, 16 or 32 bit signed value can be
- * encoded as integers to save space */
-static int rdbTryIntegerEncoding(char *s, size_t len, unsigned char *enc) {
- long long value;
- char *endptr, buf[32];
-
- /* Check if it's possible to encode this value as a number */
- value = strtoll(s, &endptr, 10);
- if (endptr[0] != '\0') return 0;
- snprintf(buf,32,"%lld",value);
-
- /* If the number converted back into a string is not identical
- * then it's not possible to encode the string as integer */
- if (strlen(buf) != len || memcmp(buf,s,len)) return 0;
-
+/* Encode 'value' as an integer if possible (if integer will fit the
+ * supported range). If the function sucessful encoded the integer
+ * then the (up to 5 bytes) encoded representation is written in the
+ * string pointed by 'enc' and the length is returned. Otherwise
+ * 0 is returned. */
+static int rdbEncodeInteger(long long value, unsigned char *enc) {
/* Finally check if it fits in our ranges */
if (value >= -(1<<7) && value <= (1<<7)-1) {
enc[0] = (REDIS_RDB_ENCVAL<<6)|REDIS_RDB_ENC_INT8;
}
}
+/* String objects in the form "2391" "-100" without any space and with a
+ * range of values that can fit in an 8, 16 or 32 bit signed value can be
+ * encoded as integers to save space */
+static int rdbTryIntegerEncoding(char *s, size_t len, unsigned char *enc) {
+ long long value;
+ char *endptr, buf[32];
+
+ /* Check if it's possible to encode this value as a number */
+ value = strtoll(s, &endptr, 10);
+ if (endptr[0] != '\0') return 0;
+ ll2string(buf,32,value);
+
+ /* If the number converted back into a string is not identical
+ * then it's not possible to encode the string as integer */
+ if (strlen(buf) != len || memcmp(buf,s,len)) return 0;
+
+ return rdbEncodeInteger(value,enc);
+}
+
static int rdbSaveLzfStringObject(FILE *fp, unsigned char *s, size_t len) {
size_t comprlen, outlen;
unsigned char byte;
return 0;
}
+/* Save a long long value as either an encoded string or a string. */
+static int rdbSaveLongLongAsStringObject(FILE *fp, long long value) {
+ unsigned char buf[32];
+ int enclen = rdbEncodeInteger(value,buf);
+ if (enclen > 0) {
+ if (fwrite(buf,enclen,1,fp) == 0) return -1;
+ } else {
+ /* Encode as string */
+ enclen = ll2string((char*)buf,32,value);
+ redisAssert(enclen < 32);
+ if (rdbSaveLen(fp,enclen) == -1) return -1;
+ if (fwrite(buf,enclen,1,fp) == 0) return -1;
+ }
+ return 0;
+}
+
/* Like rdbSaveStringObjectRaw() but handle encoded objects */
static int rdbSaveStringObject(FILE *fp, robj *obj) {
- int retval;
-
- /* Avoid incr/decr ref count business when possible.
- * This plays well with copy-on-write given that we are probably
- * in a child process (BGSAVE). Also this makes sure key objects
- * of swapped objects are not incRefCount-ed (an assert does not allow
- * this in order to avoid bugs) */
- if (obj->encoding != REDIS_ENCODING_RAW) {
- obj = getDecodedObject(obj);
- retval = rdbSaveRawString(fp,obj->ptr,sdslen(obj->ptr));
- decrRefCount(obj);
+ /* Avoid to decode the object, then encode it again, if the
+ * object is alrady integer encoded. */
+ if (obj->encoding == REDIS_ENCODING_INT) {
+ return rdbSaveLongLongAsStringObject(fp,(long)obj->ptr);
} else {
- retval = rdbSaveRawString(fp,obj->ptr,sdslen(obj->ptr));
+ redisAssert(obj->encoding == REDIS_ENCODING_RAW);
+ return rdbSaveRawString(fp,obj->ptr,sdslen(obj->ptr));
}
- return retval;
}
/* Save a double value. Doubles are saved as strings prefixed by an unsigned
len = 1;
buf[0] = (val < 0) ? 255 : 254;
} else {
- snprintf((char*)buf+1,sizeof(buf)-1,"%.17g",val);
+#if (DBL_MANT_DIG >= 52) && (LLONG_MAX == 0x7fffffffffffffffLL)
+ /* Check if the float is in a safe range to be casted into a
+ * long long. We are assuming that long long is 64 bit here.
+ * Also we are assuming that there are no implementations around where
+ * double has precision < 52 bit.
+ *
+ * Under this assumptions we test if a double is inside an interval
+ * where casting to long long is safe. Then using two castings we
+ * make sure the decimal part is zero. If all this is true we use
+ * integer printing function that is much faster. */
+ double min = -4503599627370495; /* (2^52)-1 */
+ double max = 4503599627370496; /* -(2^52) */
+ if (val > min && val < max && val == ((double)((long long)val)))
+ ll2string((char*)buf+1,sizeof(buf),(long long)val);
+ else
+#endif
+ snprintf((char*)buf+1,sizeof(buf)-1,"%.17g",val);
buf[0] = strlen((char*)buf+1);
len = buf[0]+1;
}
if (rdbSaveStringObject(fp,o) == -1) return -1;
} else if (o->type == REDIS_LIST) {
/* Save a list value */
- list *list = o->ptr;
- listIter li;
- listNode *ln;
-
- if (rdbSaveLen(fp,listLength(list)) == -1) return -1;
- listRewind(list,&li);
- while((ln = listNext(&li))) {
- robj *eleobj = listNodeValue(ln);
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *p;
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vlong;
+
+ if (rdbSaveLen(fp,ziplistLen(o->ptr)) == -1) return -1;
+ p = ziplistIndex(o->ptr,0);
+ while(ziplistGet(p,&vstr,&vlen,&vlong)) {
+ if (vstr) {
+ if (rdbSaveRawString(fp,vstr,vlen) == -1)
+ return -1;
+ } else {
+ if (rdbSaveLongLongAsStringObject(fp,vlong) == -1)
+ return -1;
+ }
+ p = ziplistNext(o->ptr,p);
+ }
+ } else if (o->encoding == REDIS_ENCODING_LINKEDLIST) {
+ list *list = o->ptr;
+ listIter li;
+ listNode *ln;
- if (rdbSaveStringObject(fp,eleobj) == -1) return -1;
+ if (rdbSaveLen(fp,listLength(list)) == -1) return -1;
+ listRewind(list,&li);
+ while((ln = listNext(&li))) {
+ robj *eleobj = listNodeValue(ln);
+ if (rdbSaveStringObject(fp,eleobj) == -1) return -1;
+ }
+ } else {
+ redisPanic("Unknown list encoding");
}
} else if (o->type == REDIS_SET) {
/* Save a set value */
/* Iterate this DB writing every entry */
while((de = dictNext(di)) != NULL) {
- robj *key = dictGetEntryKey(de);
- robj *o = dictGetEntryVal(de);
- time_t expiretime = getExpire(db,key);
+ sds keystr = dictGetEntryKey(de);
+ robj key, *o = dictGetEntryVal(de);
+ time_t expiretime;
+
+ initStaticStringObject(key,keystr);
+ expiretime = getExpire(db,&key);
/* Save the expire time */
if (expiretime != -1) {
}
/* Save the key and associated value. This requires special
* handling if the value is swapped out. */
- if (!server.vm_enabled || key->storage == REDIS_VM_MEMORY ||
- key->storage == REDIS_VM_SWAPPING) {
+ if (!server.vm_enabled || o->storage == REDIS_VM_MEMORY ||
+ o->storage == REDIS_VM_SWAPPING) {
/* Save type, key, value */
if (rdbSaveType(fp,o->type) == -1) goto werr;
- if (rdbSaveStringObject(fp,key) == -1) goto werr;
+ if (rdbSaveStringObject(fp,&key) == -1) goto werr;
if (rdbSaveObject(fp,o) == -1) goto werr;
} else {
/* REDIS_VM_SWAPPED or REDIS_VM_LOADING */
robj *po;
/* Get a preview of the object in memory */
- po = vmPreviewObject(key);
+ po = vmPreviewObject(o);
/* Save type, key, value */
- if (rdbSaveType(fp,key->vtype) == -1) goto werr;
- if (rdbSaveStringObject(fp,key) == -1) goto werr;
+ if (rdbSaveType(fp,po->type) == -1) goto werr;
+ if (rdbSaveStringObject(fp,&key) == -1) goto werr;
if (rdbSaveObject(fp,po) == -1) goto werr;
/* Remove the loaded object from memory */
decrRefCount(po);
}
}
-static robj *rdbLoadIntegerObject(FILE *fp, int enctype) {
+/* Load an integer-encoded object from file 'fp', with the specified
+ * encoding type 'enctype'. If encode is true the function may return
+ * an integer-encoded object as reply, otherwise the returned object
+ * will always be encoded as a raw string. */
+static robj *rdbLoadIntegerObject(FILE *fp, int enctype, int encode) {
unsigned char enc[4];
long long val;
val = 0; /* anti-warning */
redisPanic("Unknown RDB integer encoding type");
}
- return createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",val));
+ if (encode)
+ return createStringObjectFromLongLong(val);
+ else
+ return createObject(REDIS_STRING,sdsfromlonglong(val));
}
static robj *rdbLoadLzfStringObject(FILE*fp) {
return NULL;
}
-static robj *rdbLoadStringObject(FILE*fp) {
+static robj *rdbGenericLoadStringObject(FILE*fp, int encode) {
int isencoded;
uint32_t len;
sds val;
case REDIS_RDB_ENC_INT8:
case REDIS_RDB_ENC_INT16:
case REDIS_RDB_ENC_INT32:
- return rdbLoadIntegerObject(fp,len);
+ return rdbLoadIntegerObject(fp,len,encode);
case REDIS_RDB_ENC_LZF:
return rdbLoadLzfStringObject(fp);
default:
return createObject(REDIS_STRING,val);
}
+static robj *rdbLoadStringObject(FILE *fp) {
+ return rdbGenericLoadStringObject(fp,0);
+}
+
+static robj *rdbLoadEncodedStringObject(FILE *fp) {
+ return rdbGenericLoadStringObject(fp,1);
+}
+
/* For information about double serialization check rdbSaveDoubleValue() */
static int rdbLoadDoubleValue(FILE *fp, double *val) {
char buf[128];
/* Load a Redis object of the specified type from the specified file.
* On success a newly allocated object is returned, otherwise NULL. */
static robj *rdbLoadObject(int type, FILE *fp) {
- robj *o;
+ robj *o, *ele, *dec;
+ size_t len;
redisLog(REDIS_DEBUG,"LOADING OBJECT %d (at %d)\n",type,ftell(fp));
if (type == REDIS_STRING) {
/* Read string value */
- if ((o = rdbLoadStringObject(fp)) == NULL) return NULL;
+ if ((o = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
o = tryObjectEncoding(o);
- } else if (type == REDIS_LIST || type == REDIS_SET) {
- /* Read list/set value */
- uint32_t listlen;
+ } else if (type == REDIS_LIST) {
+ /* Read list value */
+ if ((len = rdbLoadLen(fp,NULL)) == REDIS_RDB_LENERR) return NULL;
- if ((listlen = rdbLoadLen(fp,NULL)) == REDIS_RDB_LENERR) return NULL;
- o = (type == REDIS_LIST) ? createListObject() : createSetObject();
+ /* Use a real list when there are too many entries */
+ if (len > server.list_max_ziplist_entries) {
+ o = createListObject();
+ } else {
+ o = createZiplistObject();
+ }
+
+ /* Load every single element of the list */
+ while(len--) {
+ if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
+
+ /* If we are using a ziplist and the value is too big, convert
+ * the object to a real list. */
+ if (o->encoding == REDIS_ENCODING_ZIPLIST &&
+ ele->encoding == REDIS_ENCODING_RAW &&
+ sdslen(ele->ptr) > server.list_max_ziplist_value)
+ listTypeConvert(o,REDIS_ENCODING_LINKEDLIST);
+
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ dec = getDecodedObject(ele);
+ o->ptr = ziplistPush(o->ptr,dec->ptr,sdslen(dec->ptr),REDIS_TAIL);
+ decrRefCount(dec);
+ decrRefCount(ele);
+ } else {
+ ele = tryObjectEncoding(ele);
+ listAddNodeTail(o->ptr,ele);
+ }
+ }
+ } else if (type == REDIS_SET) {
+ /* Read list/set value */
+ if ((len = rdbLoadLen(fp,NULL)) == REDIS_RDB_LENERR) return NULL;
+ o = createSetObject();
/* It's faster to expand the dict to the right size asap in order
* to avoid rehashing */
- if (type == REDIS_SET && listlen > DICT_HT_INITIAL_SIZE)
- dictExpand(o->ptr,listlen);
+ if (len > DICT_HT_INITIAL_SIZE)
+ dictExpand(o->ptr,len);
/* Load every single element of the list/set */
- while(listlen--) {
- robj *ele;
-
- if ((ele = rdbLoadStringObject(fp)) == NULL) return NULL;
+ while(len--) {
+ if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
ele = tryObjectEncoding(ele);
- if (type == REDIS_LIST) {
- listAddNodeTail((list*)o->ptr,ele);
- } else {
- dictAdd((dict*)o->ptr,ele,NULL);
- }
+ dictAdd((dict*)o->ptr,ele,NULL);
}
} else if (type == REDIS_ZSET) {
/* Read list/set value */
robj *ele;
double *score = zmalloc(sizeof(double));
- if ((ele = rdbLoadStringObject(fp)) == NULL) return NULL;
+ if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
ele = tryObjectEncoding(ele);
if (rdbLoadDoubleValue(fp,score) == -1) return NULL;
dictAdd(zs->dict,ele,score);
while(hashlen--) {
robj *key, *val;
- if ((key = rdbLoadStringObject(fp)) == NULL) return NULL;
- if ((val = rdbLoadStringObject(fp)) == NULL) return NULL;
+ if ((key = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
+ if ((val = rdbLoadEncodedStringObject(fp)) == NULL) return NULL;
/* If we are using a zipmap and there are too big values
* the object is converted to real hash table encoding. */
if (o->encoding != REDIS_ENCODING_HT &&
- (sdslen(key->ptr) > server.hash_max_zipmap_value ||
- sdslen(val->ptr) > server.hash_max_zipmap_value))
+ ((key->encoding == REDIS_ENCODING_RAW &&
+ sdslen(key->ptr) > server.hash_max_zipmap_value) ||
+ (val->encoding == REDIS_ENCODING_RAW &&
+ sdslen(val->ptr) > server.hash_max_zipmap_value)))
{
convertToRealHash(o);
}
if (o->encoding == REDIS_ENCODING_ZIPMAP) {
unsigned char *zm = o->ptr;
+ robj *deckey, *decval;
- zm = zipmapSet(zm,key->ptr,sdslen(key->ptr),
- val->ptr,sdslen(val->ptr),NULL);
+ /* We need raw string objects to add them to the zipmap */
+ deckey = getDecodedObject(key);
+ decval = getDecodedObject(val);
+ zm = zipmapSet(zm,deckey->ptr,sdslen(deckey->ptr),
+ decval->ptr,sdslen(decval->ptr),NULL);
o->ptr = zm;
+ decrRefCount(deckey);
+ decrRefCount(decval);
decrRefCount(key);
decrRefCount(val);
} else {
static int rdbLoad(char *filename) {
FILE *fp;
- robj *keyobj = NULL;
uint32_t dbid;
int type, retval, rdbver;
- dict *d = server.db[0].dict;
+ int swap_all_values = 0;
redisDb *db = server.db+0;
char buf[1024];
- time_t expiretime = -1, now = time(NULL);
- long long loadedkeys = 0;
+ time_t expiretime, now = time(NULL);
fp = fopen(filename,"r");
if (!fp) return REDIS_ERR;
return REDIS_ERR;
}
while(1) {
- robj *o;
+ robj *key, *val;
+ int force_swapout;
+ expiretime = -1;
/* Read type. */
if ((type = rdbLoadType(fp)) == -1) goto eoferr;
if (type == REDIS_EXPIRETIME) {
exit(1);
}
db = server.db+dbid;
- d = db->dict;
continue;
}
/* Read key */
- if ((keyobj = rdbLoadStringObject(fp)) == NULL) goto eoferr;
+ if ((key = rdbLoadStringObject(fp)) == NULL) goto eoferr;
/* Read value */
- if ((o = rdbLoadObject(type,fp)) == NULL) goto eoferr;
+ if ((val = rdbLoadObject(type,fp)) == NULL) goto eoferr;
+ /* Check if the key already expired */
+ if (expiretime != -1 && expiretime < now) {
+ decrRefCount(key);
+ decrRefCount(val);
+ continue;
+ }
/* Add the new object in the hash table */
- retval = dictAdd(d,keyobj,o);
- if (retval == DICT_ERR) {
- redisLog(REDIS_WARNING,"Loading DB, duplicated key (%s) found! Unrecoverable error, exiting now.", keyobj->ptr);
+ retval = dbAdd(db,key,val);
+ if (retval == REDIS_ERR) {
+ redisLog(REDIS_WARNING,"Loading DB, duplicated key (%s) found! Unrecoverable error, exiting now.", key->ptr);
exit(1);
}
/* Set the expire time if needed */
- if (expiretime != -1) {
- setExpire(db,keyobj,expiretime);
- /* Delete this key if already expired */
- if (expiretime < now) deleteKey(db,keyobj);
- expiretime = -1;
- }
- keyobj = o = NULL;
+ if (expiretime != -1) setExpire(db,key,expiretime);
+
/* Handle swapping while loading big datasets when VM is on */
- loadedkeys++;
- if (server.vm_enabled && (loadedkeys % 5000) == 0) {
+
+ /* If we detecter we are hopeless about fitting something in memory
+ * we just swap every new key on disk. Directly...
+ * Note that's important to check for this condition before resorting
+ * to random sampling, otherwise we may try to swap already
+ * swapped keys. */
+ if (swap_all_values) {
+ dictEntry *de = dictFind(db->dict,key->ptr);
+
+ /* de may be NULL since the key already expired */
+ if (de) {
+ vmpointer *vp;
+ val = dictGetEntryVal(de);
+
+ if (val->refcount == 1 &&
+ (vp = vmSwapObjectBlocking(val)) != NULL)
+ dictGetEntryVal(de) = vp;
+ }
+ decrRefCount(key);
+ continue;
+ }
+ decrRefCount(key);
+
+ /* Flush data on disk once 32 MB of additional RAM are used... */
+ force_swapout = 0;
+ if ((zmalloc_used_memory() - server.vm_max_memory) > 1024*1024*32)
+ force_swapout = 1;
+
+ /* If we have still some hope of having some value fitting memory
+ * then we try random sampling. */
+ if (!swap_all_values && server.vm_enabled && force_swapout) {
while (zmalloc_used_memory() > server.vm_max_memory) {
if (vmSwapOneObjectBlocking() == REDIS_ERR) break;
}
+ if (zmalloc_used_memory() > server.vm_max_memory)
+ swap_all_values = 1; /* We are already using too much mem */
}
}
fclose(fp);
return REDIS_OK;
eoferr: /* unexpected end of file is handled here with a fatal exit */
- if (keyobj) decrRefCount(keyobj);
redisLog(REDIS_WARNING,"Short read or OOM loading DB. Unrecoverable error, aborting now.");
exit(1);
return REDIS_ERR; /* Just to avoid warning */
}
+/*================================== Shutdown =============================== */
+static int prepareForShutdown() {
+ redisLog(REDIS_WARNING,"User requested shutdown, saving DB...");
+ /* Kill the saving child if there is a background saving in progress.
+ We want to avoid race conditions, for instance our saving child may
+ overwrite the synchronous saving did by SHUTDOWN. */
+ if (server.bgsavechildpid != -1) {
+ redisLog(REDIS_WARNING,"There is a live saving child. Killing it!");
+ kill(server.bgsavechildpid,SIGKILL);
+ rdbRemoveTempFile(server.bgsavechildpid);
+ }
+ if (server.appendonly) {
+ /* Append only file: fsync() the AOF and exit */
+ aof_fsync(server.appendfd);
+ if (server.vm_enabled) unlink(server.vm_swap_file);
+ } else {
+ /* Snapshotting. Perform a SYNC SAVE and exit */
+ if (rdbSave(server.dbfilename) == REDIS_OK) {
+ if (server.daemonize)
+ unlink(server.pidfile);
+ redisLog(REDIS_WARNING,"%zu bytes used at exit",zmalloc_used_memory());
+ } else {
+ /* Ooops.. error saving! The best we can do is to continue
+ * operating. Note that if there was a background saving process,
+ * in the next cron() Redis will be notified that the background
+ * saving aborted, handling special stuff like slaves pending for
+ * synchronization... */
+ redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit");
+ return REDIS_ERR;
+ }
+ }
+ redisLog(REDIS_WARNING,"Server exit now, bye bye...");
+ return REDIS_OK;
+}
+
/*================================== Commands =============================== */
static void authCommand(redisClient *c) {
}
}
+ touchWatchedKey(c->db,key);
if (nx) deleteIfVolatile(c->db,key);
- retval = dictAdd(c->db->dict,key,val);
- if (retval == DICT_ERR) {
+ retval = dbAdd(c->db,key,val);
+ if (retval == REDIS_ERR) {
if (!nx) {
- /* If the key is about a swapped value, we want a new key object
- * to overwrite the old. So we delete the old key in the database.
