/*
* LookupOperName
* Given a possibly-qualified operator name and exact input datatypes,
* look up the operator.
*
* Pass oprleft = InvalidOid for a prefix op, oprright = InvalidOid for
* a postfix op.
*
* If the operator name is not schema-qualified, it is sought in the current
* namespace search path.
*
* If the operator is not found, we return InvalidOid if noError is true,
* else raise an error. pstate and location are used only to report the
* error position; pass NULL/-1 if not available.
*/
Oid
LookupOperName(ParseState *pstate, List *opername, Oid oprleft, Oid oprright,
bool noError, int location)
{
Oid result;
result = OpernameGetOprid(opername, oprleft, oprright);
if (OidIsValid(result))
return result;
/* we don't use op_error here because only an exact match is wanted */
if (!noError)
{
char oprkind;
if (!OidIsValid(oprleft))
oprkind = 'l';
else if (!OidIsValid(oprright))
oprkind = 'r';
else
oprkind = 'b';
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_FUNCTION),
errmsg("operator does not exist: %s",
op_signature_string(opername, oprkind,
oprleft, oprright)),
errOmitLocation(true),
parser_errposition(pstate, location)));
}
return InvalidOid;
}
/* timestamptz_abstime()
* Convert timestamp with time zone to abstime.
*/
Datum
timestamptz_abstime(PG_FUNCTION_ARGS)
{
TimestampTz timestamp = PG_GETARG_TIMESTAMP(0);
AbsoluteTime result;
fsec_t fsec;
struct pg_tm tt,
*tm = &tt;
if (TIMESTAMP_IS_NOBEGIN(timestamp))
result = NOSTART_ABSTIME;
else if (TIMESTAMP_IS_NOEND(timestamp))
result = NOEND_ABSTIME;
else if (timestamp2tm(timestamp, NULL, tm, &fsec, NULL, NULL) == 0)
result = tm2abstime(tm, 0);
else
{
ereport(ERROR,
(errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE),
errmsg("timestamp out of range"),
errOmitLocation(true)));
result = INVALID_ABSTIME;
}
PG_RETURN_ABSOLUTETIME(result);
}
/* compatible_oper()
* given an opname and input datatypes, find a compatible binary operator
*
* This is tighter than oper() because it will not return an operator that
* requires coercion of the input datatypes (but binary-compatible operators
* are accepted). Otherwise, the semantics are the same.
*/
Operator
compatible_oper(ParseState *pstate, List *op, Oid arg1, Oid arg2,
bool noError, int location)
{
Operator optup;
Form_pg_operator opform;
/* oper() will find the best available match */
optup = oper(pstate, op, arg1, arg2, noError, location);
if (optup == (Operator) NULL)
return (Operator) NULL; /* must be noError case */
/* but is it good enough? */
opform = (Form_pg_operator) GETSTRUCT(optup);
if (IsBinaryCoercible(arg1, opform->oprleft) &&
IsBinaryCoercible(arg2, opform->oprright))
return optup;
/* nope... */
ReleaseOperator(optup);
if (!noError)
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_FUNCTION),
errmsg("operator requires run-time type coercion: %s",
op_signature_string(op, 'b', arg1, arg2)),
errOmitLocation(true),
parser_errposition(pstate, location)));
return (Operator) NULL;
}
/*
* reverse_ordering_oper - identify DESC sort operator (">") for a datatype
*
* On failure, return NULL if noError, else report a standard error
*/
Operator
reverse_ordering_oper(Oid argtype, bool noError)
{
TypeCacheEntry *typentry;
Oid oproid;
Operator optup;
/*
* Look for a ">" operator for the datatype. We require it to be an exact
* or binary-compatible match, since most callers are not prepared to cope
* with adding any run-time type coercion steps.
*
* Note: the search algorithm used by typcache.c ensures that if a ">"
* operator is returned, it will be consistent with the "=" operator
* returned by equality_oper. This is critical for sorting and grouping
* purposes.
*/
typentry = lookup_type_cache(argtype, TYPECACHE_GT_OPR);
oproid = typentry->gt_opr;
/*
* If the datatype is an array, then we can use array_gt ... but only if
* there is a suitable greater-than operator for the element type. (This
* check is not in the raw typcache.c code ... should it be?)
