diff options
Diffstat (limited to 'lang/sql/adapter/db_pragma.c')
| -rw-r--r-- | lang/sql/adapter/db_pragma.c | 386 |
1 files changed, 373 insertions, 13 deletions
diff --git a/lang/sql/adapter/db_pragma.c b/lang/sql/adapter/db_pragma.c index 4c70afec..e7f8fa0e 100644 --- a/lang/sql/adapter/db_pragma.c +++ b/lang/sql/adapter/db_pragma.c @@ -1,7 +1,7 @@ /*- * See the file LICENSE for redistribution information. * - * Copyright (c) 2010, 2012 Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2010, 2015 Oracle and/or its affiliates. All rights reserved. */ /* @@ -11,7 +11,6 @@ #include "btreeInt.h" extern void returnSingleInt(Parse *, const char *, i64); -extern u8 getBoolean(const char *); extern int __os_exists (ENV *, const char *, int *); extern int __os_unlink (ENV *, const char *, int); extern int __os_mkdir (ENV *, const char *, int); @@ -23,6 +22,16 @@ extern int __env_ref_get (DB_ENV *, u_int32_t *); static const char *PRAGMA_FILE = "pragma"; static const char *PRAGMA_VERSION = "1.0"; +static const char *ACK_POLICY_ALL = "all_sites"; +static const char *ACK_POLICY_ALL_AVAILABLE = "all_available"; +static const char *ACK_POLICY_NONE = "none"; +static const char *ACK_POLICY_ONE = "one"; +static const char *ACK_POLICY_QUORUM = "quorum"; + +static const char *REP_SITE_MASTER = "MASTER"; +static const char *REP_SITE_CLIENT = "CLIENT"; +static const char *REP_SITE_UNKNOWN = "UNKNOWN"; + static const u32 HDR_SIZE = 256; static const u32 RECORD_HDR_SIZE = 8; static const u32 VERSION_RECORD_SIZE = 12; @@ -47,6 +56,57 @@ static const u32 DEFINED_PRAGMAS = 8; #define dbExists (pDb->pBt->pBt->full_name != NULL && \ !__os_exists(NULL, pDb->pBt->pBt->full_name, NULL)) +/* Translates a text ack policy to its DB value. */ +static int textToAckPolicy(const char *policy) +{ + int len; + if (policy == NULL) + return (-1); + + len = (int)strlen(policy); + + if ((sqlite3StrNICmp(policy, ACK_POLICY_ALL, len)) == 0) + return (DB_REPMGR_ACKS_ALL); + else if ((sqlite3StrNICmp(policy, ACK_POLICY_ALL_AVAILABLE, len)) == 0) + return (DB_REPMGR_ACKS_ALL_AVAILABLE); + else if ((sqlite3StrNICmp(policy, ACK_POLICY_NONE, len)) == 0) + return (DB_REPMGR_ACKS_NONE); + else if ((sqlite3StrNICmp(policy, ACK_POLICY_ONE, len)) == 0) + return (DB_REPMGR_ACKS_ONE); + else if ((sqlite3StrNICmp(policy, ACK_POLICY_QUORUM, len)) == 0) + return (DB_REPMGR_ACKS_QUORUM); + else + return (-1); +} + +/* Translates a DB value ack policy to text. */ +static const char *ackPolicyToText(int policy) +{ + if (policy == DB_REPMGR_ACKS_ALL) + return (ACK_POLICY_ALL); + else if (policy == DB_REPMGR_ACKS_ALL_AVAILABLE) + return (ACK_POLICY_ALL_AVAILABLE); + else if (policy == DB_REPMGR_ACKS_NONE) + return (ACK_POLICY_NONE); + else if (policy == DB_REPMGR_ACKS_ONE) + return (ACK_POLICY_ONE); + else if (policy == DB_REPMGR_ACKS_QUORUM) + return (ACK_POLICY_QUORUM); + else + return (NULL); +} + +/* Translates a replication site type to text. */ +static const char *repSiteTypeToText(rep_site_type_t type) +{ + if (type == BDBSQL_REP_MASTER) + return (REP_SITE_MASTER); + else if (type == BDBSQL_REP_CLIENT) + return (REP_SITE_CLIENT); + else + return (REP_SITE_UNKNOWN); +} + static u8 envIsClosed(Parse *pParse, Btree *p, const char *pragmaName) { int rc; @@ -82,6 +142,12 @@ int bdbsqlPragmaMultiversion(Parse *pParse, Btree *p, u8 