- * This will also make sure that swap pages about the old object
- * will be marked as free. */
- if (server.vm_enabled && deleteIfSwapped(c->db,key))
- incrRefCount(key);
- dictReplace(c->db->dict,key,val);
+ dbReplace(c->db,key,val);
incrRefCount(val);
} else {
addReply(c,shared.czero);
return;
}
} else {
- incrRefCount(key);
incrRefCount(val);
}
server.dirty++;
static void getsetCommand(redisClient *c) {
if (getGenericCommand(c) == REDIS_ERR) return;
- if (dictAdd(c->db->dict,c->argv[1],c->argv[2]) == DICT_ERR) {
- dictReplace(c->db->dict,c->argv[1],c->argv[2]);
- } else {
- incrRefCount(c->argv[1]);
- }
+ dbReplace(c->db,c->argv[1],c->argv[2]);
incrRefCount(c->argv[2]);
server.dirty++;
removeExpire(c->db,c->argv[1]);
}
for (j = 1; j < c->argc; j += 2) {
- int retval;
-
c->argv[j+1] = tryObjectEncoding(c->argv[j+1]);
- retval = dictAdd(c->db->dict,c->argv[j],c->argv[j+1]);
- if (retval == DICT_ERR) {
- dictReplace(c->db->dict,c->argv[j],c->argv[j+1]);
- incrRefCount(c->argv[j+1]);
- } else {
- incrRefCount(c->argv[j]);
- incrRefCount(c->argv[j+1]);
- }
+ dbReplace(c->db,c->argv[j],c->argv[j+1]);
+ incrRefCount(c->argv[j+1]);
removeExpire(c->db,c->argv[j]);
}
server.dirty += (c->argc-1)/2;
static void incrDecrCommand(redisClient *c, long long incr) {
long long value;
- int retval;
robj *o;
o = lookupKeyWrite(c->db,c->argv[1]);
-
- if (getLongLongFromObjectOrReply(c, o, &value, NULL) != REDIS_OK) return;
+ if (o != NULL && checkType(c,o,REDIS_STRING)) return;
+ if (getLongLongFromObjectOrReply(c,o,&value,NULL) != REDIS_OK) return;
value += incr;
- o = createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",value));
- o = tryObjectEncoding(o);
- retval = dictAdd(c->db->dict,c->argv[1],o);
- if (retval == DICT_ERR) {
- dictReplace(c->db->dict,c->argv[1],o);
- removeExpire(c->db,c->argv[1]);
- } else {
- incrRefCount(c->argv[1]);
- }
+ o = createStringObjectFromLongLong(value);
+ dbReplace(c->db,c->argv[1],o);
server.dirty++;
addReply(c,shared.colon);
addReply(c,o);
o = lookupKeyWrite(c->db,c->argv[1]);
if (o == NULL) {
/* Create the key */
- retval = dictAdd(c->db->dict,c->argv[1],c->argv[2]);
- incrRefCount(c->argv[1]);
+ retval = dbAdd(c->db,c->argv[1],c->argv[2]);
incrRefCount(c->argv[2]);
totlen = stringObjectLen(c->argv[2]);
} else {
- dictEntry *de;
-
- de = dictFind(c->db->dict,c->argv[1]);
- assert(de != NULL);
-
- o = dictGetEntryVal(de);
if (o->type != REDIS_STRING) {
addReply(c,shared.wrongtypeerr);
return;
o = createStringObject(decoded->ptr, sdslen(decoded->ptr));
decrRefCount(decoded);
- dictReplace(c->db->dict,c->argv[1],o);
+ dbReplace(c->db,c->argv[1],o);
}
/* APPEND! */
if (c->argv[2]->encoding == REDIS_ENCODING_RAW) {
int deleted = 0, j;
for (j = 1; j < c->argc; j++) {
- if (deleteKey(c->db,c->argv[j])) {
+ if (dbDelete(c->db,c->argv[j])) {
+ touchWatchedKey(c->db,c->argv[j]);
server.dirty++;
deleted++;
}
}
- addReplyLong(c,deleted);
+ addReplyLongLong(c,deleted);
}
static void existsCommand(redisClient *c) {
- addReply(c,lookupKeyRead(c->db,c->argv[1]) ? shared.cone : shared.czero);
+ expireIfNeeded(c->db,c->argv[1]);
+ if (dbExists(c->db,c->argv[1])) {
+ addReply(c, shared.cone);
+ } else {
+ addReply(c, shared.czero);
+ }
}
static void selectCommand(redisClient *c) {
}
static void randomkeyCommand(redisClient *c) {
- dictEntry *de;
robj *key;
- while(1) {
- de = dictGetRandomKey(c->db->dict);
- if (!de || expireIfNeeded(c->db,dictGetEntryKey(de)) == 0) break;
- }
-
- if (de == NULL) {
+ if ((key = dbRandomKey(c->db)) == NULL) {
addReply(c,shared.nullbulk);
return;
}
- key = dictGetEntryKey(de);
- if (server.vm_enabled) {
- key = dupStringObject(key);
- addReplyBulk(c,key);
- decrRefCount(key);
- } else {
- addReplyBulk(c,key);
- }
+ addReplyBulk(c,key);
+ decrRefCount(key);
}
static void keysCommand(redisClient *c) {
addReply(c,lenobj);
decrRefCount(lenobj);
while((de = dictNext(di)) != NULL) {
- robj *keyobj = dictGetEntryKey(de);
+ sds key = dictGetEntryKey(de);
+ robj *keyobj;
- sds key = keyobj->ptr;
if ((pattern[0] == '*' && pattern[1] == '\0') ||
stringmatchlen(pattern,plen,key,sdslen(key),0)) {
+ keyobj = createStringObject(key,sdslen(key));
if (expireIfNeeded(c->db,keyobj) == 0) {
addReplyBulk(c,keyobj);
numkeys++;
}
+ decrRefCount(keyobj);
}
}
dictReleaseIterator(di);
}
static void shutdownCommand(redisClient *c) {
- redisLog(REDIS_WARNING,"User requested shutdown, saving DB...");
- /* Kill the saving child if there is a background saving in progress.
- We want to avoid race conditions, for instance our saving child may
- overwrite the synchronous saving did by SHUTDOWN. */
- if (server.bgsavechildpid != -1) {
- redisLog(REDIS_WARNING,"There is a live saving child. Killing it!");
- kill(server.bgsavechildpid,SIGKILL);
- rdbRemoveTempFile(server.bgsavechildpid);
- }
- if (server.appendonly) {
- /* Append only file: fsync() the AOF and exit */
- fsync(server.appendfd);
- if (server.vm_enabled) unlink(server.vm_swap_file);
+ if (prepareForShutdown() == REDIS_OK)
exit(0);
- } else {
- /* Snapshotting. Perform a SYNC SAVE and exit */
- if (rdbSave(server.dbfilename) == REDIS_OK) {
- if (server.daemonize)
- unlink(server.pidfile);
- redisLog(REDIS_WARNING,"%zu bytes used at exit",zmalloc_used_memory());
- redisLog(REDIS_WARNING,"Server exit now, bye bye...");
- if (server.vm_enabled) unlink(server.vm_swap_file);
- exit(0);
- } else {
- /* Ooops.. error saving! The best we can do is to continue
- * operating. Note that if there was a background saving process,
- * in the next cron() Redis will be notified that the background
- * saving aborted, handling special stuff like slaves pending for
- * synchronization... */
- redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit");
- addReplySds(c,
- sdsnew("-ERR can't quit, problems saving the DB\r\n"));
- }
- }
+ addReplySds(c, sdsnew("-ERR Errors trying to SHUTDOWN. Check logs.\r\n"));
}
static void renameGenericCommand(redisClient *c, int nx) {
incrRefCount(o);
deleteIfVolatile(c->db,c->argv[2]);
- if (dictAdd(c->db->dict,c->argv[2],o) == DICT_ERR) {
+ if (dbAdd(c->db,c->argv[2],o) == REDIS_ERR) {
if (nx) {
decrRefCount(o);
addReply(c,shared.czero);
return;
}
- dictReplace(c->db->dict,c->argv[2],o);
- } else {
- incrRefCount(c->argv[2]);
+ dbReplace(c->db,c->argv[2],o);
}
- deleteKey(c->db,c->argv[1]);
+ dbDelete(c->db,c->argv[1]);
+ touchWatchedKey(c->db,c->argv[2]);
server.dirty++;
addReply(c,nx ? shared.cone : shared.ok);
}
/* Try to add the element to the target DB */
deleteIfVolatile(dst,c->argv[1]);
- if (dictAdd(dst->dict,c->argv[1],o) == DICT_ERR) {
+ if (dbAdd(dst,c->argv[1],o) == REDIS_ERR) {
addReply(c,shared.czero);
return;
}
- incrRefCount(c->argv[1]);
incrRefCount(o);
/* OK! key moved, free the entry in the source DB */
- deleteKey(src,c->argv[1]);
+ dbDelete(src,c->argv[1]);
server.dirty++;
addReply(c,shared.cone);
}
/* =================================== Lists ================================ */
-static void pushGenericCommand(redisClient *c, int where) {
- robj *lobj;
- list *list;
- lobj = lookupKeyWrite(c->db,c->argv[1]);
+
+/* Check the argument length to see if it requires us to convert the ziplist
+ * to a real list. Only check raw-encoded objects because integer encoded
+ * objects are never too long. */
+static void listTypeTryConversion(robj *subject, robj *value) {
+ if (subject->encoding != REDIS_ENCODING_ZIPLIST) return;
+ if (value->encoding == REDIS_ENCODING_RAW &&
+ sdslen(value->ptr) > server.list_max_ziplist_value)
+ listTypeConvert(subject,REDIS_ENCODING_LINKEDLIST);
+}
+
+static void listTypePush(robj *subject, robj *value, int where) {
+ /* Check if we need to convert the ziplist */
+ listTypeTryConversion(subject,value);
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST &&
+ ziplistLen(subject->ptr) >= server.list_max_ziplist_entries)
+ listTypeConvert(subject,REDIS_ENCODING_LINKEDLIST);
+
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST) {
+ int pos = (where == REDIS_HEAD) ? ZIPLIST_HEAD : ZIPLIST_TAIL;
+ value = getDecodedObject(value);
+ subject->ptr = ziplistPush(subject->ptr,value->ptr,sdslen(value->ptr),pos);
+ decrRefCount(value);
+ } else if (subject->encoding == REDIS_ENCODING_LINKEDLIST) {
+ if (where == REDIS_HEAD) {
+ listAddNodeHead(subject->ptr,value);
+ } else {
+ listAddNodeTail(subject->ptr,value);
+ }
+ incrRefCount(value);
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+}
+
+static robj *listTypePop(robj *subject, int where) {
+ robj *value = NULL;
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *p;
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vlong;
+ int pos = (where == REDIS_HEAD) ? 0 : -1;
+ p = ziplistIndex(subject->ptr,pos);
+ if (ziplistGet(p,&vstr,&vlen,&vlong)) {
+ if (vstr) {
+ value = createStringObject((char*)vstr,vlen);
+ } else {
+ value = createStringObjectFromLongLong(vlong);
+ }
+ /* We only need to delete an element when it exists */
+ subject->ptr = ziplistDelete(subject->ptr,&p);
+ }
+ } else if (subject->encoding == REDIS_ENCODING_LINKEDLIST) {
+ list *list = subject->ptr;
+ listNode *ln;
+ if (where == REDIS_HEAD) {
+ ln = listFirst(list);
+ } else {
+ ln = listLast(list);
+ }
+ if (ln != NULL) {
+ value = listNodeValue(ln);
+ incrRefCount(value);
+ listDelNode(list,ln);
+ }
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+ return value;
+}
+
+static unsigned long listTypeLength(robj *subject) {
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST) {
+ return ziplistLen(subject->ptr);
+ } else if (subject->encoding == REDIS_ENCODING_LINKEDLIST) {
+ return listLength((list*)subject->ptr);
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+}
+
+/* Structure to hold set iteration abstraction. */
+typedef struct {
+ robj *subject;
+ unsigned char encoding;
+ unsigned char direction; /* Iteration direction */
+ unsigned char *zi;
+ listNode *ln;
+} listTypeIterator;
+
+/* Structure for an entry while iterating over a list. */
+typedef struct {
+ listTypeIterator *li;
+ unsigned char *zi; /* Entry in ziplist */
+ listNode *ln; /* Entry in linked list */
+} listTypeEntry;
+
+/* Initialize an iterator at the specified index. */
+static listTypeIterator *listTypeInitIterator(robj *subject, int index, unsigned char direction) {
+ listTypeIterator *li = zmalloc(sizeof(listTypeIterator));
+ li->subject = subject;
+ li->encoding = subject->encoding;
+ li->direction = direction;
+ if (li->encoding == REDIS_ENCODING_ZIPLIST) {
+ li->zi = ziplistIndex(subject->ptr,index);
+ } else if (li->encoding == REDIS_ENCODING_LINKEDLIST) {
+ li->ln = listIndex(subject->ptr,index);
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+ return li;
+}
+
+/* Clean up the iterator. */
+static void listTypeReleaseIterator(listTypeIterator *li) {
+ zfree(li);
+}
+
+/* Stores pointer to current the entry in the provided entry structure
+ * and advances the position of the iterator. Returns 1 when the current
+ * entry is in fact an entry, 0 otherwise. */
+static int listTypeNext(listTypeIterator *li, listTypeEntry *entry) {
+ /* Protect from converting when iterating */
+ redisAssert(li->subject->encoding == li->encoding);
+
+ entry->li = li;
+ if (li->encoding == REDIS_ENCODING_ZIPLIST) {
+ entry->zi = li->zi;
+ if (entry->zi != NULL) {
+ if (li->direction == REDIS_TAIL)
+ li->zi = ziplistNext(li->subject->ptr,li->zi);
+ else
+ li->zi = ziplistPrev(li->subject->ptr,li->zi);
+ return 1;
+ }
+ } else if (li->encoding == REDIS_ENCODING_LINKEDLIST) {
+ entry->ln = li->ln;
+ if (entry->ln != NULL) {
+ if (li->direction == REDIS_TAIL)
+ li->ln = li->ln->next;
+ else
+ li->ln = li->ln->prev;
+ return 1;
+ }
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+ return 0;
+}
+
+/* Return entry or NULL at the current position of the iterator. */
+static robj *listTypeGet(listTypeEntry *entry) {
+ listTypeIterator *li = entry->li;
+ robj *value = NULL;
+ if (li->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vlong;
+ redisAssert(entry->zi != NULL);
+ if (ziplistGet(entry->zi,&vstr,&vlen,&vlong)) {
+ if (vstr) {
+ value = createStringObject((char*)vstr,vlen);
+ } else {
+ value = createStringObjectFromLongLong(vlong);
+ }
+ }
+ } else if (li->encoding == REDIS_ENCODING_LINKEDLIST) {
+ redisAssert(entry->ln != NULL);
+ value = listNodeValue(entry->ln);
+ incrRefCount(value);
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+ return value;
+}
+
+static void listTypeInsert(listTypeEntry *entry, robj *value, int where) {
+ robj *subject = entry->li->subject;
+ if (entry->li->encoding == REDIS_ENCODING_ZIPLIST) {
+ value = getDecodedObject(value);
+ if (where == REDIS_TAIL) {
+ unsigned char *next = ziplistNext(subject->ptr,entry->zi);
+
+ /* When we insert after the current element, but the current element
+ * is the tail of the list, we need to do a push. */
+ if (next == NULL) {
+ subject->ptr = ziplistPush(subject->ptr,value->ptr,sdslen(value->ptr),REDIS_TAIL);
+ } else {
+ subject->ptr = ziplistInsert(subject->ptr,next,value->ptr,sdslen(value->ptr));
+ }
+ } else {
+ subject->ptr = ziplistInsert(subject->ptr,entry->zi,value->ptr,sdslen(value->ptr));
+ }
+ decrRefCount(value);
+ } else if (entry->li->encoding == REDIS_ENCODING_LINKEDLIST) {
+ if (where == REDIS_TAIL) {
+ listInsertNode(subject->ptr,entry->ln,value,AL_START_TAIL);
+ } else {
+ listInsertNode(subject->ptr,entry->ln,value,AL_START_HEAD);
+ }
+ incrRefCount(value);
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+}
+
+/* Compare the given object with the entry at the current position. */
+static int listTypeEqual(listTypeEntry *entry, robj *o) {
+ listTypeIterator *li = entry->li;
+ if (li->encoding == REDIS_ENCODING_ZIPLIST) {
+ redisAssert(o->encoding == REDIS_ENCODING_RAW);
+ return ziplistCompare(entry->zi,o->ptr,sdslen(o->ptr));
+ } else if (li->encoding == REDIS_ENCODING_LINKEDLIST) {
+ return equalStringObjects(o,listNodeValue(entry->ln));
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+}
+
+/* Delete the element pointed to. */
+static void listTypeDelete(listTypeEntry *entry) {
+ listTypeIterator *li = entry->li;
+ if (li->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *p = entry->zi;
+ li->subject->ptr = ziplistDelete(li->subject->ptr,&p);
+
+ /* Update position of the iterator depending on the direction */
+ if (li->direction == REDIS_TAIL)
+ li->zi = p;
+ else
+ li->zi = ziplistPrev(li->subject->ptr,p);
+ } else if (entry->li->encoding == REDIS_ENCODING_LINKEDLIST) {
+ listNode *next;
+ if (li->direction == REDIS_TAIL)
+ next = entry->ln->next;
+ else
+ next = entry->ln->prev;
+ listDelNode(li->subject->ptr,entry->ln);
+ li->ln = next;
+ } else {
+ redisPanic("Unknown list encoding");
+ }
+}
+
+static void listTypeConvert(robj *subject, int enc) {
+ listTypeIterator *li;
+ listTypeEntry entry;
+ redisAssert(subject->type == REDIS_LIST);
+
+ if (enc == REDIS_ENCODING_LINKEDLIST) {
+ list *l = listCreate();
+ listSetFreeMethod(l,decrRefCount);
+
+ /* listTypeGet returns a robj with incremented refcount */
+ li = listTypeInitIterator(subject,0,REDIS_TAIL);
+ while (listTypeNext(li,&entry)) listAddNodeTail(l,listTypeGet(&entry));
+ listTypeReleaseIterator(li);
+
+ subject->encoding = REDIS_ENCODING_LINKEDLIST;
+ zfree(subject->ptr);
+ subject->ptr = l;
+ } else {
+ redisPanic("Unsupported list conversion");
+ }
+}
+
+static void pushGenericCommand(redisClient *c, int where) {
+ robj *lobj = lookupKeyWrite(c->db,c->argv[1]);
if (lobj == NULL) {