*/
if (oproid == ARRAY_GT_OP)
{
Oid elem_type = get_element_type(argtype);
if (OidIsValid(elem_type))
{
optup = reverse_ordering_oper(elem_type, true);
if (optup != NULL)
ReleaseOperator(optup);
else
oproid = InvalidOid; /* element type has no ">" */
}
else
oproid = InvalidOid; /* bogus array type? */
}
if (OidIsValid(oproid))
{
optup = fetch_op_tup(oproid, true);
return optup;
}
if (!noError)
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_FUNCTION),
errmsg("could not identify an ordering operator for type %s",
format_type_be(argtype)),
errhint("Use an explicit ordering operator or modify the query."),
errOmitLocation(true)));
return NULL;
}
/* abstime_timestamptz()
* Convert abstime to timestamp with time zone.
*/
Datum
abstime_timestamptz(PG_FUNCTION_ARGS)
{
AbsoluteTime abstime = PG_GETARG_ABSOLUTETIME(0);
TimestampTz result;
struct pg_tm tt,
*tm = &tt;
int tz;
char zone[MAXDATELEN + 1],
*tzn = zone;
switch (abstime)
{
case INVALID_ABSTIME:
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("cannot convert abstime \"invalid\" to timestamp"),
errOmitLocation(true)));
TIMESTAMP_NOBEGIN(result);
break;
case NOSTART_ABSTIME:
TIMESTAMP_NOBEGIN(result);
break;
case NOEND_ABSTIME:
TIMESTAMP_NOEND(result);
break;
default:
abstime2tm(abstime, &tz, tm, &tzn);
if (tm2timestamp(tm, 0, &tz, &result) != 0)
ereport(ERROR,
(errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE),
errmsg("timestamp out of range"),
errOmitLocation(true)));
break;
};
PG_RETURN_TIMESTAMP(result);
}
Datum
gp_persistent_set_relation_bufpool_kind_all(PG_FUNCTION_ARGS)
{
Relation pg_database;
HeapScanDesc scan;
HeapTuple tuple;
if (!gp_upgrade_mode)
ereport(ERROR,
(errcode(ERRCODE_GP_FEATURE_NOT_SUPPORTED),
errmsg("function is not supported"),
errOmitLocation(true)));
// UNDONE: Verify we are in some sort of single-user mode.
/*
* Special call here to scan the persistent meta-data structures so we are open for
* business and then we can add information.
*/
PersistentFileSysObj_BuildInitScan();
pg_database = heap_open(
DatabaseRelationId,
AccessShareLock);
scan = heap_beginscan(pg_database, SnapshotNow, 0, NULL);
while ((tuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
{
Form_pg_database form_pg_database =
(Form_pg_database)GETSTRUCT(tuple);
Oid dbOid;
Oid dattablespace;
dbOid = HeapTupleGetOid(tuple);
dattablespace = form_pg_database->dattablespace;
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"gp_persistent_set_relation_bufpool_kind_all: dbOid %u",
dbOid);
PersistentBuild_SetRelationBufpoolKind(
dattablespace,
dbOid);
}
heap_endscan(scan);
heap_close(pg_database, AccessShareLock);
PG_RETURN_INT32(1);
}
static
DemoUri *ParseDemoUri(const char *uri_str)
{
DemoUri *uri = (DemoUri *) palloc0(sizeof(DemoUri));
int protocol_len;
uri->path = NULL;
uri->protocol = NULL;
/*
* parse protocol
*/
char *post_protocol = strstr(uri_str, "://");
if(!post_protocol)
{
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
errmsg("invalid demo prot URI \'%s\'", uri_str),
errOmitLocation(true)));
}
protocol_len = post_protocol - uri_str;
uri->protocol = (char *) palloc0 (protocol_len + 1);
strncpy(uri->protocol, uri_str, protocol_len);
/* make sure there is more to the uri string */
if (strlen(uri_str) <= protocol_len)
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
errmsg("invalid demo prot URI \'%s\' : missing path", uri_str),
errOmitLocation(true)));
/*
* parse path
*/
uri->path = pstrdup(uri_str + protocol_len + strlen("://"));
return uri;
}
/*
* equality_oper - identify a suitable equality operator for a datatype
*
* On failure, return NULL if noError, else report a standard error
*/
Operator
equality_oper(Oid argtype, bool noError)
{
TypeCacheEntry *typentry;
Oid oproid;
Operator optup;
/*
* Look for an "=" operator for the datatype. We require it to be an
* exact or binary-compatible match, since most callers are not prepared
* to cope with adding any run-time type coercion steps.
*/
typentry = lookup_type_cache(argtype, TYPECACHE_EQ_OPR);
oproid = typentry->eq_opr;
/*
* If the datatype is an array, then we can use array_eq ... but only if
* there is a suitable equality operator for the element type. (This check
* is not in the raw typcache.c code ... should it be?)