on) return 1; pBt = p->pBt; + if (pBt->blobs_enabled && on) { + sqlite3ErrorMsg(pParse, + "Cannot enable both multiversion and large record optimization."); + return 1; + } + /* Do not want other processes opening the environment */ mutexOpen = sqlite3MutexAlloc(OPEN_MUTEX(pBt->dbStorage)); sqlite3_mutex_enter(mutexOpen); @@ -250,15 +316,29 @@ static int bdbsqlPragmaStartReplication(Parse *pParse, Db *pDb) } if (dbExists) { + if (!pBt->pBt->env_opened) { + if ((rc = btreeOpenEnvironment(pBt, 1)) != SQLITE_OK) + sqlite3ErrorMsg(pParse, "Could not start " + "replication on an existing database"); + goto done; + } + /* - * Turning on replication requires recovery on the underlying + * Opening the environment started repmgr if it was + * configured for it, so we are done here. + */ + if (supportsReplication(pBt)) + goto done; + /* + * Turning on replication on an existing environment not + * configured for replication requires recovery on the * BDB environment, which requires single-threaded access. * Make sure there are no other processes or threads accessing * it. This is not foolproof, because there is a small chance * that another process or thread could slip in there between * this call and the recovery, but it should cover most cases. */ - if (hasDatabaseConnections(pBt)) { + if (hasDatabaseConnections(pBt) || pBt->pBt->nRef > 1) { sqlite3ErrorMsg(pParse, "Close all database " "connections before turning on replication"); goto done; @@ -275,7 +355,7 @@ static int bdbsqlPragmaStartReplication(Parse *pParse, Db *pDb) * the underlying BDB environment with replication always * performs a recovery. */ - pBt->pBt->repStartMaster = 1; + pBt->pBt->repForceRecover = 1; if ((rc = btreeReopenEnvironment(pBt, 0)) != SQLITE_OK) sqlite3ErrorMsg(pParse, "Could not " "start replication on an existing database"); @@ -310,6 +390,7 @@ done: static int bdbsqlPragmaStopReplication(Parse *pParse, Db *pDb) { Btree *pBt; + char *old_addr = NULL; int rc = SQLITE_OK; pBt = pDb->pBt; @@ -341,6 +422,13 @@ static int bdbsqlPragmaStopReplication(Parse *pParse, Db *pDb) */ pBt->pBt->repForceRecover = 1; rc = btreeReopenEnvironment(pBt, 1); + sqlite3_mutex_enter(pBt->pBt->mutex); + pBt->pBt->repRole = BDBSQL_REP_UNKNOWN; + old_addr = pBt->pBt->master_address; + pBt->pBt->master_address = NULL; + sqlite3_mutex_leave(pBt->pBt->mutex); + if (old_addr) + sqlite3_free(old_addr); done: return rc; @@ -383,14 +471,14 @@ int bdbsqlPragma(Parse *pParse, char *zLeft, char *zRight, int iDb) } } else if (sqlite3StrNICmp(zLeft, "txn_bulk", 8) == 0) { if (zRight) - pBt->txn_bulk = getBoolean(zRight); + pBt->txn_bulk = sqlite3GetBoolean(zRight, 0); returnSingleInt(pParse, "txn_bulk", (i64)pBt->txn_bulk); parsed = 1; /* Enables MVCC and transactions snapshots. */ } else if (sqlite3StrNICmp(zLeft, "multiversion", 12) == 0) { if (zRight) bdbsqlPragmaMultiversion(pParse, pDb->pBt, - getBoolean(zRight)); + sqlite3GetBoolean(zRight, 0)); returnSingleInt(pParse, "multiversion", (i64)((pDb->pBt->pBt->env_oflags & DB_MULTIVERSION)? @@ -401,7 +489,7 @@ int bdbsqlPragma(Parse *pParse, char *zLeft, char *zRight, int iDb) */ } else if (sqlite3StrNICmp(zLeft, "snapshot_isolation", 18) == 0) { if (zRight) { - if (getBoolean(zRight)) { + if (sqlite3GetBoolean(zRight, 0)) { if (pDb->pBt->pBt->env_oflags & DB_MULTIVERSION) { pDb->pBt->pBt->read_txn_flags @@ -530,7 +618,7 @@ int bdbsqlPragma(Parse *pParse, char *zLeft, char *zRight, int iDb) setValue[0] = '\0'; startedRep = stoppedRep = turningOn = 0; if (zRight) { - turningOn = (getBoolean(zRight) == 1); + turningOn = (sqlite3GetBoolean(zRight, 0) == 1); strcpy(setValue, turningOn ? "1" : "0"); rc = setPersistentPragma(pDb->pBt, zLeft, setValue, pParse); @@ -596,7 +684,7 @@ int bdbsqlPragma(Parse *pParse, char *zLeft, char *zRight, int iDb) outValue[0] = '\0'; if (zRight) pDb->pBt->pBt->repStartMaster = - getBoolean(zRight) == 1 ? 1 : 0; + sqlite3GetBoolean(zRight, 0) == 1 ? 1 : 0; strcpy(outValue, pDb->pBt->pBt->repStartMaster == 1 ? "1" : "0"); sqlite3VdbeSetNumCols(pParse->pVdbe, 1); @@ -733,6 +821,40 @@ int bdbsqlPragma(Parse *pParse, char *zLeft, char *zRight, int iDb) pDb->pBt->txn_priority); parsed = 1; /* + * PRAGMA bdbsql_log_buffer; -- DB_ENV->get_lg_bsize + * PRAGMA bdbsql_log_buffer = N; -- DB_ENV->set_lg_bsize + * N provides the size of the log buffer size in bytes. + */ + } else if (sqlite3StrNICmp(zLeft, "bdbsql_log_buffer", 17) == 0) { + int iLimit = -2; + u_int32_t val; + + if (zRight && + envIsClosed(pParse, pDb->pBt, "bdbsql_log_buffer")) { + if (!sqlite3GetInt32(zRight, &iLimit) || + iLimit < 0) + sqlite3ErrorMsg(pParse, + "Invalid value bdbsql_log_buffer %s", + zRight); + else { + val = iLimit; + if ((ret = + pDb->pBt->pBt->dbenv->set_lg_bsize( + pDb->pBt->pBt->dbenv, val)) != 0) { + sqlite3ErrorMsg(pParse, + "Failed to set log buffer size " + "error: %d.", ret); + } + } + } else if (zRight == NULL) { + /* Get existing value */ + pDb->pBt->pBt->dbenv->get_lg_bsize( + pDb->pBt->pBt->dbenv, &val); + + returnSingleInt(pParse, "bdbsql_log_buffer", val); + } + parsed = 1; + /* * PRAGMA bdbsql_single_process = boolean; * Turn on/off omit sharing. */ @@ -743,7 +865,7 @@ int bdbsqlPragma(Parse *pParse, char *zLeft, char *zRight, int iDb) if (zRight && envIsClosed(pParse, pDb->pBt, "bdbsql_single_process")) { - new_value = getBoolean(zRight); + new_value = sqlite3GetBoolean(zRight, 0); if (new_value != pDb->pBt->pBt->single_process) is_changed = 1; } @@ -868,7 +990,7 @@ int bdbsqlPragma(Parse *pParse, char *zLeft, char *zRight, int iDb) if (zRight && envIsClosed(pParse, pDb->pBt, "bdbsql_shared_resources")) { - if (pDb->pBt->pBt->need_open) { + if (pDb->pBt->pBt->database_existed) { /* * The DBENV->set_memory_max() method must be * called prior to opening/creating the database @@ -918,7 +1040,7 @@ int bdbsqlPragma(Parse *pParse, char *zLeft, char *zRight, int iDb) if (zRight && envIsClosed(pParse, pDb->pBt, "bdbsql_lock_tablesize")) { - if (pDb->pBt->pBt->need_open) { + if (pDb->pBt->pBt->database_existed) { /* * The DB_ENV->set_lk_tablesize() method must be * called prior to opening/creating the database @@ -949,6 +1071,244 @@ int bdbsqlPragma(Parse *pParse, char *zLeft, char *zRight, int iDb) returnSingleInt(pParse, "bdbsql_lock_tablesize", val); parsed = 1; + /* + * PRAGMA large_record_opt; -- Gets the blob threshold + * PRAGMA large_record_opt = N; -- Sets the blob threshold of all + * tables opened in the database. The blob threshold is set to + * N. A value of 0 disables blobs. + */ + } else if (sqlite3StrNICmp(zLeft, "large_record_opt", 16) == 0) { + int iLimit = -2; + u_int32_t val; + + if (zRight) { + if (!sqlite3GetInt32(zRight, &iLimit) || iLimit < 0) { + sqlite3ErrorMsg(pParse, + "Invalid value large_record_opt %s", + zRight); + } else { + /* + * SQL only supports records up to 1GB in + * size, so reject page counts that result + * in a blob threshold that will always be + * larger than 1 GB, to prevent overflow. + */ + if (iLimit > GIGABYTE) { + sqlite3ErrorMsg(pParse, + "large_record_opt must be less than or equal to 1 gigabyte", + zRight); + } else { + pDb->pBt->pBt->blob_threshold + = iLimit; + if (iLimit > 0) { + pDb->pBt->pBt->blobs_enabled + = 1; + } + } + } + } + + val = pDb->pBt->pBt->blob_threshold; + returnSingleInt(pParse, "large_record_opt", val); + parsed = 1; + /* + * PRAGMA replication_ack_policy; -- DB_ENV->repmgr_get_ack_policy + * PRAGMA replication_ack_policy = + * [all_sites|all_available|none|one|quorum]; + * -- DB_ENV->repmgr_set_ack_policy + * Sets the policy on how many clients must acknowledge a transaction + * commit message before the master will consider it durable. + */ + } else if (sqlite3StrNICmp(zLeft, "replication_ack_policy", 22) == 0) { + int val = -1; + const char *policy; + + if (zRight) { + if ((val = textToAckPolicy(zRight)) == -1) + sqlite3ErrorMsg(pParse, + "Invalid value replication_ack_policy %s", + zRight); + else { + if ((ret = + pDb->pBt->pBt->dbenv->repmgr_set_ack_policy( + pDb->pBt->pBt->dbenv, val)) != 0) { + sqlite3ErrorMsg(pParse, + "Failed to set " + "replication_ack_policy. " + "Error: %d.", ret); + } + } + } + + /* Get existing value */ + pDb->pBt->pBt->dbenv->repmgr_get_ack_policy( + pDb->pBt->pBt->dbenv, &val); + + policy = ackPolicyToText(val); + sqlite3VdbeSetNumCols(pParse->pVdbe, 1); + sqlite3VdbeSetColName(pParse->pVdbe, 0, COLNAME_NAME, + zLeft, SQLITE_STATIC); + if (policy != NULL) + sqlite3VdbeAddOp4(pParse->pVdbe, OP_String8, 0, 1, 0, + policy, 0); + else + sqlite3VdbeAddOp4(pParse->pVdbe, OP_String8, 0, 1, 0, + "No policy", 0); + sqlite3VdbeAddOp2(pParse->pVdbe, OP_ResultRow, 1, 1); + parsed = 1; + /* + * PRAGMA replication_ack_timeout; -- DB_ENV->rep_get_timeout + * PRAGMA replication_ack_timeout = N; -- DB_ENV->rep_set_timeout + * N provides the greater than -1 timeout on acks from clients + * when the master sends a transaction commit message. + */ + } else if + (sqlite3StrNICmp(zLeft, "replication_ack_timeout", 23) == 0) { + int iLimit = -2; + db_timeout_t val; + + if (zRight) { + if (!sqlite3GetInt32(zRight, &iLimit) || iLimit < 0) + sqlite3ErrorMsg(pParse, + "Invalid value replication_ack_timeout %s", + zRight); + else { + val = iLimit; + if ((ret = + pDb->pBt->pBt->dbenv->rep_set_timeout( + pDb->pBt->pBt->dbenv, DB_REP_ACK_TIMEOUT, + val)) != 0) { + sqlite3ErrorMsg(pParse, + "Failed to set " + "replication_ack_timeout. " + "Error: %d.", ret); + } + } + } + + /* Get existing value */ + pDb->pBt->pBt->dbenv->rep_get_timeout( + pDb->pBt->pBt->dbenv, DB_REP_ACK_TIMEOUT, &val); + + returnSingleInt(pParse, "replication_ack_timeout", val); + parsed = 1; + /* + * PRAGMA replication_priority; -- DB_ENV->rep_get_priority + * PRAGMA replication_priority = N; -- DB_ENV->rep_set_priority + * N provides the 1 or greater priority of the replication site, + * which is used to decide what site is elected master. + */ + } else if (sqlite3StrNICmp(zLeft, "replication_priority", 20) == 0) { + int iLimit = -2; + u_int32_t val; + + if (zRight) { + if (!sqlite3GetInt32(zRight, &iLimit) || iLimit < 1) + sqlite3ErrorMsg(pParse, + "Invalid value replication_priority %s", + zRight); + else { + val = iLimit; + if ((ret = + pDb->pBt->pBt->dbenv->rep_set_priority( + pDb->pBt->pBt->dbenv, val)) != 0) { + sqlite3ErrorMsg(pParse, + "Failed to set replication " + "priority. Error: %d.", ret); + } + } + } + + /* Get existing value */ + pDb->pBt->pBt->dbenv->rep_get_priority( + pDb->pBt->pBt->dbenv, &val); + + returnSingleInt(pParse, "replication_priority", val); + parsed = 1; + /* + * PRAGMA replication_num_sites; -- DB_ENV->repmgr_site_list + * Returns the number of sites in the replication group. + */ + } else if (sqlite3StrNICmp(zLeft, "replication_num_sites", 21) == 0) { + u_int32_t val = -1; + DB_REPMGR_SITE *sites = NULL; + + /* Get existing value */ + pDb->pBt->pBt->dbenv->repmgr_site_list( + pDb->pBt->pBt->dbenv, &val, &sites); + + /* + * repmgr_site_list only gets the list of remote sites, + * so add one for the local site if replication has been + * started. + */ + if (pDb->pBt->pBt->repStarted) + val++; + + if (sites) + sqlite3_free(sites); + + returnSingleInt(pParse, "replication_num_sites", val); + parsed = 1; + /* + * PRAGMA replication_perm_failed; -- DB_REP_PERM_FAILED + * Returns the number of times since the last time this pragma was + * called the master was unable to get enough acknowledgments from + * clients when committing a transaction. + */ + } else if (sqlite3StrNICmp(zLeft, "replication_perm_failed", 23) == 0) { + u_int32_t val = 0; + + sqlite3_mutex_enter(pDb->pBt->pBt->mutex); + val = pDb->pBt->pBt->permFailures; + pDb->pBt->pBt->permFailures = 0; + sqlite3_mutex_leave(pDb->pBt->pBt->mutex); + + returnSingleInt(pParse, "replication_perm_failed", val); + parsed = 1; + /* + * PRAGMA replication_site_status; + * Returns MASTER if the site is a replication master, CLIENT if the + * site is a replication client, and UNKNOWN if replication is not + * enabled, or the site status is still being decided (such as during + * an election). + */ + } else if (sqlite3StrNICmp(zLeft, "replication_site_status", 23) == 0) { + rep_site_type_t val; + const char *type; + + sqlite3_mutex_enter(pDb->pBt->pBt->mutex); + val = pDb->pBt->pBt->repRole; + sqlite3_mutex_leave(pDb->pBt->pBt->mutex); + + type = repSiteTypeToText(val); + sqlite3VdbeSetNumCols(pParse->pVdbe, 1); + sqlite3VdbeSetColName(pParse->pVdbe, 0, COLNAME_NAME, + zLeft, SQLITE_STATIC); + sqlite3VdbeAddOp4(pParse->pVdbe, OP_String8, 0, 1, 0, type, 0); + sqlite3VdbeAddOp2(pParse->pVdbe, OP_ResultRow, 1, 1); + parsed = 1; + /* + * PRAGMA replication_get_master; + * Returns the host:port of the master, or NULL if replication has + * not started, or there is none. + */ + } else if (sqlite3StrNICmp(zLeft, "replication_get_master", 23) == 0) { + const char *address; + + sqlite3_mutex_enter(pDb->pBt->pBt->mutex); + address = pDb->pBt->pBt->master_address; + sqlite3_mutex_leave(pDb->pBt->pBt->mutex); + + sqlite3VdbeSetNumCols(pParse->pVdbe, 1); + sqlite3VdbeSetColName(pParse->pVdbe, 0, COLNAME_NAME, + zLeft, SQLITE_STATIC); + if (address) + sqlite3VdbeAddOp4(pParse->pVdbe, OP_String8, 0, 1, 0, address, 0); + else + sqlite3VdbeAddOp4(pParse->pVdbe, OP_String8, 0, 1, 0, "NULL", 0); + sqlite3VdbeAddOp2(pParse->pVdbe, OP_ResultRow, 1, 1); + parsed = 1; } /* Return semantics to match strcmp. */ |