if (handleClientsWaitingListPush(c,c->argv[1],c->argv[2])) {
addReply(c,shared.cone);
return;
}
- lobj = createListObject();
- list = lobj->ptr;
- if (where == REDIS_HEAD) {
- listAddNodeHead(list,c->argv[2]);
- } else {
- listAddNodeTail(list,c->argv[2]);
- }
- dictAdd(c->db->dict,c->argv[1],lobj);
- incrRefCount(c->argv[1]);
- incrRefCount(c->argv[2]);
+ lobj = createZiplistObject();
+ dbAdd(c->db,c->argv[1],lobj);
} else {
if (lobj->type != REDIS_LIST) {
addReply(c,shared.wrongtypeerr);
addReply(c,shared.cone);
return;
}
- list = lobj->ptr;
- if (where == REDIS_HEAD) {
- listAddNodeHead(list,c->argv[2]);
- } else {
- listAddNodeTail(list,c->argv[2]);
- }
- incrRefCount(c->argv[2]);
}
+ listTypePush(lobj,c->argv[2],where);
+ addReplyLongLong(c,listTypeLength(lobj));
server.dirty++;
- addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",listLength(list)));
}
static void lpushCommand(redisClient *c) {
pushGenericCommand(c,REDIS_TAIL);
}
+static void pushxGenericCommand(redisClient *c, robj *refval, robj *val, int where) {
+ robj *subject;
+ listTypeIterator *iter;
+ listTypeEntry entry;
+ int inserted = 0;
+
+ if ((subject = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL ||
+ checkType(c,subject,REDIS_LIST)) return;
+
+ if (refval != NULL) {
+ /* Note: we expect refval to be string-encoded because it is *not* the
+ * last argument of the multi-bulk LINSERT. */
+ redisAssert(refval->encoding == REDIS_ENCODING_RAW);
+
+ /* We're not sure if this value can be inserted yet, but we cannot
+ * convert the list inside the iterator. We don't want to loop over
+ * the list twice (once to see if the value can be inserted and once
+ * to do the actual insert), so we assume this value can be inserted
+ * and convert the ziplist to a regular list if necessary. */
+ listTypeTryConversion(subject,val);
+
+ /* Seek refval from head to tail */
+ iter = listTypeInitIterator(subject,0,REDIS_TAIL);
+ while (listTypeNext(iter,&entry)) {
+ if (listTypeEqual(&entry,refval)) {
+ listTypeInsert(&entry,val,where);
+ inserted = 1;
+ break;
+ }
+ }
+ listTypeReleaseIterator(iter);
+
+ if (inserted) {
+ /* Check if the length exceeds the ziplist length threshold. */
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST &&
+ ziplistLen(subject->ptr) > server.list_max_ziplist_entries)
+ listTypeConvert(subject,REDIS_ENCODING_LINKEDLIST);
+ server.dirty++;
+ } else {
+ /* Notify client of a failed insert */
+ addReply(c,shared.cnegone);
+ return;
+ }
+ } else {
+ listTypePush(subject,val,where);
+ server.dirty++;
+ }
+
+ addReplyUlong(c,listTypeLength(subject));
+}
+
+static void lpushxCommand(redisClient *c) {
+ pushxGenericCommand(c,NULL,c->argv[2],REDIS_HEAD);
+}
+
+static void rpushxCommand(redisClient *c) {
+ pushxGenericCommand(c,NULL,c->argv[2],REDIS_TAIL);
+}
+
+static void linsertCommand(redisClient *c) {
+ if (strcasecmp(c->argv[2]->ptr,"after") == 0) {
+ pushxGenericCommand(c,c->argv[3],c->argv[4],REDIS_TAIL);
+ } else if (strcasecmp(c->argv[2]->ptr,"before") == 0) {
+ pushxGenericCommand(c,c->argv[3],c->argv[4],REDIS_HEAD);
+ } else {
+ addReply(c,shared.syntaxerr);
+ }
+}
+
static void llenCommand(redisClient *c) {
- robj *o;
- list *l;
-
- if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL ||
- checkType(c,o,REDIS_LIST)) return;
-
- l = o->ptr;
- addReplyUlong(c,listLength(l));
+ robj *o = lookupKeyReadOrReply(c,c->argv[1],shared.czero);
+ if (o == NULL || checkType(c,o,REDIS_LIST)) return;
+ addReplyUlong(c,listTypeLength(o));
}
static void lindexCommand(redisClient *c) {
- robj *o;
+ robj *o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk);
+ if (o == NULL || checkType(c,o,REDIS_LIST)) return;
int index = atoi(c->argv[2]->ptr);
- list *list;
- listNode *ln;
-
- if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL ||
- checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
+ robj *value = NULL;
- ln = listIndex(list, index);
- if (ln == NULL) {
- addReply(c,shared.nullbulk);
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *p;
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vlong;
+ p = ziplistIndex(o->ptr,index);
+ if (ziplistGet(p,&vstr,&vlen,&vlong)) {
+ if (vstr) {
+ value = createStringObject((char*)vstr,vlen);
+ } else {
+ value = createStringObjectFromLongLong(vlong);
+ }
+ addReplyBulk(c,value);
+ decrRefCount(value);
+ } else {
+ addReply(c,shared.nullbulk);
+ }
+ } else if (o->encoding == REDIS_ENCODING_LINKEDLIST) {
+ listNode *ln = listIndex(o->ptr,index);
+ if (ln != NULL) {
+ value = listNodeValue(ln);
+ addReplyBulk(c,value);
+ } else {
+ addReply(c,shared.nullbulk);
+ }
} else {
- robj *ele = listNodeValue(ln);
- addReplyBulk(c,ele);
+ redisPanic("Unknown list encoding");
}
}
static void lsetCommand(redisClient *c) {
- robj *o;
+ robj *o = lookupKeyWriteOrReply(c,c->argv[1],shared.nokeyerr);
+ if (o == NULL || checkType(c,o,REDIS_LIST)) return;
int index = atoi(c->argv[2]->ptr);
- list *list;
- listNode *ln;
-
- if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.nokeyerr)) == NULL ||
- checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
-
- ln = listIndex(list, index);
- if (ln == NULL) {
- addReply(c,shared.outofrangeerr);
+ robj *value = c->argv[3];
+
+ listTypeTryConversion(o,value);
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *p, *zl = o->ptr;
+ p = ziplistIndex(zl,index);
+ if (p == NULL) {
+ addReply(c,shared.outofrangeerr);
+ } else {
+ o->ptr = ziplistDelete(o->ptr,&p);
+ value = getDecodedObject(value);
+ o->ptr = ziplistInsert(o->ptr,p,value->ptr,sdslen(value->ptr));
+ decrRefCount(value);
+ addReply(c,shared.ok);
+ server.dirty++;
+ }
+ } else if (o->encoding == REDIS_ENCODING_LINKEDLIST) {
+ listNode *ln = listIndex(o->ptr,index);
+ if (ln == NULL) {
+ addReply(c,shared.outofrangeerr);
+ } else {
+ decrRefCount((robj*)listNodeValue(ln));
+ listNodeValue(ln) = value;
+ incrRefCount(value);
+ addReply(c,shared.ok);
+ server.dirty++;
+ }
} else {
- robj *ele = listNodeValue(ln);
-
- decrRefCount(ele);
- listNodeValue(ln) = c->argv[3];
- incrRefCount(c->argv[3]);
- addReply(c,shared.ok);
- server.dirty++;
+ redisPanic("Unknown list encoding");
}
}
static void popGenericCommand(redisClient *c, int where) {
- robj *o;
- list *list;
- listNode *ln;
-
- if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.nullbulk)) == NULL ||
- checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
-
- if (where == REDIS_HEAD)
- ln = listFirst(list);
- else
- ln = listLast(list);
+ robj *o = lookupKeyWriteOrReply(c,c->argv[1],shared.nullbulk);
+ if (o == NULL || checkType(c,o,REDIS_LIST)) return;
- if (ln == NULL) {
+ robj *value = listTypePop(o,where);
+ if (value == NULL) {
addReply(c,shared.nullbulk);
} else {
- robj *ele = listNodeValue(ln);
- addReplyBulk(c,ele);
- listDelNode(list,ln);
- if (listLength(list) == 0) deleteKey(c->db,c->argv[1]);
+ addReplyBulk(c,value);
+ decrRefCount(value);
+ if (listTypeLength(o) == 0) dbDelete(c->db,c->argv[1]);
server.dirty++;
}
}
}
static void lrangeCommand(redisClient *c) {
- robj *o;
+ robj *o, *value;
int start = atoi(c->argv[2]->ptr);
int end = atoi(c->argv[3]->ptr);
int llen;
int rangelen, j;
- list *list;
- listNode *ln;
- robj *ele;
+ listTypeEntry entry;
if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptymultibulk)) == NULL
|| checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
- llen = listLength(list);
+ llen = listTypeLength(o);
/* convert negative indexes */
if (start < 0) start = llen+start;
rangelen = (end-start)+1;
/* Return the result in form of a multi-bulk reply */
- ln = listIndex(list, start);
addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",rangelen));
+ listTypeIterator *li = listTypeInitIterator(o,start,REDIS_TAIL);
for (j = 0; j < rangelen; j++) {
- ele = listNodeValue(ln);
- addReplyBulk(c,ele);
- ln = ln->next;
+ redisAssert(listTypeNext(li,&entry));
+ value = listTypeGet(&entry);
+ addReplyBulk(c,value);
+ decrRefCount(value);
}
+ listTypeReleaseIterator(li);
}
static void ltrimCommand(redisClient *c) {
if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.ok)) == NULL ||
checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
- llen = listLength(list);
+ llen = listTypeLength(o);
/* convert negative indexes */
if (start < 0) start = llen+start;
}
/* Remove list elements to perform the trim */
- for (j = 0; j < ltrim; j++) {
- ln = listFirst(list);
- listDelNode(list,ln);
- }
- for (j = 0; j < rtrim; j++) {
- ln = listLast(list);
- listDelNode(list,ln);
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ o->ptr = ziplistDeleteRange(o->ptr,0,ltrim);
+ o->ptr = ziplistDeleteRange(o->ptr,-rtrim,rtrim);
+ } else if (o->encoding == REDIS_ENCODING_LINKEDLIST) {
+ list = o->ptr;
+ for (j = 0; j < ltrim; j++) {
+ ln = listFirst(list);
+ listDelNode(list,ln);
+ }
+ for (j = 0; j < rtrim; j++) {
+ ln = listLast(list);
+ listDelNode(list,ln);
+ }
+ } else {
+ redisPanic("Unknown list encoding");
}
- if (listLength(list) == 0) deleteKey(c->db,c->argv[1]);
+ if (listTypeLength(o) == 0) dbDelete(c->db,c->argv[1]);
server.dirty++;
addReply(c,shared.ok);
}
static void lremCommand(redisClient *c) {
- robj *o;
- list *list;
- listNode *ln, *next;
+ robj *subject, *obj = c->argv[3];
int toremove = atoi(c->argv[2]->ptr);
int removed = 0;
- int fromtail = 0;
+ listTypeEntry entry;
- if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL ||
- checkType(c,o,REDIS_LIST)) return;
- list = o->ptr;
+ subject = lookupKeyWriteOrReply(c,c->argv[1],shared.czero);
+ if (subject == NULL || checkType(c,subject,REDIS_LIST)) return;
+
+ /* Make sure obj is raw when we're dealing with a ziplist */
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST)
+ obj = getDecodedObject(obj);
+ listTypeIterator *li;
if (toremove < 0) {
toremove = -toremove;
- fromtail = 1;
+ li = listTypeInitIterator(subject,-1,REDIS_HEAD);
+ } else {
+ li = listTypeInitIterator(subject,0,REDIS_TAIL);
}
- ln = fromtail ? list->tail : list->head;
- while (ln) {
- robj *ele = listNodeValue(ln);
- next = fromtail ? ln->prev : ln->next;
- if (compareStringObjects(ele,c->argv[3]) == 0) {
- listDelNode(list,ln);
+ while (listTypeNext(li,&entry)) {
+ if (listTypeEqual(&entry,obj)) {
+ listTypeDelete(&entry);
server.dirty++;
removed++;
if (toremove && removed == toremove) break;
}
- ln = next;
}
- if (listLength(list) == 0) deleteKey(c->db,c->argv[1]);
+ listTypeReleaseIterator(li);
+
+ /* Clean up raw encoded object */
+ if (subject->encoding == REDIS_ENCODING_ZIPLIST)
+ decrRefCount(obj);
+
+ if (listTypeLength(subject) == 0) dbDelete(c->db,c->argv[1]);
addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",removed));
}
* as well. This command was originally proposed by Ezra Zygmuntowicz.
*/
static void rpoplpushcommand(redisClient *c) {
- robj *sobj;
- list *srclist;
- listNode *ln;
-
+ robj *sobj, *value;
if ((sobj = lookupKeyWriteOrReply(c,c->argv[1],shared.nullbulk)) == NULL ||
checkType(c,sobj,REDIS_LIST)) return;
- srclist = sobj->ptr;
- ln = listLast(srclist);
- if (ln == NULL) {
+ if (listTypeLength(sobj) == 0) {
addReply(c,shared.nullbulk);
} else {
robj *dobj = lookupKeyWrite(c->db,c->argv[2]);
- robj *ele = listNodeValue(ln);
- list *dstlist;
-
- if (dobj && dobj->type != REDIS_LIST) {
- addReply(c,shared.wrongtypeerr);
- return;
- }
+ if (dobj && checkType(c,dobj,REDIS_LIST)) return;
+ value = listTypePop(sobj,REDIS_TAIL);
/* Add the element to the target list (unless it's directly
* passed to some BLPOP-ing client */
- if (!handleClientsWaitingListPush(c,c->argv[2],ele)) {
- if (dobj == NULL) {
- /* Create the list if the key does not exist */
- dobj = createListObject();
- dictAdd(c->db->dict,c->argv[2],dobj);
- incrRefCount(c->argv[2]);
+ if (!handleClientsWaitingListPush(c,c->argv[2],value)) {
+ /* Create the list if the key does not exist */
+ if (!dobj) {
+ dobj = createZiplistObject();
+ dbAdd(c->db,c->argv[2],dobj);
}
- dstlist = dobj->ptr;
- listAddNodeHead(dstlist,ele);
- incrRefCount(ele);
+ listTypePush(dobj,value,REDIS_HEAD);
}
/* Send the element to the client as reply as well */
- addReplyBulk(c,ele);
+ addReplyBulk(c,value);
+
+ /* listTypePop returns an object with its refcount incremented */
+ decrRefCount(value);
- /* Finally remove the element from the source list */
- listDelNode(srclist,ln);
- if (listLength(srclist) == 0) deleteKey(c->db,c->argv[1]);
+ /* Delete the source list when it is empty */
+ if (listTypeLength(sobj) == 0) dbDelete(c->db,c->argv[1]);
server.dirty++;
}
}
set = lookupKeyWrite(c->db,c->argv[1]);
if (set == NULL) {
set = createSetObject();
- dictAdd(c->db->dict,c->argv[1],set);
- incrRefCount(c->argv[1]);
+ dbAdd(c->db,c->argv[1],set);
} else {
if (set->type != REDIS_SET) {
addReply(c,shared.wrongtypeerr);
if (dictDelete(set->ptr,c->argv[2]) == DICT_OK) {
server.dirty++;
if (htNeedsResize(set->ptr)) dictResize(set->ptr);
- if (dictSize((dict*)set->ptr) == 0) deleteKey(c->db,c->argv[1]);
+ if (dictSize((dict*)set->ptr) == 0) dbDelete(c->db,c->argv[1]);
addReply(c,shared.cone);
} else {
addReply(c,shared.czero);
return;
}
if (dictSize((dict*)srcset->ptr) == 0 && srcset != dstset)
- deleteKey(c->db,c->argv[1]);
+ dbDelete(c->db,c->argv[1]);
server.dirty++;
/* Add the element to the destination set */
if (!dstset) {
dstset = createSetObject();
- dictAdd(c->db->dict,c->argv[2],dstset);
- incrRefCount(c->argv[2]);
+ dbAdd(c->db,c->argv[2],dstset);
}
if (dictAdd(dstset->ptr,c->argv[3],NULL) == DICT_OK)
incrRefCount(c->argv[3]);
addReplyBulk(c,ele);
dictDelete(set->ptr,ele);
if (htNeedsResize(set->ptr)) dictResize(set->ptr);
- if (dictSize((dict*)set->ptr) == 0) deleteKey(c->db,c->argv[1]);
+ if (dictSize((dict*)set->ptr) == 0) dbDelete(c->db,c->argv[1]);
server.dirty++;
}
}
if (!setobj) {
zfree(dv);
if (dstkey) {
- if (deleteKey(c->db,dstkey))
+ if (dbDelete(c->db,dstkey))
server.dirty++;
addReply(c,shared.czero);
} else {
if (dstkey) {
/* Store the resulting set into the target, if the intersection
* is not an empty set. */
- deleteKey(c->db,dstkey);
+ dbDelete(c->db,dstkey);
if (dictSize((dict*)dstset->ptr) > 0) {
- dictAdd(c->db->dict,dstkey,dstset);
- incrRefCount(dstkey);
- addReplyLong(c,dictSize((dict*)dstset->ptr));
+ dbAdd(c->db,dstkey,dstset);
+ addReplyLongLong(c,dictSize((dict*)dstset->ptr));
} else {
decrRefCount(dstset);
addReply(c,shared.czero);
} else {
/* If we have a target key where to store the resulting set
* create this key with the result set inside */
- deleteKey(c->db,dstkey);
+ dbDelete(c->db,dstkey);
if (dictSize((dict*)dstset->ptr) > 0) {
- dictAdd(c->db->dict,dstkey,dstset);
- incrRefCount(dstkey);
- addReplyLong(c,dictSize((dict*)dstset->ptr));
+ dbAdd(c->db,dstkey,dstset);
+ addReplyLongLong(c,dictSize((dict*)dstset->ptr));
} else {
decrRefCount(dstset);
addReply(c,shared.czero);
zskiplistNode *zn = zmalloc(sizeof(*zn));
zn->forward = zmalloc(sizeof(zskiplistNode*) * level);
- if (level > 0)
+ if (level > 1)
zn->span = zmalloc(sizeof(unsigned int) * (level - 1));
+ else
+ zn->span = NULL;
zn->score = score;
zn->obj = obj;
return zn;
/* We may have multiple elements with the same score, what we need
* is to find the element with both the right score and object. */
x = x->forward[0];
- if (x && score == x->score && compareStringObjects(x->obj,obj) == 0) {
+ if (x && score == x->score && equalStringObjects(x->obj,obj)) {
zslDeleteNode(zsl, x, update);
zslFreeNode(x);
return 1;
* Returns 0 when the element cannot be found, rank otherwise.