*/
if (oproid == ARRAY_EQ_OP)
{
Oid elem_type = get_element_type(argtype);
if (OidIsValid(elem_type))
{
optup = equality_oper(elem_type, true);
if (optup != NULL)
ReleaseOperator(optup);
else
oproid = InvalidOid; /* element type has no "=" */
}
else
oproid = InvalidOid; /* bogus array type? */
}
if (OidIsValid(oproid))
{
optup = fetch_op_tup(oproid, true);
return optup;
}
if (!noError)
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_FUNCTION),
errmsg("could not identify an equality operator for type %s",
format_type_be(argtype)),
errOmitLocation(true)));
return NULL;
}
/*
* op_error - utility routine to complain about an unresolvable operator
*/
static void
op_error(ParseState *pstate, List *op, char oprkind,
Oid arg1, Oid arg2,
FuncDetailCode fdresult, int location)
{
if (fdresult == FUNCDETAIL_MULTIPLE)
ereport(ERROR,
(errcode(ERRCODE_AMBIGUOUS_FUNCTION),
errmsg("operator is not unique: %s",
op_signature_string(op, oprkind, arg1, arg2)),
errhint("Could not choose a best candidate operator. "
"You may need to add explicit type casts."),
errOmitLocation(true),
parser_errposition(pstate, location)));
else
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_FUNCTION),
errmsg("operator does not exist: %s",
op_signature_string(op, oprkind, arg1, arg2)),
errhint("No operator matches the given name and argument type(s). "
"You may need to add explicit type casts."),
errOmitLocation(true),
parser_errposition(pstate, location)));
}
Datum
reltime_interval(PG_FUNCTION_ARGS)
{
RelativeTime reltime = PG_GETARG_RELATIVETIME(0);
Interval *result;
int year,
month,
day;
result = (Interval *) palloc(sizeof(Interval));
switch (reltime)
{
case INVALID_RELTIME:
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("cannot convert reltime \"invalid\" to interval"),
errOmitLocation(true)));
result->time = 0;
result->day = 0;
result->month = 0;
break;
default:
#ifdef HAVE_INT64_TIMESTAMP
year = reltime / SECS_PER_YEAR;
reltime -= year * SECS_PER_YEAR;
month = reltime / (DAYS_PER_MONTH * SECS_PER_DAY);
reltime -= month * (DAYS_PER_MONTH * SECS_PER_DAY);
day = reltime / SECS_PER_DAY;
reltime -= day * SECS_PER_DAY;
result->time = (reltime * USECS_PER_SEC);
#else
TMODULO(reltime, year, SECS_PER_YEAR);
TMODULO(reltime, month, DAYS_PER_MONTH * SECS_PER_DAY);
TMODULO(reltime, day, SECS_PER_DAY);
result->time = reltime;
#endif
result->month = MONTHS_PER_YEAR * year + month;
result->day = day;
break;
}
PG_RETURN_INTERVAL_P(result);
}
/*
* tintervalrecv - converts external binary format to tinterval
*/
Datum
tintervalrecv(PG_FUNCTION_ARGS)
{
StringInfo buf = (StringInfo) PG_GETARG_POINTER(0);
TimeInterval tinterval;
tinterval = (TimeInterval) palloc(sizeof(TimeIntervalData));
tinterval->status = pq_getmsgint(buf, sizeof(tinterval->status));
if (!(tinterval->status == T_INTERVAL_INVAL ||
tinterval->status == T_INTERVAL_VALID))
ereport(ERROR,
(errcode(ERRCODE_INVALID_BINARY_REPRESENTATION),
errmsg("invalid status in external \"tinterval\" value"),
errOmitLocation(true)));
tinterval->data[0] = pq_getmsgint(buf, sizeof(tinterval->data[0]));
tinterval->data[1] = pq_getmsgint(buf, sizeof(tinterval->data[1]));
PG_RETURN_TIMEINTERVAL(tinterval);
}
void PersistentRelation_AddCreated(
RelFileNode *relFileNode,
ItemPointer persistentTid,
bool flushToXLog)
{
WRITE_PERSISTENT_STATE_ORDERED_LOCK_DECLARE;
PersistentFileSysObjName fsObjName;
ItemPointerData previousFreeTid;
int64 persistentSerialNum;
RelationDirEntry relationDirEntry;
Datum values[Natts_gp_persistent_relation_node];
if (RelFileNode_IsEmpty(relFileNode))
{
elog(ERROR, "Invalid RelFileNode (0,0,0)");
}
MemSet(&previousFreeTid, 0, sizeof(ItemPointerData));
if (!Persistent_BeforePersistenceWork())
{
elog(ERROR, "We can only add to persistent meta-data when special states");
}
/* Verify PersistentFileSysObj_BuildInitScan has been called */
PersistentRelation_VerifyInitScan();