* Note that the rank is 1-based due to the span of zsl->header to the
* first element. */
-static unsigned long zslGetRank(zskiplist *zsl, double score, robj *o) {
+static unsigned long zslistTypeGetRank(zskiplist *zsl, double score, robj *o) {
zskiplistNode *x;
unsigned long rank = 0;
int i;
}
/* x might be equal to zsl->header, so test if obj is non-NULL */
- if (x->obj && compareStringObjects(x->obj,o) == 0) {
+ if (x->obj && equalStringObjects(x->obj,o)) {
return rank;
}
}
}
/* Finds an element by its rank. The rank argument needs to be 1-based. */
-zskiplistNode* zslGetElementByRank(zskiplist *zsl, unsigned long rank) {
+zskiplistNode* zslistTypeGetElementByRank(zskiplist *zsl, unsigned long rank) {
zskiplistNode *x;
unsigned long traversed = 0;
int i;
zset *zs;
double *score;
+ if (isnan(scoreval)) {
+ addReplySds(c,sdsnew("-ERR provide score is Not A Number (nan)\r\n"));
+ return;
+ }
+
zsetobj = lookupKeyWrite(c->db,key);
if (zsetobj == NULL) {
zsetobj = createZsetObject();
- dictAdd(c->db->dict,key,zsetobj);
- incrRefCount(key);
+ dbAdd(c->db,key,zsetobj);
} else {
if (zsetobj->type != REDIS_ZSET) {
addReply(c,shared.wrongtypeerr);
} else {
*score = scoreval;
}
+ if (isnan(*score)) {
+ addReplySds(c,
+ sdsnew("-ERR resulting score is Not A Number (nan)\r\n"));
+ zfree(score);
+ /* Note that we don't need to check if the zset may be empty and
+ * should be removed here, as we can only obtain Nan as score if
+ * there was already an element in the sorted set. */
+ return;
+ }
} else {
*score = scoreval;
}
/* Delete from the hash table */
dictDelete(zs->dict,c->argv[2]);
if (htNeedsResize(zs->dict)) dictResize(zs->dict);
- if (dictSize(zs->dict) == 0) deleteKey(c->db,c->argv[1]);
+ if (dictSize(zs->dict) == 0) dbDelete(c->db,c->argv[1]);
server.dirty++;
addReply(c,shared.cone);
}
zs = zsetobj->ptr;
deleted = zslDeleteRangeByScore(zs->zsl,min,max,zs->dict);
if (htNeedsResize(zs->dict)) dictResize(zs->dict);
- if (dictSize(zs->dict) == 0) deleteKey(c->db,c->argv[1]);
+ if (dictSize(zs->dict) == 0) dbDelete(c->db,c->argv[1]);
server.dirty += deleted;
- addReplyLong(c,deleted);
+ addReplyLongLong(c,deleted);
}
static void zremrangebyrankCommand(redisClient *c) {
* use 1-based rank */
deleted = zslDeleteRangeByRank(zs->zsl,start+1,end+1,zs->dict);
if (htNeedsResize(zs->dict)) dictResize(zs->dict);
- if (dictSize(zs->dict) == 0) deleteKey(c->db,c->argv[1]);
+ if (dictSize(zs->dict) == 0) dbDelete(c->db,c->argv[1]);
server.dirty += deleted;
- addReplyLong(c, deleted);
+ addReplyLongLong(c, deleted);
}
typedef struct {
#define REDIS_AGGR_SUM 1
#define REDIS_AGGR_MIN 2
#define REDIS_AGGR_MAX 3
+#define zunionInterDictValue(_e) (dictGetEntryVal(_e) == NULL ? 1.0 : *(double*)dictGetEntryVal(_e))
inline static void zunionInterAggregate(double *target, double val, int aggregate) {
if (aggregate == REDIS_AGGR_SUM) {
}
static void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) {
- int i, j, zsetnum;
+ int i, j, setnum;
int aggregate = REDIS_AGGR_SUM;
zsetopsrc *src;
robj *dstobj;
dictIterator *di;
dictEntry *de;
- /* expect zsetnum input keys to be given */
- zsetnum = atoi(c->argv[2]->ptr);
- if (zsetnum < 1) {
- addReplySds(c,sdsnew("-ERR at least 1 input key is needed for ZUNION/ZINTER\r\n"));
+ /* expect setnum input keys to be given */
+ setnum = atoi(c->argv[2]->ptr);
+ if (setnum < 1) {
+ addReplySds(c,sdsnew("-ERR at least 1 input key is needed for ZUNIONSTORE/ZINTERSTORE\r\n"));
return;
}
/* test if the expected number of keys would overflow */
- if (3+zsetnum > c->argc) {
+ if (3+setnum > c->argc) {
addReply(c,shared.syntaxerr);
return;
}
/* read keys to be used for input */
- src = zmalloc(sizeof(zsetopsrc) * zsetnum);
- for (i = 0, j = 3; i < zsetnum; i++, j++) {
- robj *zsetobj = lookupKeyWrite(c->db,c->argv[j]);
- if (!zsetobj) {
+ src = zmalloc(sizeof(zsetopsrc) * setnum);
+ for (i = 0, j = 3; i < setnum; i++, j++) {
+ robj *obj = lookupKeyWrite(c->db,c->argv[j]);
+ if (!obj) {
src[i].dict = NULL;
} else {
- if (zsetobj->type != REDIS_ZSET) {
+ if (obj->type == REDIS_ZSET) {
+ src[i].dict = ((zset*)obj->ptr)->dict;
+ } else if (obj->type == REDIS_SET) {
+ src[i].dict = (obj->ptr);
+ } else {
zfree(src);
addReply(c,shared.wrongtypeerr);
return;
}
- src[i].dict = ((zset*)zsetobj->ptr)->dict;
}
/* default all weights to 1 */
int remaining = c->argc - j;
while (remaining) {
- if (remaining >= (zsetnum + 1) && !strcasecmp(c->argv[j]->ptr,"weights")) {
+ if (remaining >= (setnum + 1) && !strcasecmp(c->argv[j]->ptr,"weights")) {
j++; remaining--;
- for (i = 0; i < zsetnum; i++, j++, remaining--) {
+ for (i = 0; i < setnum; i++, j++, remaining--) {
if (getDoubleFromObjectOrReply(c, c->argv[j], &src[i].weight, NULL) != REDIS_OK)
return;
}
/* sort sets from the smallest to largest, this will improve our
* algorithm's performance */
- qsort(src,zsetnum,sizeof(zsetopsrc), qsortCompareZsetopsrcByCardinality);
+ qsort(src,setnum,sizeof(zsetopsrc),qsortCompareZsetopsrcByCardinality);
dstobj = createZsetObject();
dstzset = dstobj->ptr;
di = dictGetIterator(src[0].dict);
while((de = dictNext(di)) != NULL) {
double *score = zmalloc(sizeof(double)), value;
- *score = src[0].weight * (*(double*)dictGetEntryVal(de));
+ *score = src[0].weight * zunionInterDictValue(de);
- for (j = 1; j < zsetnum; j++) {
+ for (j = 1; j < setnum; j++) {
dictEntry *other = dictFind(src[j].dict,dictGetEntryKey(de));
if (other) {
- value = src[j].weight * (*(double*)dictGetEntryVal(other));
+ value = src[j].weight * zunionInterDictValue(other);
zunionInterAggregate(score, value, aggregate);
} else {
break;
}
/* skip entry when not present in every source dict */
- if (j != zsetnum) {
+ if (j != setnum) {
zfree(score);
} else {
robj *o = dictGetEntryKey(de);
dictReleaseIterator(di);
}
} else if (op == REDIS_OP_UNION) {
- for (i = 0; i < zsetnum; i++) {
+ for (i = 0; i < setnum; i++) {
if (!src[i].dict) continue;
di = dictGetIterator(src[i].dict);
if (dictFind(dstzset->dict,dictGetEntryKey(de)) != NULL) continue;
double *score = zmalloc(sizeof(double)), value;
- *score = src[i].weight * (*(double*)dictGetEntryVal(de));
+ *score = src[i].weight * zunionInterDictValue(de);
/* because the zsets are sorted by size, its only possible
* for sets at larger indices to hold this entry */
- for (j = (i+1); j < zsetnum; j++) {
+ for (j = (i+1); j < setnum; j++) {
dictEntry *other = dictFind(src[j].dict,dictGetEntryKey(de));
if (other) {
- value = src[j].weight * (*(double*)dictGetEntryVal(other));
+ value = src[j].weight * zunionInterDictValue(other);
zunionInterAggregate(score, value, aggregate);
}
}
redisAssert(op == REDIS_OP_INTER || op == REDIS_OP_UNION);
}
- deleteKey(c->db,dstkey);
+ dbDelete(c->db,dstkey);
if (dstzset->zsl->length) {
- dictAdd(c->db->dict,dstkey,dstobj);
- incrRefCount(dstkey);
- addReplyLong(c, dstzset->zsl->length);
+ dbAdd(c->db,dstkey,dstobj);
+ addReplyLongLong(c, dstzset->zsl->length);
server.dirty++;
} else {
decrRefCount(dstobj);
zfree(src);
}
-static void zunionCommand(redisClient *c) {
+static void zunionstoreCommand(redisClient *c) {
zunionInterGenericCommand(c,c->argv[1], REDIS_OP_UNION);
}
-static void zinterCommand(redisClient *c) {
+static void zinterstoreCommand(redisClient *c) {
zunionInterGenericCommand(c,c->argv[1], REDIS_OP_INTER);
}
/* check if starting point is trivial, before searching
* the element in log(N) time */
if (reverse) {
- ln = start == 0 ? zsl->tail : zslGetElementByRank(zsl, llen-start);
+ ln = start == 0 ? zsl->tail : zslistTypeGetElementByRank(zsl, llen-start);
} else {
ln = start == 0 ?
- zsl->header->forward[0] : zslGetElementByRank(zsl, start+1);
+ zsl->header->forward[0] : zslistTypeGetElementByRank(zsl, start+1);
}
/* Return the result in form of a multi-bulk reply */
if (limit > 0) limit--;
}
if (justcount) {
- addReplyLong(c,(long)rangelen);
+ addReplyLongLong(c,(long)rangelen);
} else {
lenobj->ptr = sdscatprintf(sdsempty(),"*%lu\r\n",
withscores ? (rangelen*2) : rangelen);
}
score = dictGetEntryVal(de);
- rank = zslGetRank(zsl, *score, c->argv[2]);
+ rank = zslistTypeGetRank(zsl, *score, c->argv[2]);
if (rank) {
if (reverse) {
- addReplyLong(c, zsl->length - rank);
+ addReplyLongLong(c, zsl->length - rank);
} else {
- addReplyLong(c, rank-1);
+ addReplyLongLong(c, rank-1);
}
} else {
addReply(c,shared.nullbulk);
/* Check the length of a number of objects to see if we need to convert a
* zipmap to a real hash. Note that we only check string encoded objects
* as their string length can be queried in constant time. */
-static void hashTryConversion(robj *subject, robj **argv, int start, int end) {
+static void hashTypeTryConversion(robj *subject, robj **argv, int start, int end) {
int i;
if (subject->encoding != REDIS_ENCODING_ZIPMAP) return;
}
/* Encode given objects in-place when the hash uses a dict. */
-static void hashTryObjectEncoding(robj *subject, robj **o1, robj **o2) {
+static void hashTypeTryObjectEncoding(robj *subject, robj **o1, robj **o2) {
if (subject->encoding == REDIS_ENCODING_HT) {
if (o1) *o1 = tryObjectEncoding(*o1);
if (o2) *o2 = tryObjectEncoding(*o2);
/* Get the value from a hash identified by key. Returns either a string
* object or NULL if the value cannot be found. The refcount of the object
* is always increased by 1 when the value was found. */
-static robj *hashGet(robj *o, robj *key) {
+static robj *hashTypeGet(robj *o, robj *key) {
robj *value = NULL;
if (o->encoding == REDIS_ENCODING_ZIPMAP) {
unsigned char *v;
/* Test if the key exists in the given hash. Returns 1 if the key
* exists and 0 when it doesn't. */
-static int hashExists(robj *o, robj *key) {
+static int hashTypeExists(robj *o, robj *key) {
if (o->encoding == REDIS_ENCODING_ZIPMAP) {
key = getDecodedObject(key);
if (zipmapExists(o->ptr,key->ptr,sdslen(key->ptr))) {
/* Add an element, discard the old if the key already exists.
* Return 0 on insert and 1 on update. */
-static int hashSet(robj *o, robj *key, robj *value) {
+static int hashTypeSet(robj *o, robj *key, robj *value) {
int update = 0;
if (o->encoding == REDIS_ENCODING_ZIPMAP) {
key = getDecodedObject(key);
/* Delete an element from a hash.
* Return 1 on deleted and 0 on not found. */
-static int hashDelete(robj *o, robj *key) {
+static int hashTypeDelete(robj *o, robj *key) {
int deleted = 0;
if (o->encoding == REDIS_ENCODING_ZIPMAP) {
key = getDecodedObject(key);
}
/* Return the number of elements in a hash. */
-static unsigned long hashLength(robj *o) {
+static unsigned long hashTypeLength(robj *o) {
return (o->encoding == REDIS_ENCODING_ZIPMAP) ?
zipmapLen((unsigned char*)o->ptr) : dictSize((dict*)o->ptr);
}
dictIterator *di;
dictEntry *de;
-} hashIterator;
+} hashTypeIterator;
-static hashIterator *hashInitIterator(robj *subject) {
- hashIterator *hi = zmalloc(sizeof(hashIterator));
+static hashTypeIterator *hashTypeInitIterator(robj *subject) {
+ hashTypeIterator *hi = zmalloc(sizeof(hashTypeIterator));
hi->encoding = subject->encoding;
if (hi->encoding == REDIS_ENCODING_ZIPMAP) {
hi->zi = zipmapRewind(subject->ptr);
return hi;
}
-static void hashReleaseIterator(hashIterator *hi) {
+static void hashTypeReleaseIterator(hashTypeIterator *hi) {
if (hi->encoding == REDIS_ENCODING_HT) {
dictReleaseIterator(hi->di);
}
/* Move to the next entry in the hash. Return REDIS_OK when the next entry
* could be found and REDIS_ERR when the iterator reaches the end. */
-static int hashNext(hashIterator *hi) {
+static int hashTypeNext(hashTypeIterator *hi) {
if (hi->encoding == REDIS_ENCODING_ZIPMAP) {
if ((hi->zi = zipmapNext(hi->zi, &hi->zk, &hi->zklen,
&hi->zv, &hi->zvlen)) == NULL) return REDIS_ERR;
/* Get key or value object at current iteration position.
* This increases the refcount of the field object by 1. */
-static robj *hashCurrent(hashIterator *hi, int what) {
+static robj *hashTypeCurrent(hashTypeIterator *hi, int what) {
robj *o;
if (hi->encoding == REDIS_ENCODING_ZIPMAP) {
if (what & REDIS_HASH_KEY) {
return o;
}
-static robj *hashLookupWriteOrCreate(redisClient *c, robj *key) {
+static robj *hashTypeLookupWriteOrCreate(redisClient *c, robj *key) {
robj *o = lookupKeyWrite(c->db,key);
if (o == NULL) {
o = createHashObject();
- dictAdd(c->db->dict,key,o);
- incrRefCount(key);
+ dbAdd(c->db,key,o);
} else {
if (o->type != REDIS_HASH) {
addReply(c,shared.wrongtypeerr);
int update;
robj *o;
- if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return;
- hashTryConversion(o,c->argv,2,3);
- hashTryObjectEncoding(o,&c->argv[2], &c->argv[3]);
- update = hashSet(o,c->argv[2],c->argv[3]);
+ if ((o = hashTypeLookupWriteOrCreate(c,c->argv[1])) == NULL) return;
+ hashTypeTryConversion(o,c->argv,2,3);
+ hashTypeTryObjectEncoding(o,&c->argv[2], &c->argv[3]);
+ update = hashTypeSet(o,c->argv[2],c->argv[3]);
addReply(c, update ? shared.czero : shared.cone);
server.dirty++;
}
static void hsetnxCommand(redisClient *c) {
robj *o;
- if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return;
- hashTryConversion(o,c->argv,2,3);
+ if ((o = hashTypeLookupWriteOrCreate(c,c->argv[1])) == NULL) return;
+ hashTypeTryConversion(o,c->argv,2,3);
- if (hashExists(o, c->argv[2])) {
+ if (hashTypeExists(o, c->argv[2])) {
addReply(c, shared.czero);
} else {
- hashTryObjectEncoding(o,&c->argv[2], &c->argv[3]);
- hashSet(o,c->argv[2],c->argv[3]);
+ hashTypeTryObjectEncoding(o,&c->argv[2], &c->argv[3]);
+ hashTypeSet(o,c->argv[2],c->argv[3]);
addReply(c, shared.cone);
server.dirty++;
}
return;
}
- if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return;
- hashTryConversion(o,c->argv,2,c->argc-1);
+ if ((o = hashTypeLookupWriteOrCreate(c,c->argv[1])) == NULL) return;
+ hashTypeTryConversion(o,c->argv,2,c->argc-1);
for (i = 2; i < c->argc; i += 2) {
- hashTryObjectEncoding(o,&c->argv[i], &c->argv[i+1]);
- hashSet(o,c->argv[i],c->argv[i+1]);
+ hashTypeTryObjectEncoding(o,&c->argv[i], &c->argv[i+1]);
+ hashTypeSet(o,c->argv[i],c->argv[i+1]);
}
addReply(c, shared.ok);
server.dirty++;
robj *o, *current, *new;
if (getLongLongFromObjectOrReply(c,c->argv[3],&incr,NULL) != REDIS_OK) return;
- if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return;
- if ((current = hashGet(o,c->argv[2])) != NULL) {
+ if ((o = hashTypeLookupWriteOrCreate(c,c->argv[1])) == NULL) return;
+ if ((current = hashTypeGet(o,c->argv[2])) != NULL) {
if (getLongLongFromObjectOrReply(c,current,&value,
"hash value is not an integer") != REDIS_OK) {
decrRefCount(current);
value += incr;
new = createStringObjectFromLongLong(value);
- hashTryObjectEncoding(o,&c->argv[2],NULL);
- hashSet(o,c->argv[2],new);
+ hashTypeTryObjectEncoding(o,&c->argv[2],NULL);
+ hashTypeSet(o,c->argv[2],new);
decrRefCount(new);
addReplyLongLong(c,value);
server.dirty++;
if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL ||
checkType(c,o,REDIS_HASH)) return;
- if ((value = hashGet(o,c->argv[2])) != NULL) {
+ if ((value = hashTypeGet(o,c->argv[2])) != NULL) {
addReplyBulk(c,value);
decrRefCount(value);
} else {
* an empty hash. The reply should then be a series of NULLs. */
addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",c->argc-2));
for (i = 2; i < c->argc; i++) {
- if (o != NULL && (value = hashGet(o,c->argv[i])) != NULL) {
+ if (o != NULL && (value = hashTypeGet(o,c->argv[i])) != NULL) {
addReplyBulk(c,value);
decrRefCount(value);
} else {
if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL ||
checkType(c,o,REDIS_HASH)) return;
- if (hashDelete(o,c->argv[2])) {
- if (hashLength(o) == 0) deleteKey(c->db,c->argv[1]);
+ if (hashTypeDelete(o,c->argv[2])) {
+ if (hashTypeLength(o) == 0) dbDelete(c->db,c->argv[1]);
addReply(c,shared.cone);
server.dirty++;
} else {
if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL ||
checkType(c,o,REDIS_HASH)) return;
- addReplyUlong(c,hashLength(o));
+ addReplyUlong(c,hashTypeLength(o));
}
static void genericHgetallCommand(redisClient *c, int flags) {
robj *o, *lenobj, *obj;
unsigned long count = 0;
- hashIterator *hi;
+ hashTypeIterator *hi;
if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptymultibulk)) == NULL
|| checkType(c,o,REDIS_HASH)) return;
addReply(c,lenobj);
decrRefCount(lenobj);
- hi = hashInitIterator(o);
- while (hashNext(hi) != REDIS_ERR) {
+ hi = hashTypeInitIterator(o);
+ while (hashTypeNext(hi) != REDIS_ERR) {
if (flags & REDIS_HASH_KEY) {
- obj = hashCurrent(hi,REDIS_HASH_KEY);
+ obj = hashTypeCurrent(hi,REDIS_HASH_KEY);
addReplyBulk(c,obj);
decrRefCount(obj);
count++;
}
if (flags & REDIS_HASH_VALUE) {
- obj = hashCurrent(hi,REDIS_HASH_VALUE);
+ obj = hashTypeCurrent(hi,REDIS_HASH_VALUE);
addReplyBulk(c,obj);
decrRefCount(obj);
count++;
}
}
- hashReleaseIterator(hi);
+ hashTypeReleaseIterator(hi);
lenobj->ptr = sdscatprintf(sdsempty(),"*%lu\r\n",count);
}
if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL ||
checkType(c,o,REDIS_HASH)) return;
- addReply(c, hashExists(o,c->argv[2]) ? shared.cone : shared.czero);
+ addReply(c, hashTypeExists(o,c->argv[2]) ? shared.cone : shared.czero);
}
static void convertToRealHash(robj *o) {
static void flushdbCommand(redisClient *c) {
server.dirty += dictSize(c->db->dict);
+ touchWatchedKeysOnFlush(c->db->id);
dictEmpty(c->db->dict);
dictEmpty(c->db->expires);
addReply(c,shared.ok);
}
static void flushallCommand(redisClient *c) {
+ touchWatchedKeysOnFlush(-1);
server.dirty += emptyDb();
addReply(c,shared.ok);
if (server.bgsavechildpid != -1) {
/* Retrieve value from hash by the field name. This operation
* already increases the refcount of the returned object. */
initStaticStringObject(fieldobj,((char*)&fieldname)+(sizeof(long)*2));
- o = hashGet(o, &fieldobj);
+ o = hashTypeGet(o, &fieldobj);
} else {
if (o->type != REDIS_STRING) return NULL;
* is optimized for speed and a bit less for readability */
static void sortCommand(redisClient *c) {
list *operations;
- int outputlen = 0;
+ unsigned int outputlen = 0;
int desc = 0, alpha = 0;
int limit_start = 0, limit_count = -1, start, end;
int j, dontsort = 0, vectorlen;
/* Load the sorting vector with all the objects to sort */
switch(sortval->type) {
- case REDIS_LIST: vectorlen = listLength((list*)sortval->ptr); break;
+ case REDIS_LIST: vectorlen = listTypeLength(sortval); break;
case REDIS_SET: vectorlen = dictSize((dict*)sortval->ptr); break;
case REDIS_ZSET: vectorlen = dictSize(((zset*)sortval->ptr)->dict); break;
default: vectorlen = 0; redisPanic("Bad SORT type"); /* Avoid GCC warning */
j = 0;
if (sortval->type == REDIS_LIST) {
- list *list = sortval->ptr;
- listNode *ln;
- listIter li;
-
- listRewind(list,&li);
- while((ln = listNext(&li))) {
- robj *ele = ln->value;
- vector[j].obj = ele;
+ listTypeIterator *li = listTypeInitIterator(sortval,0,REDIS_TAIL);
+ listTypeEntry entry;
+ while(listTypeNext(li,&entry)) {
+ vector[j].obj = listTypeGet(&entry);
vector[j].u.score = 0;
vector[j].u.cmpobj = NULL;
j++;
}
+ listTypeReleaseIterator(li);
} else {
dict *set;
dictIterator *di;
}
}
} else {
- robj *listObject = createListObject();
- list *listPtr = (list*) listObject->ptr;
+ robj *sobj = createZiplistObject();
/* STORE option specified, set the sorting result as a List object */
for (j = start; j <= end; j++) {
listIter li;
if (!getop) {
- listAddNodeTail(listPtr,vector[j].obj);
- incrRefCount(vector[j].obj);
- }
- listRewind(operations,&li);
- while((ln = listNext(&li))) {
- redisSortOperation *sop = ln->value;
- robj *val = lookupKeyByPattern(c->db,sop->pattern,
- vector[j].obj);
+ listTypePush(sobj,vector[j].obj,REDIS_TAIL);
+ } else {
+ listRewind(operations,&li);
+ while((ln = listNext(&li))) {
+ redisSortOperation *sop = ln->value;
+ robj *val = lookupKeyByPattern(c->db,sop->pattern,
+ vector[j].obj);
- if (sop->type == REDIS_SORT_GET) {
- if (!val) {
- listAddNodeTail(listPtr,createStringObject("",0));
+ if (sop->type == REDIS_SORT_GET) {
+ if (!val) val = createStringObject("",0);
+
+ /* listTypePush does an incrRefCount, so we should take care
+ * care of the incremented refcount caused by either
+ * lookupKeyByPattern or createStringObject("",0) */
+ listTypePush(sobj,val,REDIS_TAIL);
+ decrRefCount(val);
} else {
- /* We should do a incrRefCount on val because it is
- * added to the list, but also a decrRefCount because
- * it is returned by lookupKeyByPattern. This results
- * in doing nothing at all. */
- listAddNodeTail(listPtr,val);
+ /* always fails */
+ redisAssert(sop->type == REDIS_SORT_GET);
}
- } else {
- redisAssert(sop->type == REDIS_SORT_GET); /* always fails */
}
}
}
- if (dictReplace(c->db->dict,storekey,listObject)) {
- incrRefCount(storekey);
- }
+ dbReplace(c->db,storekey,sobj);
/* Note: we add 1 because the DB is dirty anyway since even if the
* SORT result is empty a new key is set and maybe the old content
* replaced. */
}
/* Cleanup */
+ if (sortval->type == REDIS_LIST)
+ for (j = 0; j < vectorlen; j++)
+ decrRefCount(vector[j].obj);
decrRefCount(sortval);
listRelease(operations);
for (j = 0; j < vectorlen; j++) {
bytesToHuman(hmem,zmalloc_used_memory());
info = sdscatprintf(sdsempty(),
"redis_version:%s\r\n"
+ "redis_git_sha1:%s\r\n"
+ "redis_git_dirty:%d\r\n"
"arch_bits:%s\r\n"
"multiplexing_api:%s\r\n"
"process_id:%ld\r\n"
"total_connections_received:%lld\r\n"
"total_commands_processed:%lld\r\n"
"expired_keys:%lld\r\n"
- "hash_max_zipmap_entries:%ld\r\n"
- "hash_max_zipmap_value:%ld\r\n"
+ "hash_max_zipmap_entries:%zu\r\n"
+ "hash_max_zipmap_value:%zu\r\n"
"pubsub_channels:%ld\r\n"
"pubsub_patterns:%u\r\n"
"vm_enabled:%d\r\n"
"role:%s\r\n"
,REDIS_VERSION,
+ redisGitSHA1(),
+ strtol(redisGitDirty(),NULL,10) > 0,
(sizeof(long) == 8) ? "64" : "32",
aeGetApiName(),
(long) getpid(),
/* ================================= Expire ================================= */
static int removeExpire(redisDb *db, robj *key) {
- if (dictDelete(db->expires,key) == DICT_OK) {
+ /* An expire may only be removed if there is a corresponding entry in the
+ * main dict. Otherwise, the key will never be freed. */
+ redisAssert(dictFind(db->dict,key->ptr) != NULL);
+ if (dictDelete(db->expires,key->ptr) == DICT_OK) {
return 1;
} else {
return 0;
}
static int setExpire(redisDb *db, robj *key, time_t when) {
- if (dictAdd(db->expires,key,(void*)when) == DICT_ERR) {
+ dictEntry *de;
+
+ /* Reuse the sds from the main dict in the expire dict */
+ redisAssert((de = dictFind(db->dict,key->ptr)) != NULL);
+ if (dictAdd(db->expires,dictGetEntryKey(de),(void*)when) == DICT_ERR) {
return 0;
} else {
- incrRefCount(key);
return 1;
}
}
/* No expire? return ASAP */
if (dictSize(db->expires) == 0 ||
- (de = dictFind(db->expires,key)) == NULL) return -1;
+ (de = dictFind(db->expires,key->ptr)) == NULL) return -1;
+ /* The entry was found in the expire dict, this means it should also
+ * be present in the main dict (safety check). */
+ redisAssert(dictFind(db->dict,key->ptr) != NULL);
return (time_t) dictGetEntryVal(de);
}
static int expireIfNeeded(redisDb *db, robj *key) {
- time_t when;
- dictEntry *de;
-
- /* No expire? return ASAP */
- if (dictSize(db->expires) == 0 ||
- (de = dictFind(db->expires,key)) == NULL) return 0;
+ time_t when = getExpire(db,key);
+ if (when < 0) return 0;
- /* Lookup the expire */
- when = (time_t) dictGetEntryVal(de);
+ /* Return when this key has not expired */
if (time(NULL) <= when) return 0;
/* Delete the key */
- dictDelete(db->expires,key);
server.stat_expiredkeys++;
- return dictDelete(db->dict,key) == DICT_OK;
+ server.dirty++;
+ return dbDelete(db,key);
}
static int deleteIfVolatile(redisDb *db, robj *key) {
- dictEntry *de;
-
- /* No expire? return ASAP */
- if (dictSize(db->expires) == 0 ||
- (de = dictFind(db->expires,key)) == NULL) return 0;
+ if (getExpire(db,key) < 0) return 0;
/* Delete the key */
- server.dirty++;
server.stat_expiredkeys++;
- dictDelete(db->expires,key);
- return dictDelete(db->dict,key) == DICT_OK;
+ server.dirty++;
+ return dbDelete(db,key);
}
static void expireGenericCommand(redisClient *c, robj *key, robj *param, long offset) {
seconds -= offset;
- de = dictFind(c->db->dict,key);
+ de = dictFind(c->db->dict,key->ptr);
if (de == NULL) {
addReply(c,shared.czero);
return;
}
if (seconds <= 0) {
- if (deleteKey(c->db,key)) server.dirty++;
+ if (dbDelete(c->db,key)) server.dirty++;
addReply(c, shared.cone);
return;
} else {
}
static void multiCommand(redisClient *c) {
+ if (c->flags & REDIS_MULTI) {
+ addReplySds(c,sdsnew("-ERR MULTI calls can not be nested\r\n"));
+ return;
+ }
c->flags |= REDIS_MULTI;
addReply(c,shared.ok);
}
freeClientMultiState(c);
initClientMultiState(c);
c->flags &= (~REDIS_MULTI);
+ unwatchAllKeys(c);
addReply(c,shared.ok);
}
return;
}
+ /* Check if we need to abort the EXEC if some WATCHed key was touched.
+ * A failed EXEC will return a multi bulk nil object. */
+ if (c->flags & REDIS_DIRTY_CAS) {
+ freeClientMultiState(c);
+ initClientMultiState(c);
+ c->flags &= ~(REDIS_MULTI|REDIS_DIRTY_CAS);
+ unwatchAllKeys(c);
+ addReply(c,shared.nullmultibulk);
+ return;
+ }
+
/* Replicate a MULTI request now that we are sure the block is executed.
* This way we'll deliver the MULTI/..../EXEC block as a whole and
* both the AOF and the replication link will have the same consistency
execCommandReplicateMulti(c);
/* Exec all the queued commands */
+ unwatchAllKeys(c); /* Unwatch ASAP otherwise we'll waste CPU cycles */
orig_argv = c->argv;
orig_argc = c->argc;
addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",c->mstate.count));
c->argc = orig_argc;
freeClientMultiState(c);
initClientMultiState(c);
- c->flags &= (~REDIS_MULTI);
+ c->flags &= ~(REDIS_MULTI|REDIS_DIRTY_CAS);
/* Make sure the EXEC command is always replicated / AOF, since we
* always send the MULTI command (we can't know beforehand if the
* next operations will contain at least a modification to the DB). */
* empty we need to block. In order to do so we remove the notification for
* new data to read in the client socket (so that we'll not serve new
* requests if the blocking request is not served). Also we put the client
- * in a dictionary (db->blockingkeys) mapping keys to a list of clients
+ * in a dictionary (db->blocking_keys) mapping keys to a list of clients
* blocking for this keys.
* - If a PUSH operation against a key with blocked clients waiting is
* performed, we serve the first in the list: basically instead to push
list *l;
int j;
- c->blockingkeys = zmalloc(sizeof(robj*)*numkeys);
- c->blockingkeysnum = numkeys;
+ c->blocking_keys = zmalloc(sizeof(robj*)*numkeys);
+ c->blocking_keys_num = numkeys;
c->blockingto = timeout;
for (j = 0; j < numkeys; j++) {
/* Add the key in the client structure, to map clients -> keys */
- c->blockingkeys[j] = keys[j];
+ c->blocking_keys[j] = keys[j];
incrRefCount(keys[j]);
/* And in the other "side", to map keys -> clients */
- de = dictFind(c->db->blockingkeys,keys[j]);
+ de = dictFind(c->db->blocking_keys,keys[j]);
if (de == NULL) {
int retval;
/* For every key we take a list of clients blocked for it */
l = listCreate();
- retval = dictAdd(c->db->blockingkeys,keys[j],l);
+ retval = dictAdd(c->db->blocking_keys,keys[j],l);
incrRefCount(keys[j]);
assert(retval == DICT_OK);
} else {
list *l;
int j;
- assert(c->blockingkeys != NULL);
+ assert(c->blocking_keys != NULL);
/* The client may wait for multiple keys, so unblock it for every key. */
- for (j = 0; j < c->blockingkeysnum; j++) {
+ for (j = 0; j < c->blocking_keys_num; j++) {
/* Remove this client from the list of clients waiting for this key. */
- de = dictFind(c->db->blockingkeys,c->blockingkeys[j]);
+ de = dictFind(c->db->blocking_keys,c->blocking_keys[j]);
assert(de != NULL);
l = dictGetEntryVal(de);
listDelNode(l,listSearchKey(l,c));
/* If the list is empty we need to remove it to avoid wasting memory */
if (listLength(l) == 0)
- dictDelete(c->db->blockingkeys,c->blockingkeys[j]);
- decrRefCount(c->blockingkeys[j]);
+ dictDelete(c->db->blocking_keys,c->blocking_keys[j]);
+ decrRefCount(c->blocking_keys[j]);
}
/* Cleanup the client structure */
- zfree(c->blockingkeys);
- c->blockingkeys = NULL;
+ zfree(c->blocking_keys);
+ c->blocking_keys = NULL;
c->flags &= (~REDIS_BLOCKED);
server.blpop_blocked_clients--;
/* We want to process data if there is some command waiting
list *l;
listNode *ln;
- de = dictFind(c->db->blockingkeys,key);
+ de = dictFind(c->db->blocking_keys,key);
if (de == NULL) return 0;
l = dictGetEntryVal(de);
ln = listFirst(l);
addReply(c,shared.wrongtypeerr);
return;
} else {
- list *list = o->ptr;
- if (listLength(list) != 0) {
+ if (listTypeLength(o) != 0) {
/* If the list contains elements fall back to the usual
* non-blocking POP operation */
robj *argv[2], **orig_argv;
minttl = t;
}
}
- deleteKey(server.db+j,minkey);
+ dbDelete(server.db+j,minkey);
}
}
if (!freed) return; /* nothing to free... */
/* ============================== Append Only file ========================== */
-static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc) {
- sds buf = sdsempty();
- int j;
- ssize_t nwritten;
- time_t now;
- robj *tmpargv[3];
+/* Called when the user switches from "appendonly yes" to "appendonly no"
+ * at runtime using the CONFIG command. */
+static void stopAppendOnly(void) {
+ flushAppendOnlyFile();
+ aof_fsync(server.appendfd);
+ close(server.appendfd);
- /* The DB this command was targetting is not the same as the last command
- * we appendend. To issue a SELECT command is needed. */
- if (dictid != server.appendseldb) {
- char seldb[64];
+ server.appendfd = -1;
+ server.appendseldb = -1;
+ server.appendonly = 0;
+ /* rewrite operation in progress? kill it, wait child exit */
+ if (server.bgsavechildpid != -1) {
+ int statloc;
- snprintf(seldb,sizeof(seldb),"%d",dictid);
- buf = sdscatprintf(buf,"*2\r\n$6\r\nSELECT\r\n$%lu\r\n%s\r\n",
- (unsigned long)strlen(seldb),seldb);
- server.appendseldb = dictid;
+ if (kill(server.bgsavechildpid,SIGKILL) != -1)
+ wait3(&statloc,0,NULL);
+ /* reset the buffer accumulating changes while the child saves */
+ sdsfree(server.bgrewritebuf);
+ server.bgrewritebuf = sdsempty();
+ server.bgsavechildpid = -1;
}
+}
- /* "Fix" the argv vector if the command is EXPIRE. We want to translate
- * EXPIREs into EXPIREATs calls */
- if (cmd->proc == expireCommand) {
- long when;
-
- tmpargv[0] = createStringObject("EXPIREAT",8);
- tmpargv[1] = argv[1];
- incrRefCount(argv[1]);
- when = time(NULL)+strtol(argv[2]->ptr,NULL,10);
- tmpargv[2] = createObject(REDIS_STRING,
- sdscatprintf(sdsempty(),"%ld",when));
- argv = tmpargv;
+/* Called when the user switches from "appendonly no" to "appendonly yes"
+ * at runtime using the CONFIG command. */
+static int startAppendOnly(void) {
+ server.appendonly = 1;
+ server.lastfsync = time(NULL);
+ server.appendfd = open(server.appendfilename,O_WRONLY|O_APPEND|O_CREAT,0644);
+ if (server.appendfd == -1) {
+ redisLog(REDIS_WARNING,"Used tried to switch on AOF via CONFIG, but I can't open the AOF file: %s",strerror(errno));
+ return REDIS_ERR;
}
-
- /* Append the actual command */
- buf = sdscatprintf(buf,"*%d\r\n",argc);
- for (j = 0; j < argc; j++) {
- robj *o = argv[j];
-
- o = getDecodedObject(o);
- buf = sdscatprintf(buf,"$%lu\r\n",(unsigned long)sdslen(o->ptr));
- buf = sdscatlen(buf,o->ptr,sdslen(o->ptr));
- buf = sdscatlen(buf,"\r\n",2);
- decrRefCount(o);
+ if (rewriteAppendOnlyFileBackground() == REDIS_ERR) {
+ server.appendonly = 0;
+ close(server.appendfd);
+ redisLog(REDIS_WARNING,"Used tried to switch on AOF via CONFIG, I can't trigger a background AOF rewrite operation. Check the above logs for more info about the error.",strerror(errno));
+ return REDIS_ERR;
}
+ return REDIS_OK;
+}
- /* Free the objects from the modified argv for EXPIREAT */
- if (cmd->proc == expireCommand) {
- for (j = 0; j < 3; j++)
- decrRefCount(argv[j]);
- }
+/* Write the append only file buffer on disk.
+ *
+ * Since we are required to write the AOF before replying to the client,
+ * and the only way the client socket can get a write is entering when the
+ * the event loop, we accumulate all the AOF writes in a memory
+ * buffer and write it on disk using this function just before entering
+ * the event loop again. */
+static void flushAppendOnlyFile(void) {
+ time_t now;
+ ssize_t nwritten;
+
+ if (sdslen(server.aofbuf) == 0) return;
/* We want to perform a single write. This should be guaranteed atomic
* at least if the filesystem we are writing is a real physical one.
* While this will save us against the server being killed I don't think
* there is much to do about the whole server stopping for power problems
* or alike */
- nwritten = write(server.appendfd,buf,sdslen(buf));
- if (nwritten != (signed)sdslen(buf)) {
+ nwritten = write(server.appendfd,server.aofbuf,sdslen(server.aofbuf));
+ if (nwritten != (signed)sdslen(server.aofbuf)) {
/* Ooops, we are in troubles. The best thing to do for now is
- * to simply exit instead to give the illusion that everything is
+ * aborting instead of giving the illusion that everything is
* working as expected. */
if (nwritten == -1) {
redisLog(REDIS_WARNING,"Exiting on error writing to the append-only file: %s",strerror(errno));
}
exit(1);
}
- /* If a background append only file rewriting is in progress we want to
- * accumulate the differences between the child DB and the current one
- * in a buffer, so that when the child process will do its work we
- * can append the differences to the new append only file. */
- if (server.bgrewritechildpid != -1)
- server.bgrewritebuf = sdscatlen(server.bgrewritebuf,buf,sdslen(buf));
+ sdsfree(server.aofbuf);
+ server.aofbuf = sdsempty();
- sdsfree(buf);
+ /* Don't Fsync if no-appendfsync-on-rewrite is set to yes and we have
+ * childs performing heavy I/O on disk. */
+ if (server.no_appendfsync_on_rewrite &&
+ (server.bgrewritechildpid != -1 || server.bgsavechildpid != -1))
+ return;
+ /* Fsync if needed */
now = time(NULL);
if (server.appendfsync == APPENDFSYNC_ALWAYS ||
(server.appendfsync == APPENDFSYNC_EVERYSEC &&
}
}
+static sds catAppendOnlyGenericCommand(sds buf, int argc, robj **argv) {
+ int j;
+ buf = sdscatprintf(buf,"*%d\r\n",argc);
+ for (j = 0; j < argc; j++) {
+ robj *o = getDecodedObject(argv[j]);
+ buf = sdscatprintf(buf,"$%lu\r\n",(unsigned long)sdslen(o->ptr));
+ buf = sdscatlen(buf,o->ptr,sdslen(o->ptr));
+ buf = sdscatlen(buf,"\r\n",2);
+ decrRefCount(o);
+ }
+ return buf;
+}
+
+static sds catAppendOnlyExpireAtCommand(sds buf, robj *key, robj *seconds) {
+ int argc = 3;
+ long when;
+ robj *argv[3];
+
+ /* Make sure we can use strtol */
+ seconds = getDecodedObject(seconds);
+ when = time(NULL)+strtol(seconds->ptr,NULL,10);
+ decrRefCount(seconds);
+
+ argv[0] = createStringObject("EXPIREAT",8);
+ argv[1] = key;
+ argv[2] = createObject(REDIS_STRING,
+ sdscatprintf(sdsempty(),"%ld",when));
+ buf = catAppendOnlyGenericCommand(buf, argc, argv);
+ decrRefCount(argv[0]);
+ decrRefCount(argv[2]);
+ return buf;
+}
+
+static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc) {
+ sds buf = sdsempty();
+ robj *tmpargv[3];
+
+ /* The DB this command was targetting is not the same as the last command
+ * we appendend. To issue a SELECT command is needed. */
+ if (dictid != server.appendseldb) {
+ char seldb[64];
+
+ snprintf(seldb,sizeof(seldb),"%d",dictid);
+ buf = sdscatprintf(buf,"*2\r\n$6\r\nSELECT\r\n$%lu\r\n%s\r\n",
+ (unsigned long)strlen(seldb),seldb);
+ server.appendseldb = dictid;
+ }
+
+ if (cmd->proc == expireCommand) {
+ /* Translate EXPIRE into EXPIREAT */
+ buf = catAppendOnlyExpireAtCommand(buf,argv[1],argv[2]);
+ } else if (cmd->proc == setexCommand) {
+ /* Translate SETEX to SET and EXPIREAT */
+ tmpargv[0] = createStringObject("SET",3);
+ tmpargv[1] = argv[1];
+ tmpargv[2] = argv[3];
+ buf = catAppendOnlyGenericCommand(buf,3,tmpargv);
+ decrRefCount(tmpargv[0]);
+ buf = catAppendOnlyExpireAtCommand(buf,argv[1],argv[2]);
+ } else {
+ buf = catAppendOnlyGenericCommand(buf,argc,argv);
+ }
+
+ /* Append to the AOF buffer. This will be flushed on disk just before
+ * of re-entering the event loop, so before the client will get a
+ * positive reply about the operation performed. */
+ server.aofbuf = sdscatlen(server.aofbuf,buf,sdslen(buf));
+
+ /* If a background append only file rewriting is in progress we want to
+ * accumulate the differences between the child DB and the current one
+ * in a buffer, so that when the child process will do its work we
+ * can append the differences to the new append only file. */
+ if (server.bgrewritechildpid != -1)
+ server.bgrewritebuf = sdscatlen(server.bgrewritebuf,buf,sdslen(buf));
+
+ sdsfree(buf);
+}
+
/* In Redis commands are always executed in the context of a client, so in
* order to load the append only file we need to create a fake client. */
static struct redisClient *createFakeClient(void) {
struct redisClient *fakeClient;
FILE *fp = fopen(filename,"r");
struct redis_stat sb;
- unsigned long long loadedkeys = 0;
int appendonly = server.appendonly;
if (redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0)
char buf[128];
sds argsds;
struct redisCommand *cmd;
+ int force_swapout;
if (fgets(buf,sizeof(buf),fp) == NULL) {
if (feof(fp))
for (j = 0; j < argc; j++) decrRefCount(argv[j]);
zfree(argv);
/* Handle swapping while loading big datasets when VM is on */
- loadedkeys++;
- if (server.vm_enabled && (loadedkeys % 5000) == 0) {
+ force_swapout = 0;
+ if ((zmalloc_used_memory() - server.vm_max_memory) > 1024*1024*32)
+ force_swapout = 1;
+
+ if (server.vm_enabled && force_swapout) {
while (zmalloc_used_memory() > server.vm_max_memory) {
if (vmSwapOneObjectBlocking() == REDIS_ERR) break;
}
exit(1);
}
-/* Write an object into a file in the bulk format $<count>\r\n<payload>\r\n */
-static int fwriteBulkObject(FILE *fp, robj *obj) {
- char buf[128];
- int decrrc = 0;
-
- /* Avoid the incr/decr ref count business if possible to help
- * copy-on-write (we are often in a child process when this function
- * is called).