PersistentFileSysObjName_SetRelationDir(
&fsObjName,
relFileNode,
is_tablespace_shared);
WRITE_PERSISTENT_STATE_ORDERED_LOCK;
relationDirEntry = PersistentRelation_CreateEntryUnderLock(relFileNode);
if (relationDirEntry == NULL)
{
/* If out of shared memory, no need to promote to PANIC. */
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
ereport(ERROR,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("Out of shared-memory for persistent relations"),
errhint("You may need to increase the gp_max_relations value"),
errOmitLocation(true)));
}
relationDirEntry->state = PersistentFileSysState_Created;
GpPersistentRelationNode_SetDatumValues(
values,
relFileNode->spcNode,
relFileNode->dbNode,
relFileNode->relNode,
PersistentFileSysState_Created,
/* reserved */ 0,
/* parentXid */ InvalidTransactionId,
/* persistentSerialNum */ 0,
&previousFreeTid,
is_tablespace_shared(relFileNode->spcNode));
PersistentFileSysObj_AddTuple(
PersistentFsObjType_RelationDir,
values,
flushToXLog,
persistentTid,
&persistentSerialNum);
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
if (Debug_persistent_print)
{
elog(Persistent_DebugPrintLevel(),
"Persistent relation: Add '%s', in state 'Created', serial number " INT64_FORMAT " at TID %s",
PersistentFileSysObjName_ObjectName(&fsObjName),
persistentSerialNum,
ItemPointerToString(persistentTid));
}
}
void PersistentDatabase_MarkJustInTimeCreatePending(
DbDirNode *dbDirNode,
ItemPointer persistentTid,
int64 *persistentSerialNum)
{
WRITE_PERSISTENT_STATE_ORDERED_LOCK_DECLARE;
PersistentFileSysObjName fsObjName;
DatabaseDirEntry databaseDirEntry;
SharedOidSearchAddResult addResult;
if (Persistent_BeforePersistenceWork())
{
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Skipping persistent database '%s' because we are before persistence work",
GetDatabasePath(
dbDirNode->database,
dbDirNode->tablespace));
return; // The initdb process will load the persistent table once we out of bootstrap mode.
}
PersistentDatabase_VerifyInitScan();
PersistentFileSysObjName_SetDatabaseDir(&fsObjName,dbDirNode->tablespace,dbDirNode->database,is_tablespace_shared);
WRITE_PERSISTENT_STATE_ORDERED_LOCK;
databaseDirEntry =
(DatabaseDirEntry)
SharedOidSearch_Find(
&persistentDatabaseSharedData->databaseDirSearchTable,
dbDirNode->database,
dbDirNode->tablespace);
if (databaseDirEntry != NULL)
{
/*
* An existence check should have been done before calling this routine.
*/
elog(ERROR, "Persistent database entry '%s' already exists in state '%s'",
GetDatabasePath(
dbDirNode->database,
dbDirNode->tablespace),
PersistentFileSysObjState_Name(databaseDirEntry->state));
}
addResult =
SharedOidSearch_Add(
&persistentDatabaseSharedData->databaseDirSearchTable,
dbDirNode->database,
dbDirNode->tablespace,
(SharedOidSearchObjHeader**)&databaseDirEntry);
if (addResult == SharedOidSearchAddResult_NoMemory)
{
/* If out of shared memory, no need to promote to PANIC. */
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
ereport(ERROR,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("Out of shared-memory for persistent databases"),
errhint("You may need to increase the gp_max_databases and "
"gp_max_tablespaces value"),
errOmitLocation(true)));
}
else if (addResult == SharedOidSearchAddResult_Exists)
elog(PANIC, "Persistent database entry '%s' already exists in state '%s'",
GetDatabasePath(
dbDirNode->database,
dbDirNode->tablespace),
PersistentFileSysObjState_Name(databaseDirEntry->state));
else
Assert(addResult == SharedOidSearchAddResult_Ok);
databaseDirEntry->state = PersistentFileSysState_JustInTimeCreatePending;
PersistentDatabase_AddTuple(
databaseDirEntry,
/* reserved */ 0,
/* parentXid */ InvalidTransactionId,
/* flushToXLog */ true);
*persistentTid = databaseDirEntry->persistentTid;
*persistentSerialNum = databaseDirEntry->persistentSerialNum;
/*
* This XLOG must be generated under the persistent write-lock.