- * Also makes sure that key objects don't get incrRefCount-ed when VM
- * is enabled */
- if (obj->encoding != REDIS_ENCODING_RAW) {
- obj = getDecodedObject(obj);
- decrrc = 1;
- }
- snprintf(buf,sizeof(buf),"$%ld\r\n",(long)sdslen(obj->ptr));
- if (fwrite(buf,strlen(buf),1,fp) == 0) goto err;
- if (sdslen(obj->ptr) && fwrite(obj->ptr,sdslen(obj->ptr),1,fp) == 0)
- goto err;
- if (fwrite("\r\n",2,1,fp) == 0) goto err;
- if (decrrc) decrRefCount(obj);
- return 1;
-err:
- if (decrrc) decrRefCount(obj);
- return 0;
-}
-
/* Write binary-safe string into a file in the bulkformat
* $<count>\r\n<payload>\r\n */
static int fwriteBulkString(FILE *fp, char *s, unsigned long len) {
- char buf[128];
-
- snprintf(buf,sizeof(buf),"$%ld\r\n",(unsigned long)len);
- if (fwrite(buf,strlen(buf),1,fp) == 0) return 0;
- if (len && fwrite(s,len,1,fp) == 0) return 0;
+ char cbuf[128];
+ int clen;
+ cbuf[0] = '$';
+ clen = 1+ll2string(cbuf+1,sizeof(cbuf)-1,len);
+ cbuf[clen++] = '\r';
+ cbuf[clen++] = '\n';
+ if (fwrite(cbuf,clen,1,fp) == 0) return 0;
+ if (len > 0 && fwrite(s,len,1,fp) == 0) return 0;
if (fwrite("\r\n",2,1,fp) == 0) return 0;
return 1;
}
}
/* Write a long value in bulk format $<count>\r\n<payload>\r\n */
-static int fwriteBulkLong(FILE *fp, long l) {
- char buf[128], lbuf[128];
-
- snprintf(lbuf,sizeof(lbuf),"%ld\r\n",l);
- snprintf(buf,sizeof(buf),"$%lu\r\n",(unsigned long)strlen(lbuf)-2);
- if (fwrite(buf,strlen(buf),1,fp) == 0) return 0;
- if (fwrite(lbuf,strlen(lbuf),1,fp) == 0) return 0;
+static int fwriteBulkLongLong(FILE *fp, long long l) {
+ char bbuf[128], lbuf[128];
+ unsigned int blen, llen;
+ llen = ll2string(lbuf,32,l);
+ blen = snprintf(bbuf,sizeof(bbuf),"$%u\r\n%s\r\n",llen,lbuf);
+ if (fwrite(bbuf,blen,1,fp) == 0) return 0;
return 1;
}
+/* Delegate writing an object to writing a bulk string or bulk long long. */
+static int fwriteBulkObject(FILE *fp, robj *obj) {
+ /* Avoid using getDecodedObject to help copy-on-write (we are often
+ * in a child process when this function is called). */
+ if (obj->encoding == REDIS_ENCODING_INT) {
+ return fwriteBulkLongLong(fp,(long)obj->ptr);
+ } else if (obj->encoding == REDIS_ENCODING_RAW) {
+ return fwriteBulkString(fp,obj->ptr,sdslen(obj->ptr));
+ } else {
+ redisPanic("Unknown string encoding");
+ }
+}
+
/* Write a sequence of commands able to fully rebuild the dataset into
* "filename". Used both by REWRITEAOF and BGREWRITEAOF. */
static int rewriteAppendOnlyFile(char *filename) {
/* SELECT the new DB */
if (fwrite(selectcmd,sizeof(selectcmd)-1,1,fp) == 0) goto werr;
- if (fwriteBulkLong(fp,j) == 0) goto werr;
+ if (fwriteBulkLongLong(fp,j) == 0) goto werr;
/* Iterate this DB writing every entry */
while((de = dictNext(di)) != NULL) {
- robj *key, *o;
+ sds keystr = dictGetEntryKey(de);
+ robj key, *o;
time_t expiretime;
int swapped;
- key = dictGetEntryKey(de);
+ keystr = dictGetEntryKey(de);
+ o = dictGetEntryVal(de);
+ initStaticStringObject(key,keystr);
/* If the value for this key is swapped, load a preview in memory.
* We use a "swapped" flag to remember if we need to free the
* value object instead to just increment the ref count anyway
* in order to avoid copy-on-write of pages if we are forked() */
- if (!server.vm_enabled || key->storage == REDIS_VM_MEMORY ||
- key->storage == REDIS_VM_SWAPPING) {
- o = dictGetEntryVal(de);
+ if (!server.vm_enabled || o->storage == REDIS_VM_MEMORY ||
+ o->storage == REDIS_VM_SWAPPING) {
swapped = 0;
} else {
- o = vmPreviewObject(key);
+ o = vmPreviewObject(o);
swapped = 1;
}
- expiretime = getExpire(db,key);
+ expiretime = getExpire(db,&key);
/* Save the key and associated value */
if (o->type == REDIS_STRING) {
char cmd[]="*3\r\n$3\r\nSET\r\n";
if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
/* Key and value */
- if (fwriteBulkObject(fp,key) == 0) goto werr;
+ if (fwriteBulkObject(fp,&key) == 0) goto werr;
if (fwriteBulkObject(fp,o) == 0) goto werr;
} else if (o->type == REDIS_LIST) {
/* Emit the RPUSHes needed to rebuild the list */
- list *list = o->ptr;
- listNode *ln;
- listIter li;
+ char cmd[]="*3\r\n$5\r\nRPUSH\r\n";
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *zl = o->ptr;
+ unsigned char *p = ziplistIndex(zl,0);
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vlong;
+
+ while(ziplistGet(p,&vstr,&vlen,&vlong)) {
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulkObject(fp,&key) == 0) goto werr;
+ if (vstr) {
+ if (fwriteBulkString(fp,(char*)vstr,vlen) == 0)
+ goto werr;
+ } else {
+ if (fwriteBulkLongLong(fp,vlong) == 0)
+ goto werr;
+ }
+ p = ziplistNext(zl,p);
+ }
+ } else if (o->encoding == REDIS_ENCODING_LINKEDLIST) {
+ list *list = o->ptr;
+ listNode *ln;
+ listIter li;
- listRewind(list,&li);
- while((ln = listNext(&li))) {
- char cmd[]="*3\r\n$5\r\nRPUSH\r\n";
- robj *eleobj = listNodeValue(ln);
+ listRewind(list,&li);
+ while((ln = listNext(&li))) {
+ robj *eleobj = listNodeValue(ln);
- if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
- if (fwriteBulkObject(fp,key) == 0) goto werr;
- if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulkObject(fp,&key) == 0) goto werr;
+ if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
+ }
+ } else {
+ redisPanic("Unknown list encoding");
}
} else if (o->type == REDIS_SET) {
/* Emit the SADDs needed to rebuild the set */
robj *eleobj = dictGetEntryKey(de);
if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
- if (fwriteBulkObject(fp,key) == 0) goto werr;
+ if (fwriteBulkObject(fp,&key) == 0) goto werr;
if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
}
dictReleaseIterator(di);
double *score = dictGetEntryVal(de);
if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
- if (fwriteBulkObject(fp,key) == 0) goto werr;
+ if (fwriteBulkObject(fp,&key) == 0) goto werr;
if (fwriteBulkDouble(fp,*score) == 0) goto werr;
if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
}
while((p = zipmapNext(p,&field,&flen,&val,&vlen)) != NULL) {
if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
- if (fwriteBulkObject(fp,key) == 0) goto werr;
+ if (fwriteBulkObject(fp,&key) == 0) goto werr;
if (fwriteBulkString(fp,(char*)field,flen) == -1)
return -1;
if (fwriteBulkString(fp,(char*)val,vlen) == -1)
robj *val = dictGetEntryVal(de);
if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
- if (fwriteBulkObject(fp,key) == 0) goto werr;
+ if (fwriteBulkObject(fp,&key) == 0) goto werr;
if (fwriteBulkObject(fp,field) == -1) return -1;
if (fwriteBulkObject(fp,val) == -1) return -1;
}
/* If this key is already expired skip it */
if (expiretime < now) continue;
if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
- if (fwriteBulkObject(fp,key) == 0) goto werr;
- if (fwriteBulkLong(fp,expiretime) == 0) goto werr;
+ if (fwriteBulkObject(fp,&key) == 0) goto werr;
+ if (fwriteBulkLongLong(fp,expiretime) == 0) goto werr;
}
if (swapped) decrRefCount(o);
}
/* Make sure data will not remain on the OS's output buffers */
fflush(fp);
- fsync(fileno(fp));
+ aof_fsync(fileno(fp));
fclose(fp);
/* Use RENAME to make sure the DB file is changed atomically only
/* =================== Virtual Memory - Blocking Side ====================== */
-/* substitute the first occurrence of '%p' with the process pid in the
- * swap file name. */
-static void expandVmSwapFilename(void) {
- char *p = strstr(server.vm_swap_file,"%p");
- sds new;
+/* Create a VM pointer object. This kind of objects are used in place of
+ * values in the key -> value hash table, for swapped out objects. */
+static vmpointer *createVmPointer(int vtype) {
+ vmpointer *vp = zmalloc(sizeof(vmpointer));
- if (!p) return;
- new = sdsempty();
- *p = '\0';
- new = sdscat(new,server.vm_swap_file);
- new = sdscatprintf(new,"%ld",(long) getpid());
- new = sdscat(new,p+2);
- zfree(server.vm_swap_file);
- server.vm_swap_file = new;
+ vp->type = REDIS_VMPOINTER;
+ vp->storage = REDIS_VM_SWAPPED;
+ vp->vtype = vtype;
+ return vp;
}
static void vmInit(void) {
off_t totsize;
int pipefds[2];
size_t stacksize;
+ struct flock fl;
if (server.vm_max_threads != 0)
zmalloc_enable_thread_safeness(); /* we need thread safe zmalloc() */
- expandVmSwapFilename();
redisLog(REDIS_NOTICE,"Using '%s' as swap file",server.vm_swap_file);
+ /* Try to open the old swap file, otherwise create it */
if ((server.vm_fp = fopen(server.vm_swap_file,"r+b")) == NULL) {
server.vm_fp = fopen(server.vm_swap_file,"w+b");
}
if (server.vm_fp == NULL) {
redisLog(REDIS_WARNING,
- "Impossible to open the swap file: %s. Exiting.",
+ "Can't open the swap file: %s. Exiting.",
strerror(errno));
exit(1);
}
server.vm_fd = fileno(server.vm_fp);
+ /* Lock the swap file for writing, this is useful in order to avoid
+ * another instance to use the same swap file for a config error. */
+ fl.l_type = F_WRLCK;
+ fl.l_whence = SEEK_SET;
+ fl.l_start = fl.l_len = 0;
+ if (fcntl(server.vm_fd,F_SETLK,&fl) == -1) {
+ redisLog(REDIS_WARNING,
+ "Can't lock the swap file at '%s': %s. Make sure it is not used by another Redis instance.", server.vm_swap_file, strerror(errno));
+ exit(1);
+ }
+ /* Initialize */
server.vm_next_page = 0;
server.vm_near_pages = 0;
server.vm_stats_used_pages = 0;
return REDIS_OK;
}
-/* Swap the 'val' object relative to 'key' into disk. Store all the information
- * needed to later retrieve the object into the key object.
+/* Transfers the 'val' object to disk. Store all the information
+ * a 'vmpointer' object containing all the information needed to load the
+ * object back later is returned.
+ *
* If we can't find enough contiguous empty pages to swap the object on disk
- * REDIS_ERR is returned. */
-static int vmSwapObjectBlocking(robj *key, robj *val) {
+ * NULL is returned. */
+static vmpointer *vmSwapObjectBlocking(robj *val) {
off_t pages = rdbSavedObjectPages(val,NULL);
off_t page;
+ vmpointer *vp;
- assert(key->storage == REDIS_VM_MEMORY);
- assert(key->refcount == 1);
- if (vmFindContiguousPages(&page,pages) == REDIS_ERR) return REDIS_ERR;
- if (vmWriteObjectOnSwap(val,page) == REDIS_ERR) return REDIS_ERR;
- key->vm.page = page;
- key->vm.usedpages = pages;
- key->storage = REDIS_VM_SWAPPED;
- key->vtype = val->type;
+ assert(val->storage == REDIS_VM_MEMORY);
+ assert(val->refcount == 1);
+ if (vmFindContiguousPages(&page,pages) == REDIS_ERR) return NULL;
+ if (vmWriteObjectOnSwap(val,page) == REDIS_ERR) return NULL;
+
+ vp = createVmPointer(val->type);
+ vp->page = page;
+ vp->usedpages = pages;
decrRefCount(val); /* Deallocate the object from memory. */
vmMarkPagesUsed(page,pages);
- redisLog(REDIS_DEBUG,"VM: object %s swapped out at %lld (%lld pages)",
- (unsigned char*) key->ptr,
+ redisLog(REDIS_DEBUG,"VM: object %p swapped out at %lld (%lld pages)",
+ (void*) val,
(unsigned long long) page, (unsigned long long) pages);
server.vm_stats_swapped_objects++;
server.vm_stats_swapouts++;
- return REDIS_OK;
+ return vp;
}
static robj *vmReadObjectFromSwap(off_t page, int type) {
return o;
}
-/* Load the value object relative to the 'key' object from swap to memory.
+/* Load the specified object from swap to memory.
* The newly allocated object is returned.
*
* If preview is true the unserialized object is returned to the caller but
- * no changes are made to the key object, nor the pages are marked as freed */
-static robj *vmGenericLoadObject(robj *key, int preview) {
+ * the pages are not marked as freed, nor the vp object is freed. */
+static robj *vmGenericLoadObject(vmpointer *vp, int preview) {
robj *val;
- redisAssert(key->storage == REDIS_VM_SWAPPED || key->storage == REDIS_VM_LOADING);
- val = vmReadObjectFromSwap(key->vm.page,key->vtype);
+ redisAssert(vp->type == REDIS_VMPOINTER &&
+ (vp->storage == REDIS_VM_SWAPPED || vp->storage == REDIS_VM_LOADING));
+ val = vmReadObjectFromSwap(vp->page,vp->vtype);
if (!preview) {
- key->storage = REDIS_VM_MEMORY;
- key->vm.atime = server.unixtime;
- vmMarkPagesFree(key->vm.page,key->vm.usedpages);
- redisLog(REDIS_DEBUG, "VM: object %s loaded from disk",
- (unsigned char*) key->ptr);
+ redisLog(REDIS_DEBUG, "VM: object %p loaded from disk", (void*)vp);
+ vmMarkPagesFree(vp->page,vp->usedpages);
+ zfree(vp);
server.vm_stats_swapped_objects--;
} else {
- redisLog(REDIS_DEBUG, "VM: object %s previewed from disk",
- (unsigned char*) key->ptr);
+ redisLog(REDIS_DEBUG, "VM: object %p previewed from disk", (void*)vp);
}
server.vm_stats_swapins++;
return val;
}
-/* Plain object loading, from swap to memory */
-static robj *vmLoadObject(robj *key) {
+/* Plain object loading, from swap to memory.
+ *
+ * 'o' is actually a redisVmPointer structure that will be freed by the call.
+ * The return value is the loaded object. */
+static robj *vmLoadObject(robj *o) {
/* If we are loading the object in background, stop it, we
* need to load this object synchronously ASAP. */
- if (key->storage == REDIS_VM_LOADING)
- vmCancelThreadedIOJob(key);
- return vmGenericLoadObject(key,0);
+ if (o->storage == REDIS_VM_LOADING)
+ vmCancelThreadedIOJob(o);
+ return vmGenericLoadObject((vmpointer*)o,0);
}
/* Just load the value on disk, without to modify the key.
* This is useful when we want to perform some operation on the value
* without to really bring it from swap to memory, like while saving the
* dataset or rewriting the append only log. */
-static robj *vmPreviewObject(robj *key) {
- return vmGenericLoadObject(key,1);
+static robj *vmPreviewObject(robj *o) {
+ return vmGenericLoadObject((vmpointer*)o,1);
}
/* How a good candidate is this object for swapping?
* proportionally, this is why we use the logarithm. This algorithm is
* just a first try and will probably be tuned later. */
static double computeObjectSwappability(robj *o) {
- time_t age = server.unixtime - o->vm.atime;
- long asize = 0;
+ /* actual age can be >= minage, but not < minage. As we use wrapping
+ * 21 bit clocks with minutes resolution for the LRU. */
+ time_t minage = abs(server.lruclock - o->lru);
+ long asize = 0, elesize;
+ robj *ele;
list *l;
+ listNode *ln;
dict *d;
struct dictEntry *de;
int z;
- if (age <= 0) return 0;
+ if (minage <= 0) return 0;
switch(o->type) {
case REDIS_STRING:
if (o->encoding != REDIS_ENCODING_RAW) {
}
break;
case REDIS_LIST:
- l = o->ptr;
- listNode *ln = listFirst(l);
-
- asize = sizeof(list);
- if (ln) {
- robj *ele = ln->value;
- long elesize;
-
- elesize = (ele->encoding == REDIS_ENCODING_RAW) ?
- (sizeof(*o)+sdslen(ele->ptr)) :
- sizeof(*o);
- asize += (sizeof(listNode)+elesize)*listLength(l);
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ asize = sizeof(*o)+ziplistSize(o->ptr);
+ } else {
+ l = o->ptr;
+ ln = listFirst(l);
+ asize = sizeof(list);
+ if (ln) {
+ ele = ln->value;
+ elesize = (ele->encoding == REDIS_ENCODING_RAW) ?
+ (sizeof(*o)+sdslen(ele->ptr)) : sizeof(*o);
+ asize += (sizeof(listNode)+elesize)*listLength(l);
+ }
}
break;
case REDIS_SET:
asize = sizeof(dict)+(sizeof(struct dictEntry*)*dictSlots(d));
if (z) asize += sizeof(zset)-sizeof(dict);
if (dictSize(d)) {
- long elesize;
- robj *ele;
-
de = dictGetRandomKey(d);
ele = dictGetEntryKey(de);
elesize = (ele->encoding == REDIS_ENCODING_RAW) ?
- (sizeof(*o)+sdslen(ele->ptr)) :
- sizeof(*o);
+ (sizeof(*o)+sdslen(ele->ptr)) : sizeof(*o);
asize += (sizeof(struct dictEntry)+elesize)*dictSize(d);
if (z) asize += sizeof(zskiplistNode)*dictSize(d);
}
d = o->ptr;
asize = sizeof(dict)+(sizeof(struct dictEntry*)*dictSlots(d));
if (dictSize(d)) {
- long elesize;
- robj *ele;
-
de = dictGetRandomKey(d);
ele = dictGetEntryKey(de);
elesize = (ele->encoding == REDIS_ENCODING_RAW) ?
- (sizeof(*o)+sdslen(ele->ptr)) :
- sizeof(*o);
+ (sizeof(*o)+sdslen(ele->ptr)) : sizeof(*o);
ele = dictGetEntryVal(de);
elesize = (ele->encoding == REDIS_ENCODING_RAW) ?
- (sizeof(*o)+sdslen(ele->ptr)) :
- sizeof(*o);
+ (sizeof(*o)+sdslen(ele->ptr)) : sizeof(*o);
asize += (sizeof(struct dictEntry)+elesize)*dictSize(d);
}
}
break;
}
- return (double)age*log(1+asize);
+ return (double)minage*log(1+asize);
}
/* Try to swap an object that's a good candidate for swapping.
struct dictEntry *best = NULL;
double best_swappability = 0;
redisDb *best_db = NULL;
- robj *key, *val;
+ robj *val;
+ sds key;
for (j = 0; j < server.dbnum; j++) {
redisDb *db = server.db+j;
if (maxtries) maxtries--;
de = dictGetRandomKey(db->dict);
- key = dictGetEntryKey(de);
val = dictGetEntryVal(de);
/* Only swap objects that are currently in memory.
*
- * Also don't swap shared objects if threaded VM is on, as we
- * try to ensure that the main thread does not touch the
+ * Also don't swap shared objects: not a good idea in general and
+ * we need to ensure that the main thread does not touch the
* object while the I/O thread is using it, but we can't
* control other keys without adding additional mutex. */
- if (key->storage != REDIS_VM_MEMORY ||
- (server.vm_max_threads != 0 && val->refcount != 1)) {
+ if (val->storage != REDIS_VM_MEMORY || val->refcount != 1) {
if (maxtries) i--; /* don't count this try */
continue;
}
val = dictGetEntryVal(best);
redisLog(REDIS_DEBUG,"Key with best swappability: %s, %f",
- key->ptr, best_swappability);
+ key, best_swappability);
- /* Unshare the key if needed */
- if (key->refcount > 1) {
- robj *newkey = dupStringObject(key);
- decrRefCount(key);
- key = dictGetEntryKey(best) = newkey;
- }
/* Swap it */
if (usethreads) {
- vmSwapObjectThreaded(key,val,best_db);
+ robj *keyobj = createStringObject(key,sdslen(key));
+ vmSwapObjectThreaded(keyobj,val,best_db);
+ decrRefCount(keyobj);
return REDIS_OK;
} else {
- if (vmSwapObjectBlocking(key,val) == REDIS_OK) {
- dictGetEntryVal(best) = NULL;
+ vmpointer *vp;
+
+ if ((vp = vmSwapObjectBlocking(val)) != NULL) {
+ dictGetEntryVal(best) = vp;
return REDIS_OK;
} else {
return REDIS_ERR;
return (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1);
}
-/* Delete a key if swapped. Returns 1 if the key was found, was swapped
- * and was deleted. Otherwise 0 is returned. */
-static int deleteIfSwapped(redisDb *db, robj *key) {
- dictEntry *de;
- robj *foundkey;
-
- if ((de = dictFind(db->dict,key)) == NULL) return 0;
- foundkey = dictGetEntryKey(de);
- if (foundkey->storage == REDIS_VM_MEMORY) return 0;
- deleteKey(db,key);
- return 1;
-}
-
/* =================== Virtual Memory - Threaded I/O ======================= */
static void freeIOJob(iojob *j) {
if ((j->type == REDIS_IOJOB_PREPARE_SWAP ||
j->type == REDIS_IOJOB_DO_SWAP ||
j->type == REDIS_IOJOB_LOAD) && j->val != NULL)
+ {
+ /* we fix the storage type, otherwise decrRefCount() will try to
+ * kill the I/O thread Job (that does no longer exists). */
+ if (j->val->storage == REDIS_VM_SWAPPING)
+ j->val->storage = REDIS_VM_MEMORY;
decrRefCount(j->val);
- /* We don't decrRefCount the j->key field as we did't incremented
- * the count creating IO Jobs. This is because the key field here is
- * just used as an indentifier and if a key is removed the Job should
- * never be touched again. */
+ }
+ decrRefCount(j->key);
zfree(j);
}
while((retval = read(fd,buf,1)) == 1) {
iojob *j;
listNode *ln;
- robj *key;
struct dictEntry *de;
redisLog(REDIS_DEBUG,"Processing I/O completed job");
}
/* Post process it in the main thread, as there are things we
* can do just here to avoid race conditions and/or invasive locks */
- redisLog(REDIS_DEBUG,"Job %p type: %d, key at %p (%s) refcount: %d\n", (void*) j, j->type, (void*)j->key, (char*)j->key->ptr, j->key->refcount);
- de = dictFind(j->db->dict,j->key);
- assert(de != NULL);
- key = dictGetEntryKey(de);
+ redisLog(REDIS_DEBUG,"COMPLETED Job type: %d, ID %p, key: %s", j->type, (void*)j->id, (unsigned char*)j->key->ptr);
+ de = dictFind(j->db->dict,j->key->ptr);
+ redisAssert(de != NULL);
if (j->type == REDIS_IOJOB_LOAD) {
redisDb *db;
+ vmpointer *vp = dictGetEntryVal(de);
/* Key loaded, bring it at home */
- key->storage = REDIS_VM_MEMORY;
- key->vm.atime = server.unixtime;
- vmMarkPagesFree(key->vm.page,key->vm.usedpages);
+ vmMarkPagesFree(vp->page,vp->usedpages);
redisLog(REDIS_DEBUG, "VM: object %s loaded from disk (threaded)",
- (unsigned char*) key->ptr);
+ (unsigned char*) j->key->ptr);
server.vm_stats_swapped_objects--;
server.vm_stats_swapins++;
dictGetEntryVal(de) = j->val;
incrRefCount(j->val);
db = j->db;
- freeIOJob(j);
/* Handle clients waiting for this key to be loaded. */
- handleClientsBlockedOnSwappedKey(db,key);
+ handleClientsBlockedOnSwappedKey(db,j->key);
+ freeIOJob(j);
+ zfree(vp);
} else if (j->type == REDIS_IOJOB_PREPARE_SWAP) {
/* Now we know the amount of pages required to swap this object.
* Let's find some space for it, and queue this task again
{
/* Ooops... no space or we can't swap as there is
* a fork()ed Redis trying to save stuff on disk. */
+ j->val->storage = REDIS_VM_MEMORY; /* undo operation */
freeIOJob(j);
- key->storage = REDIS_VM_MEMORY; /* undo operation */
} else {
/* Note that we need to mark this pages as used now,
* if the job will be canceled, we'll mark them as freed
unlockThreadedIO();
}
} else if (j->type == REDIS_IOJOB_DO_SWAP) {
- robj *val;
+ vmpointer *vp;
/* Key swapped. We can finally free some memory. */
- if (key->storage != REDIS_VM_SWAPPING) {
- printf("key->storage: %d\n",key->storage);
- printf("key->name: %s\n",(char*)key->ptr);
- printf("key->refcount: %d\n",key->refcount);
+ if (j->val->storage != REDIS_VM_SWAPPING) {
+ vmpointer *vp = (vmpointer*) j->id;
+ printf("storage: %d\n",vp->storage);
+ printf("key->name: %s\n",(char*)j->key->ptr);
printf("val: %p\n",(void*)j->val);
printf("val->type: %d\n",j->val->type);
printf("val->ptr: %s\n",(char*)j->val->ptr);
}
- redisAssert(key->storage == REDIS_VM_SWAPPING);
- val = dictGetEntryVal(de);
- key->vm.page = j->page;
- key->vm.usedpages = j->pages;
- key->storage = REDIS_VM_SWAPPED;
- key->vtype = j->val->type;
- decrRefCount(val); /* Deallocate the object from memory. */
- dictGetEntryVal(de) = NULL;
+ redisAssert(j->val->storage == REDIS_VM_SWAPPING);
+ vp = createVmPointer(j->val->type);
+ vp->page = j->page;
+ vp->usedpages = j->pages;
+ dictGetEntryVal(de) = vp;
+ /* Fix the storage otherwise decrRefCount will attempt to
+ * remove the associated I/O job */
+ j->val->storage = REDIS_VM_MEMORY;
+ decrRefCount(j->val);
redisLog(REDIS_DEBUG,
"VM: object %s swapped out at %lld (%lld pages) (threaded)",
- (unsigned char*) key->ptr,
+ (unsigned char*) j->key->ptr,
(unsigned long long) j->page, (unsigned long long) j->pages);
server.vm_stats_swapped_objects++;
server.vm_stats_swapouts++;
assert(o->storage == REDIS_VM_LOADING || o->storage == REDIS_VM_SWAPPING);
again:
lockThreadedIO();
- /* Search for a matching key in one of the queues */
+ /* Search for a matching object in one of the queues */
for (i = 0; i < 3; i++) {
listNode *ln;
listIter li;
iojob *job = ln->value;
if (job->canceled) continue; /* Skip this, already canceled. */
- if (job->key == o) {
- redisLog(REDIS_DEBUG,"*** CANCELED %p (%s) (type %d) (LIST ID %d)\n",
- (void*)job, (char*)o->ptr, job->type, i);
+ if (job->id == o) {
+ redisLog(REDIS_DEBUG,"*** CANCELED %p (key %s) (type %d) (LIST ID %d)\n",
+ (void*)job, (char*)job->key->ptr, job->type, i);
/* Mark the pages as free since the swap didn't happened
* or happened but is now discarded. */
if (i != 1 && job->type == REDIS_IOJOB_DO_SWAP)
else if (o->storage == REDIS_VM_SWAPPING)
o->storage = REDIS_VM_MEMORY;
unlockThreadedIO();
+ redisLog(REDIS_DEBUG,"*** DONE");
return;
}
}
}
unlockThreadedIO();
- assert(1 != 1); /* We should never reach this */
+ printf("Not found: %p\n", (void*)o);
+ redisAssert(1 != 1); /* We should never reach this */
}
static void *IOThreadEntryPoint(void *arg) {
/* Process the Job */
if (j->type == REDIS_IOJOB_LOAD) {
- j->val = vmReadObjectFromSwap(j->page,j->key->vtype);
+ vmpointer *vp = (vmpointer*)j->id;
+ j->val = vmReadObjectFromSwap(j->page,vp->vtype);
} else if (j->type == REDIS_IOJOB_PREPARE_SWAP) {
FILE *fp = fopen("/dev/null","w+");
j->pages = rdbSavedObjectPages(j->val,fp);
static int vmSwapObjectThreaded(robj *key, robj *val, redisDb *db) {
iojob *j;
- assert(key->storage == REDIS_VM_MEMORY);
- assert(key->refcount == 1);
-
j = zmalloc(sizeof(*j));
j->type = REDIS_IOJOB_PREPARE_SWAP;
j->db = db;
j->key = key;
- j->val = val;
+ incrRefCount(key);
+ j->id = j->val = val;
incrRefCount(val);
j->canceled = 0;
j->thread = (pthread_t) -1;
- key->storage = REDIS_VM_SWAPPING;
+ val->storage = REDIS_VM_SWAPPING;
lockThreadedIO();
queueIOJob(j);
/* If the key does not exist or is already in RAM we don't need to
* block the client at all. */
- de = dictFind(c->db->dict,key);
+ de = dictFind(c->db->dict,key->ptr);
if (de == NULL) return 0;
- o = dictGetEntryKey(de);
+ o = dictGetEntryVal(de);
if (o->storage == REDIS_VM_MEMORY) {
return 0;
} else if (o->storage == REDIS_VM_SWAPPING) {
/* Are we already loading the key from disk? If not create a job */
if (o->storage == REDIS_VM_SWAPPED) {
iojob *j;
+ vmpointer *vp = (vmpointer*)o;
o->storage = REDIS_VM_LOADING;
j = zmalloc(sizeof(*j));
j->type = REDIS_IOJOB_LOAD;
j->db = c->db;
- j->key = o;
- j->key->vtype = o->vtype;
- j->page = o->vm.page;
+ j->id = (robj*)vp;
+ j->key = key;
+ incrRefCount(key);
+ j->page = vp->page;
j->val = NULL;
j->canceled = 0;
j->thread = (pthread_t) -1;
return 1;
}
-/* Preload keys needed for the ZUNION and ZINTER commands. */
-static void zunionInterBlockClientOnSwappedKeys(redisClient *c) {
+/* Preload keys for any command with first, last and step values for
+ * the command keys prototype, as defined in the command table. */
+static void waitForMultipleSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv) {
+ int j, last;
+ if (cmd->vm_firstkey == 0) return;
+ last = cmd->vm_lastkey;
+ if (last < 0) last = argc+last;
+ for (j = cmd->vm_firstkey; j <= last; j += cmd->vm_keystep) {
+ redisAssert(j < argc);
+ waitForSwappedKey(c,argv[j]);
+ }
+}
+
+/* Preload keys needed for the ZUNIONSTORE and ZINTERSTORE commands.
+ * Note that the number of keys to preload is user-defined, so we need to
+ * apply a sanity check against argc. */
+static void zunionInterBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv) {
int i, num;
- num = atoi(c->argv[2]->ptr);
+ REDIS_NOTUSED(cmd);
+
+ num = atoi(argv[2]->ptr);
+ if (num > (argc-3)) return;
for (i = 0; i < num; i++) {
- waitForSwappedKey(c,c->argv[3+i]);
+ waitForSwappedKey(c,argv[3+i]);
+ }
+}
+
+/* Preload keys needed to execute the entire MULTI/EXEC block.
+ *
+ * This function is called by blockClientOnSwappedKeys when EXEC is issued,
+ * and will block the client when any command requires a swapped out value. */
+static void execBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv) {
+ int i, margc;
+ struct redisCommand *mcmd;
+ robj **margv;
+ REDIS_NOTUSED(cmd);
+ REDIS_NOTUSED(argc);
+ REDIS_NOTUSED(argv);
+
+ if (!(c->flags & REDIS_MULTI)) return;
+ for (i = 0; i < c->mstate.count; i++) {
+ mcmd = c->mstate.commands[i].cmd;
+ margc = c->mstate.commands[i].argc;
+ margv = c->mstate.commands[i].argv;
+
+ if (mcmd->vm_preload_proc != NULL) {
+ mcmd->vm_preload_proc(c,mcmd,margc,margv);
+ } else {
+ waitForMultipleSwappedKeys(c,mcmd,margc,margv);
+ }
}
}
*
* Return 1 if the client is marked as blocked, 0 if the client can
* continue as the keys it is going to access appear to be in memory. */
-static int blockClientOnSwappedKeys(struct redisCommand *cmd, redisClient *c) {
- int j, last;
-
+static int blockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd) {
if (cmd->vm_preload_proc != NULL) {
- cmd->vm_preload_proc(c);
+ cmd->vm_preload_proc(c,cmd,c->argc,c->argv);
} else {
- if (cmd->vm_firstkey == 0) return 0;
- last = cmd->vm_lastkey;
- if (last < 0) last = c->argc+last;
- for (j = cmd->vm_firstkey; j <= last; j += cmd->vm_keystep)
- waitForSwappedKey(c,c->argv[j]);
+ waitForMultipleSwappedKeys(c,cmd,c->argc,c->argv);
}
/* If the client was blocked for at least one key, mark it as blocked. */
/* Remove the key from the list of keys this client is waiting for. */
listRewind(c->io_keys,&li);
while ((ln = listNext(&li)) != NULL) {
- if (compareStringObjects(ln->value,key) == 0) {
+ if (equalStringObjects(ln->value,key)) {
listDelNode(c->io_keys,ln);
break;
}
return listLength(c->io_keys) == 0;
}
+/* Every time we now a key was loaded back in memory, we handle clients
+ * waiting for this key if any. */
static void handleClientsBlockedOnSwappedKey(redisDb *db, robj *key) {
struct dictEntry *de;
list *l;
static void configSetCommand(redisClient *c) {
robj *o = getDecodedObject(c->argv[3]);
+ long long ll;
+
if (!strcasecmp(c->argv[2]->ptr,"dbfilename")) {
zfree(server.dbfilename);
server.dbfilename = zstrdup(o->ptr);
zfree(server.masterauth);
server.masterauth = zstrdup(o->ptr);
} else if (!strcasecmp(c->argv[2]->ptr,"maxmemory")) {
- server.maxmemory = strtoll(o->ptr, NULL, 10);
+ if (getLongLongFromObject(o,&ll) == REDIS_ERR ||
+ ll < 0) goto badfmt;
+ server.maxmemory = ll;
+ } else if (!strcasecmp(c->argv[2]->ptr,"timeout")) {
+ if (getLongLongFromObject(o,&ll) == REDIS_ERR ||
+ ll < 0 || ll > LONG_MAX) goto badfmt;
+ server.maxidletime = ll;
+ } else if (!strcasecmp(c->argv[2]->ptr,"appendfsync")) {
+ if (!strcasecmp(o->ptr,"no")) {
+ server.appendfsync = APPENDFSYNC_NO;
+ } else if (!strcasecmp(o->ptr,"everysec")) {
+ server.appendfsync = APPENDFSYNC_EVERYSEC;
+ } else if (!strcasecmp(o->ptr,"always")) {
+ server.appendfsync = APPENDFSYNC_ALWAYS;
+ } else {
+ goto badfmt;
+ }
+ } else if (!strcasecmp(c->argv[2]->ptr,"no-appendfsync-on-rewrite")) {
+ int yn = yesnotoi(o->ptr);
+
+ if (yn == -1) goto badfmt;
+ server.no_appendfsync_on_rewrite = yn;
+ } else if (!strcasecmp(c->argv[2]->ptr,"appendonly")) {
+ int old = server.appendonly;
+ int new = yesnotoi(o->ptr);
+
+ if (new == -1) goto badfmt;
+ if (old != new) {
+ if (new == 0) {
+ stopAppendOnly();
+ } else {
+ if (startAppendOnly() == REDIS_ERR) {
+ addReplySds(c,sdscatprintf(sdsempty(),
+ "-ERR Unable to turn on AOF. Check server logs.\r\n"));
+ decrRefCount(o);
+ return;
+ }
+ }
+ }
+ } else if (!strcasecmp(c->argv[2]->ptr,"save")) {
+ int vlen, j;
+ sds *v = sdssplitlen(o->ptr,sdslen(o->ptr)," ",1,&vlen);
+
+ /* Perform sanity check before setting the new config:
+ * - Even number of args
+ * - Seconds >= 1, changes >= 0 */
+ if (vlen & 1) {
+ sdsfreesplitres(v,vlen);
+ goto badfmt;
+ }
+ for (j = 0; j < vlen; j++) {
+ char *eptr;
+ long val;
+
+ val = strtoll(v[j], &eptr, 10);
+ if (eptr[0] != '\0' ||
+ ((j & 1) == 0 && val < 1) ||
+ ((j & 1) == 1 && val < 0)) {
+ sdsfreesplitres(v,vlen);
+ goto badfmt;
+ }
+ }
+ /* Finally set the new config */
+ resetServerSaveParams();
+ for (j = 0; j < vlen; j += 2) {
+ time_t seconds;
+ int changes;
+
+ seconds = strtoll(v[j],NULL,10);
+ changes = strtoll(v[j+1],NULL,10);
+ appendServerSaveParams(seconds, changes);
+ }
+ sdsfreesplitres(v,vlen);
} else {
addReplySds(c,sdscatprintf(sdsempty(),
"-ERR not supported CONFIG parameter %s\r\n",
}
decrRefCount(o);
addReply(c,shared.ok);
+ return;
+
+badfmt: /* Bad format errors */
+ addReplySds(c,sdscatprintf(sdsempty(),
+ "-ERR invalid argument '%s' for CONFIG SET '%s'\r\n",
+ (char*)o->ptr,
+ (char*)c->argv[2]->ptr));
+ decrRefCount(o);
}
static void configGetCommand(redisClient *c) {
if (stringmatch(pattern,"maxmemory",0)) {
char buf[128];
- snprintf(buf,128,"%llu\n",server.maxmemory);
+ ll2string(buf,128,server.maxmemory);
addReplyBulkCString(c,"maxmemory");
addReplyBulkCString(c,buf);
matches++;
}
+ if (stringmatch(pattern,"timeout",0)) {
+ char buf[128];
+
+ ll2string(buf,128,server.maxidletime);
+ addReplyBulkCString(c,"timeout");
+ addReplyBulkCString(c,buf);
+ matches++;
+ }
+ if (stringmatch(pattern,"appendonly",0)) {
+ addReplyBulkCString(c,"appendonly");
+ addReplyBulkCString(c,server.appendonly ? "yes" : "no");
+ matches++;
+ }
+ if (stringmatch(pattern,"no-appendfsync-on-rewrite",0)) {
+ addReplyBulkCString(c,"no-appendfsync-on-rewrite");
+ addReplyBulkCString(c,server.no_appendfsync_on_rewrite ? "yes" : "no");
+ matches++;
+ }
+ if (stringmatch(pattern,"appendfsync",0)) {
+ char *policy;
+
+ switch(server.appendfsync) {
+ case APPENDFSYNC_NO: policy = "no"; break;
+ case APPENDFSYNC_EVERYSEC: policy = "everysec"; break;
+ case APPENDFSYNC_ALWAYS: policy = "always"; break;
+ default: policy = "unknown"; break; /* too harmless to panic */
+ }
+ addReplyBulkCString(c,"appendfsync");
+ addReplyBulkCString(c,policy);
+ matches++;
+ }
+ if (stringmatch(pattern,"save",0)) {
+ sds buf = sdsempty();
+ int j;
+
+ for (j = 0; j < server.saveparamslen; j++) {
+ buf = sdscatprintf(buf,"%ld %d",
+ server.saveparams[j].seconds,
+ server.saveparams[j].changes);
+ if (j != server.saveparamslen-1)
+ buf = sdscatlen(buf," ",1);
+ }
+ addReplyBulkCString(c,"save");
+ addReplyBulkCString(c,buf);
+ sdsfree(buf);
+ matches++;
+ }
decrRefCount(o);
lenobj->ptr = sdscatprintf(sdsempty(),"*%d\r\n",matches*2);
}
pubsubPattern *pa = a, *pb = b;
return (pa->client == pb->client) &&
- (compareStringObjects(pa->pattern,pb->pattern) == 0);
+ (equalStringObjects(pa->pattern,pb->pattern));
}
/* Subscribe a client to a channel. Returns 1 if the operation succeeded, or
addReply(c,shared.mbulk3);
addReply(c,shared.subscribebulk);
addReplyBulk(c,channel);
- addReplyLong(c,dictSize(c->pubsub_channels)+listLength(c->pubsub_patterns));
+ addReplyLongLong(c,dictSize(c->pubsub_channels)+listLength(c->pubsub_patterns));
return retval;
}
addReply(c,shared.mbulk3);
addReply(c,shared.unsubscribebulk);
addReplyBulk(c,channel);
- addReplyLong(c,dictSize(c->pubsub_channels)+
+ addReplyLongLong(c,dictSize(c->pubsub_channels)+
listLength(c->pubsub_patterns));
}
addReply(c,shared.mbulk3);
addReply(c,shared.psubscribebulk);
addReplyBulk(c,pattern);
- addReplyLong(c,dictSize(c->pubsub_channels)+listLength(c->pubsub_patterns));
+ addReplyLongLong(c,dictSize(c->pubsub_channels)+listLength(c->pubsub_patterns));
return retval;
}
addReply(c,shared.mbulk3);
addReply(c,shared.punsubscribebulk);
addReplyBulk(c,pattern);
- addReplyLong(c,dictSize(c->pubsub_channels)+
+ addReplyLongLong(c,dictSize(c->pubsub_channels)+
listLength(c->pubsub_patterns));
}
decrRefCount(pattern);
static void publishCommand(redisClient *c) {
int receivers = pubsubPublishMessage(c->argv[1],c->argv[2]);
- addReplyLong(c,receivers);
+ addReplyLongLong(c,receivers);
+}
+
+/* ===================== WATCH (CAS alike for MULTI/EXEC) ===================
+ *
+ * The implementation uses a per-DB hash table mapping keys to list of clients
+ * WATCHing those keys, so that given a key that is going to be modified
+ * we can mark all the associated clients as dirty.