*/
#ifdef MASTER_MIRROR_SYNC
mmxlog_log_create_database(dbDirNode->tablespace,
dbDirNode->database);
#endif
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
}
static int send_snmp_inform_or_trap(const GpErrorData * errorData, const char * subject, const char * severity)
{
netsnmp_session session, *ss = NULL;
netsnmp_pdu *pdu, *response;
int status;
char csysuptime[20];
static bool snmp_initialized = false;
static char myhostname[255]; /* gethostname usually is limited to 65 chars out, but make this big to be safe */
char *rawstring = NULL;
List *elemlist = NIL;
ListCell *l = NULL;
/*
* "inform" messages get a positive acknowledgement response from the SNMP manager.
* If it doesn't come, the message might be resent.
*
* "trap" messages are one-way, and we have no idea if the manager received it.
* But, it's faster and cheaper, and no need to retry. So some people might prefer it.
*/
bool inform = strcmp(gp_snmp_use_inform_or_trap,"inform") == 0;
if (gp_snmp_monitor_address == NULL || gp_snmp_monitor_address[0] == '\0')
{
static bool firsttime = 1;
ereport(firsttime ? LOG : DEBUG1,(errmsg("SNMP inform/trap alerts are disabled"),errOmitLocation(true)));
firsttime = false;
return -1;
}
/*
* SNMP managers are required to handle messages up to at least 484 bytes long, but I believe most existing
* managers support messages up to one packet (ethernet frame) in size, 1472 bytes.
*
* But, should we take that chance? Or play it safe and limit the message to 484 bytes?
*/
elog(DEBUG2,"send_snmp_inform_or_trap");
if (!snmp_initialized)
{
snmp_enable_stderrlog();
if (gp_snmp_debug_log != NULL && gp_snmp_debug_log[0] != '\0')
{
snmp_enable_filelog(gp_snmp_debug_log, 1);
//debug_register_tokens("ALL");
snmp_set_do_debugging(1);
}
/*
* Initialize the SNMP library. This also reads the MIB database.
*/
/* Add GPDB-MIB to the list to be loaded */
putenv("MIBS=+GPDB-MIB:SNMP-FRAMEWORK-MIB:SNMPv2-CONF:SNMPv2-TC:SNMPv2-TC");
init_snmp("sendalert");
snmp_initialized = true;
{
char portnum[16];
myhostname[0] = '\0';
if (gethostname(myhostname, sizeof(myhostname)) == 0)
{
strcat(myhostname,":");
pg_ltoa(PostPortNumber,portnum);
strcat(myhostname,portnum);
}
}
}
/*
* Trap/Inform messages always start with the system up time. (SysUpTime.0)
*
* This presumably would be the uptime of GPDB, not the machine it is running on, I think.
*
* Use Postmaster's "MyStartTime" as a way to get that.
*/
sprintf(csysuptime, "%ld", (long)(time(NULL) - MyStartTime));
/*
// ERRCODE_DISK_FULL could be reported vi rbmsMIB rdbmsTraps rdbmsOutOfSpace trap.
// But it appears we never generate that error?
// ERRCODE_ADMIN_SHUTDOWN means SysAdmin aborted somebody's request. Not interesting?
// ERRCODE_CRASH_SHUTDOWN sounds interesting, but I don't see that we ever generate it.
// ERRCODE_CANNOT_CONNECT_NOW means we are starting up, shutting down, in recovery, or Too many users are logged on.