+ *
+ * Also every client contains a list of WATCHed keys so that's possible to
+ * un-watch such keys when the client is freed or when UNWATCH is called. */
+
+/* In the client->watched_keys list we need to use watchedKey structures
+ * as in order to identify a key in Redis we need both the key name and the
+ * DB */
+typedef struct watchedKey {
+ robj *key;
+ redisDb *db;
+} watchedKey;
+
+/* Watch for the specified key */
+static void watchForKey(redisClient *c, robj *key) {
+ list *clients = NULL;
+ listIter li;
+ listNode *ln;
+ watchedKey *wk;
+
+ /* Check if we are already watching for this key */
+ listRewind(c->watched_keys,&li);
+ while((ln = listNext(&li))) {
+ wk = listNodeValue(ln);
+ if (wk->db == c->db && equalStringObjects(key,wk->key))
+ return; /* Key already watched */
+ }
+ /* This key is not already watched in this DB. Let's add it */
+ clients = dictFetchValue(c->db->watched_keys,key);
+ if (!clients) {
+ clients = listCreate();
+ dictAdd(c->db->watched_keys,key,clients);
+ incrRefCount(key);
+ }
+ listAddNodeTail(clients,c);
+ /* Add the new key to the lits of keys watched by this client */
+ wk = zmalloc(sizeof(*wk));
+ wk->key = key;
+ wk->db = c->db;
+ incrRefCount(key);
+ listAddNodeTail(c->watched_keys,wk);
+}
+
+/* Unwatch all the keys watched by this client. To clean the EXEC dirty
+ * flag is up to the caller. */
+static void unwatchAllKeys(redisClient *c) {
+ listIter li;
+ listNode *ln;
+
+ if (listLength(c->watched_keys) == 0) return;
+ listRewind(c->watched_keys,&li);
+ while((ln = listNext(&li))) {
+ list *clients;
+ watchedKey *wk;
+
+ /* Lookup the watched key -> clients list and remove the client
+ * from the list */
+ wk = listNodeValue(ln);
+ clients = dictFetchValue(wk->db->watched_keys, wk->key);
+ assert(clients != NULL);
+ listDelNode(clients,listSearchKey(clients,c));
+ /* Kill the entry at all if this was the only client */
+ if (listLength(clients) == 0)
+ dictDelete(wk->db->watched_keys, wk->key);
+ /* Remove this watched key from the client->watched list */
+ listDelNode(c->watched_keys,ln);
+ decrRefCount(wk->key);
+ zfree(wk);
+ }
+}
+
+/* "Touch" a key, so that if this key is being WATCHed by some client the
+ * next EXEC will fail. */
+static void touchWatchedKey(redisDb *db, robj *key) {
+ list *clients;
+ listIter li;
+ listNode *ln;
+
+ if (dictSize(db->watched_keys) == 0) return;
+ clients = dictFetchValue(db->watched_keys, key);
+ if (!clients) return;
+
+ /* Mark all the clients watching this key as REDIS_DIRTY_CAS */
+ /* Check if we are already watching for this key */
+ listRewind(clients,&li);
+ while((ln = listNext(&li))) {
+ redisClient *c = listNodeValue(ln);
+
+ c->flags |= REDIS_DIRTY_CAS;
+ }
+}
+
+/* On FLUSHDB or FLUSHALL all the watched keys that are present before the
+ * flush but will be deleted as effect of the flushing operation should
+ * be touched. "dbid" is the DB that's getting the flush. -1 if it is
+ * a FLUSHALL operation (all the DBs flushed). */
+static void touchWatchedKeysOnFlush(int dbid) {
+ listIter li1, li2;
+ listNode *ln;
+
+ /* For every client, check all the waited keys */
+ listRewind(server.clients,&li1);
+ while((ln = listNext(&li1))) {
+ redisClient *c = listNodeValue(ln);
+ listRewind(c->watched_keys,&li2);
+ while((ln = listNext(&li2))) {
+ watchedKey *wk = listNodeValue(ln);
+
+ /* For every watched key matching the specified DB, if the
+ * key exists, mark the client as dirty, as the key will be
+ * removed. */
+ if (dbid == -1 || wk->db->id == dbid) {
+ if (dictFind(wk->db->dict, wk->key->ptr) != NULL)
+ c->flags |= REDIS_DIRTY_CAS;
+ }
+ }
+ }
+}
+
+static void watchCommand(redisClient *c) {
+ int j;
+
+ if (c->flags & REDIS_MULTI) {
+ addReplySds(c,sdsnew("-ERR WATCH inside MULTI is not allowed\r\n"));
+ return;
+ }
+ for (j = 1; j < c->argc; j++)
+ watchForKey(c,c->argv[j]);
+ addReply(c,shared.ok);
+}
+
+static void unwatchCommand(redisClient *c) {
+ unwatchAllKeys(c);
+ c->flags &= (~REDIS_DIRTY_CAS);
+ addReply(c,shared.ok);
}
/* ================================= Debugging ============================== */
+/* Compute the sha1 of string at 's' with 'len' bytes long.
+ * The SHA1 is then xored againt the string pointed by digest.
+ * Since xor is commutative, this operation is used in order to
+ * "add" digests relative to unordered elements.
+ *
+ * So digest(a,b,c,d) will be the same of digest(b,a,c,d) */
+static void xorDigest(unsigned char *digest, void *ptr, size_t len) {
+ SHA1_CTX ctx;
+ unsigned char hash[20], *s = ptr;
+ int j;
+
+ SHA1Init(&ctx);
+ SHA1Update(&ctx,s,len);
+ SHA1Final(hash,&ctx);
+
+ for (j = 0; j < 20; j++)
+ digest[j] ^= hash[j];
+}
+
+static void xorObjectDigest(unsigned char *digest, robj *o) {
+ o = getDecodedObject(o);
+ xorDigest(digest,o->ptr,sdslen(o->ptr));
+ decrRefCount(o);
+}
+
+/* This function instead of just computing the SHA1 and xoring it
+ * against diget, also perform the digest of "digest" itself and
+ * replace the old value with the new one.
+ *
+ * So the final digest will be:
+ *
+ * digest = SHA1(digest xor SHA1(data))
+ *
+ * This function is used every time we want to preserve the order so
+ * that digest(a,b,c,d) will be different than digest(b,c,d,a)
+ *
+ * Also note that mixdigest("foo") followed by mixdigest("bar")
+ * will lead to a different digest compared to "fo", "obar".
+ */
+static void mixDigest(unsigned char *digest, void *ptr, size_t len) {
+ SHA1_CTX ctx;
+ char *s = ptr;
+
+ xorDigest(digest,s,len);
+ SHA1Init(&ctx);
+ SHA1Update(&ctx,digest,20);
+ SHA1Final(digest,&ctx);
+}
+
+static void mixObjectDigest(unsigned char *digest, robj *o) {
+ o = getDecodedObject(o);
+ mixDigest(digest,o->ptr,sdslen(o->ptr));
+ decrRefCount(o);
+}
+
+/* Compute the dataset digest. Since keys, sets elements, hashes elements
+ * are not ordered, we use a trick: every aggregate digest is the xor
+ * of the digests of their elements. This way the order will not change
+ * the result. For list instead we use a feedback entering the output digest
+ * as input in order to ensure that a different ordered list will result in
+ * a different digest. */
+static void computeDatasetDigest(unsigned char *final) {
+ unsigned char digest[20];
+ char buf[128];
+ dictIterator *di = NULL;
+ dictEntry *de;
+ int j;
+ uint32_t aux;
+
+ memset(final,0,20); /* Start with a clean result */
+
+ for (j = 0; j < server.dbnum; j++) {
+ redisDb *db = server.db+j;
+
+ if (dictSize(db->dict) == 0) continue;
+ di = dictGetIterator(db->dict);
+
+ /* hash the DB id, so the same dataset moved in a different
+ * DB will lead to a different digest */
+ aux = htonl(j);
+ mixDigest(final,&aux,sizeof(aux));
+
+ /* Iterate this DB writing every entry */
+ while((de = dictNext(di)) != NULL) {
+ sds key;
+ robj *keyobj, *o;
+ time_t expiretime;
+
+ memset(digest,0,20); /* This key-val digest */
+ key = dictGetEntryKey(de);
+ keyobj = createStringObject(key,sdslen(key));
+
+ mixDigest(digest,key,sdslen(key));
+
+ /* Make sure the key is loaded if VM is active */
+ o = lookupKeyRead(db,keyobj);
+
+ aux = htonl(o->type);
+ mixDigest(digest,&aux,sizeof(aux));
+ expiretime = getExpire(db,keyobj);
+
+ /* Save the key and associated value */
+ if (o->type == REDIS_STRING) {
+ mixObjectDigest(digest,o);
+ } else if (o->type == REDIS_LIST) {
+ listTypeIterator *li = listTypeInitIterator(o,0,REDIS_TAIL);
+ listTypeEntry entry;
+ while(listTypeNext(li,&entry)) {
+ robj *eleobj = listTypeGet(&entry);
+ mixObjectDigest(digest,eleobj);
+ decrRefCount(eleobj);
+ }
+ listTypeReleaseIterator(li);
+ } else if (o->type == REDIS_SET) {
+ dict *set = o->ptr;
+ dictIterator *di = dictGetIterator(set);
+ dictEntry *de;
+
+ while((de = dictNext(di)) != NULL) {
+ robj *eleobj = dictGetEntryKey(de);
+
+ xorObjectDigest(digest,eleobj);
+ }
+ dictReleaseIterator(di);
+ } else if (o->type == REDIS_ZSET) {
+ zset *zs = o->ptr;
+ dictIterator *di = dictGetIterator(zs->dict);
+ dictEntry *de;
+
+ while((de = dictNext(di)) != NULL) {
+ robj *eleobj = dictGetEntryKey(de);
+ double *score = dictGetEntryVal(de);
+ unsigned char eledigest[20];
+
+ snprintf(buf,sizeof(buf),"%.17g",*score);
+ memset(eledigest,0,20);
+ mixObjectDigest(eledigest,eleobj);
+ mixDigest(eledigest,buf,strlen(buf));
+ xorDigest(digest,eledigest,20);
+ }
+ dictReleaseIterator(di);
+ } else if (o->type == REDIS_HASH) {
+ hashTypeIterator *hi;
+ robj *obj;
+
+ hi = hashTypeInitIterator(o);
+ while (hashTypeNext(hi) != REDIS_ERR) {
+ unsigned char eledigest[20];
+
+ memset(eledigest,0,20);
+ obj = hashTypeCurrent(hi,REDIS_HASH_KEY);
+ mixObjectDigest(eledigest,obj);
+ decrRefCount(obj);
+ obj = hashTypeCurrent(hi,REDIS_HASH_VALUE);
+ mixObjectDigest(eledigest,obj);
+ decrRefCount(obj);
+ xorDigest(digest,eledigest,20);
+ }
+ hashTypeReleaseIterator(hi);
+ } else {
+ redisPanic("Unknown object type");
+ }
+ /* If the key has an expire, add it to the mix */
+ if (expiretime != -1) xorDigest(digest,"!!expire!!",10);
+ /* We can finally xor the key-val digest to the final digest */
+ xorDigest(final,digest,20);
+ decrRefCount(keyobj);
+ }
+ dictReleaseIterator(di);
+ }
+}
+
static void debugCommand(redisClient *c) {
if (!strcasecmp(c->argv[1]->ptr,"segfault")) {
*((char*)-1) = 'x';
redisLog(REDIS_WARNING,"Append Only File loaded by DEBUG LOADAOF");
addReply(c,shared.ok);
} else if (!strcasecmp(c->argv[1]->ptr,"object") && c->argc == 3) {
- dictEntry *de = dictFind(c->db->dict,c->argv[2]);
- robj *key, *val;
+ dictEntry *de = dictFind(c->db->dict,c->argv[2]->ptr);
+ robj *val;
if (!de) {
addReply(c,shared.nokeyerr);
return;
}
- key = dictGetEntryKey(de);
val = dictGetEntryVal(de);
- if (!server.vm_enabled || (key->storage == REDIS_VM_MEMORY ||
- key->storage == REDIS_VM_SWAPPING)) {
+ if (!server.vm_enabled || (val->storage == REDIS_VM_MEMORY ||
+ val->storage == REDIS_VM_SWAPPING)) {
char *strenc;
char buf[128];
strenc = buf;
}
addReplySds(c,sdscatprintf(sdsempty(),
- "+Key at:%p refcount:%d, value at:%p refcount:%d "
+ "+Value at:%p refcount:%d "
"encoding:%s serializedlength:%lld\r\n",
- (void*)key, key->refcount, (void*)val, val->refcount,
+ (void*)val, val->refcount,
strenc, (long long) rdbSavedObjectLen(val,NULL)));
} else {
+ vmpointer *vp = (vmpointer*) val;
addReplySds(c,sdscatprintf(sdsempty(),
- "+Key at:%p refcount:%d, value swapped at: page %llu "
+ "+Value swapped at: page %llu "
"using %llu pages\r\n",
- (void*)key, key->refcount, (unsigned long long) key->vm.page,
- (unsigned long long) key->vm.usedpages));
+ (unsigned long long) vp->page,
+ (unsigned long long) vp->usedpages));
}
} else if (!strcasecmp(c->argv[1]->ptr,"swapin") && c->argc == 3) {
lookupKeyRead(c->db,c->argv[2]);
addReply(c,shared.ok);
} else if (!strcasecmp(c->argv[1]->ptr,"swapout") && c->argc == 3) {
- dictEntry *de = dictFind(c->db->dict,c->argv[2]);
- robj *key, *val;
+ dictEntry *de = dictFind(c->db->dict,c->argv[2]->ptr);
+ robj *val;
+ vmpointer *vp;
if (!server.vm_enabled) {
addReplySds(c,sdsnew("-ERR Virtual Memory is disabled\r\n"));
addReply(c,shared.nokeyerr);
return;
}
- key = dictGetEntryKey(de);
val = dictGetEntryVal(de);
- /* If the key is shared we want to create a copy */
- if (key->refcount > 1) {
- robj *newkey = dupStringObject(key);
- decrRefCount(key);
- key = dictGetEntryKey(de) = newkey;
- }
/* Swap it */
- if (key->storage != REDIS_VM_MEMORY) {
+ if (val->storage != REDIS_VM_MEMORY) {
addReplySds(c,sdsnew("-ERR This key is not in memory\r\n"));
- } else if (vmSwapObjectBlocking(key,val) == REDIS_OK) {
- dictGetEntryVal(de) = NULL;
+ } else if (val->refcount != 1) {
+ addReplySds(c,sdsnew("-ERR Object is shared\r\n"));
+ } else if ((vp = vmSwapObjectBlocking(val)) != NULL) {
+ dictGetEntryVal(de) = vp;
addReply(c,shared.ok);
} else {
addReply(c,shared.err);
}
+ } else if (!strcasecmp(c->argv[1]->ptr,"populate") && c->argc == 3) {
+ long keys, j;
+ robj *key, *val;
+ char buf[128];
+
+ if (getLongFromObjectOrReply(c, c->argv[2], &keys, NULL) != REDIS_OK)
+ return;
+ for (j = 0; j < keys; j++) {
+ snprintf(buf,sizeof(buf),"key:%lu",j);
+ key = createStringObject(buf,strlen(buf));
+ if (lookupKeyRead(c->db,key) != NULL) {
+ decrRefCount(key);
+ continue;
+ }
+ snprintf(buf,sizeof(buf),"value:%lu",j);
+ val = createStringObject(buf,strlen(buf));
+ dbAdd(c->db,key,val);
+ decrRefCount(key);
+ }
+ addReply(c,shared.ok);
+ } else if (!strcasecmp(c->argv[1]->ptr,"digest") && c->argc == 2) {
+ unsigned char digest[20];
+ sds d = sdsnew("+");
+ int j;
+
+ computeDatasetDigest(digest);
+ for (j = 0; j < 20; j++)
+ d = sdscatprintf(d, "%02x",digest[j]);
+
+ d = sdscatlen(d,"\r\n",2);
+ addReplySds(c,d);
} else {
addReplySds(c,sdsnew(
"-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT <key>|SWAPIN <key>|SWAPOUT <key>|RELOAD]\r\n"));
static void _redisAssert(char *estr, char *file, int line) {
redisLog(REDIS_WARNING,"=== ASSERTION FAILED ===");
- redisLog(REDIS_WARNING,"==> %s:%d '%s' is not true\n",file,line,estr);
+ redisLog(REDIS_WARNING,"==> %s:%d '%s' is not true",file,line,estr);
#ifdef HAVE_BACKTRACE
redisLog(REDIS_WARNING,"(forcing SIGSEGV in order to print the stack trace)");
*((char*)-1) = 'x';
}
static void version() {
- printf("Redis server version %s\n", REDIS_VERSION);
+ printf("Redis server version %s (%s:%d)\n", REDIS_VERSION,
+ redisGitSHA1(), atoi(redisGitDirty()) > 0);
exit(0);
}
time_t start;
initServerConfig();
+ sortCommandTable();
if (argc == 2) {
if (strcmp(argv[1], "-v") == 0 ||
strcmp(argv[1], "--version") == 0) version();
_exit(0);
}
+static void sigtermHandler(int sig) {
+ REDIS_NOTUSED(sig);
+
+ redisLog(REDIS_WARNING,"SIGTERM received, scheduling shutting down...");
+ server.shutdown_asap = 1;
+}
+
static void setupSigSegvAction(void) {
struct sigaction act;
sigaction (SIGFPE, &act, NULL);
sigaction (SIGILL, &act, NULL);
sigaction (SIGBUS, &act, NULL);
+
+ act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
+ act.sa_handler = sigtermHandler;
+ sigaction (SIGTERM, &act, NULL);
return;
}
projects / redis.git / blobdiff