// abnormal database system shutdown
*/
/*
* The gpdbAlertSeverity is a crude attempt to classify some of these messages based on severity,
* where OK means everything is running normal, Down means everything is shut down, degraded would be
* for times when some segments are down, but the system is up, The others are maybe useful in the future
*
* gpdbSevUnknown(0),
* gpdbSevOk(1),
* gpdbSevWarning(2),
* gpdbSevError(3),
* gpdbSevFatal(4),
* gpdbSevPanic(5),
* gpdbSevSystemDegraded(6),
* gpdbSevSystemDown(7)
*/
char detail[MAX_ALERT_STRING+1];
snprintf(detail, MAX_ALERT_STRING, "%s", errorData->error_detail);
detail[127] = '\0';
char sqlstmt[MAX_ALERT_STRING+1];
char * sqlstmtp = errorData->debug_query_string;
if (sqlstmtp == NULL || sqlstmtp[0] == '\0')
sqlstmtp = errorData->internal_query;
if (sqlstmtp == NULL)
sqlstmtp = "";
snprintf(sqlstmt, MAX_ALERT_STRING, "%s", sqlstmtp);
sqlstmt[MAX_ALERT_STRING] = '\0';
/* Need a modifiable copy of To list */
rawstring = pstrdup(gp_snmp_monitor_address);
/* Parse string into list of identifiers */
if (!SplitMailString(rawstring, ',', &elemlist))
{
/* syntax error in list */
ereport(LOG,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("invalid list syntax for \"gp_snmp_monitor_address\"")));
return -1;
}
/*
* This session is just a template, and doesn't need to be connected.
* It is used by snmp_add(), which copies this info, opens the new session, and assigns the transport.
*/
snmp_sess_init( &session ); /* Initialize session to default values */
session.version = SNMP_VERSION_2c;
session.timeout = SNMP_DEFAULT_TIMEOUT;
session.retries = SNMP_DEFAULT_RETRIES;
session.remote_port = 162; // I think this isn't used by net-snmp any more.
/*if (strchr(session.peername,':')==NULL)
strcat(session.peername,":162");*/
session.community = (u_char *)gp_snmp_community;
session.community_len = strlen((const char *)session.community); // SNMP_DEFAULT_COMMUNITY_LEN means "public"
session.callback_magic = NULL;
foreach(l, elemlist)
{
char *cur_gp_snmp_monitor_address = (char *) lfirst(l);
if (cur_gp_snmp_monitor_address == NULL || cur_gp_snmp_monitor_address[0] == '\0')
continue;
session.peername = cur_gp_snmp_monitor_address;
/*
* If we try to "snmp_open( &session )", net-snmp will set up a connection to that
* endpoint on port 161, assuming we are the network monitor, and the other side is an agent.
*
* But we are pretending to be an agent, sending traps to the NM, so we don't need this.
*/
/*if (!snmp_open( &session )) // Don't open the session here!
{
const char *str;
int xerr;
xerr = snmp_errno;
str = snmp_api_errstring(xerr);
elog(LOG, "snmp_open: %s", str);
return -1;
}*/
/*
* This call copies the info from "session" to "ss", assigns the transport, and opens the session.
* We must specify "snmptrap" so the transport will know we want port 162 by default.
*/
ss = snmp_add(&session,
netsnmp_transport_open_client("snmptrap", cur_gp_snmp_monitor_address),
NULL, NULL);
if (ss == NULL) {
/*
* diagnose netsnmp_transport_open_client and snmp_add errors with
* the input netsnmp_session pointer
*/
{
char *err;
snmp_error(&session, NULL, NULL, &err);
elog(LOG, "send_alert snmp_add of %s failed: %s", cur_gp_snmp_monitor_address, err);
free(err);
}
return -1;
}
/*
* We need to create the pdu each time, as it gets freed when we send a trap.
*/
pdu = snmp_pdu_create(inform ? SNMP_MSG_INFORM : SNMP_MSG_TRAP2);
if (!pdu)
{
const char *str;
int xerr;
xerr = snmp_errno;
str = snmp_api_errstring(xerr);
elog(LOG, "Failed to create notification PDU: %s", str);
return -1;
}
/*
* Trap/Inform messages always start with the system up time. (SysUpTime.0)
* We use Postmaster's "MyStartTime" as a way to get that.
*/
snmp_add_var(pdu, objid_sysuptime,
sizeof(objid_sysuptime) / sizeof(oid),
't', (const char *)csysuptime);
#if 0
/*
* In the future, we might want to send RDBMS-MIB::rdbmsStateChange when the system becomes unavailable or
* partially unavailable. This code, which is not currently used, shows how to build the pdu for
* that trap.
*/
/* {iso(1) identified-organization(3) dod(6) internet(1) mgmt(2) mib-2(1) rdbmsMIB(39) rdbmsTraps(2) rdbmsStateChange(1)} */
snmp_add_var(pdu, objid_snmptrap,
sizeof(objid_snmptrap) / sizeof(oid),
'o', "1.3.6.1.2.1.39.2.1"); // rdbmsStateChange
snmp_add_var(pdu, objid_rdbmsrelstate,
sizeof(objid_rdbmsrelstate) / sizeof(oid),
'i', "5"); // 4 = restricted, 5 = unavailable
#endif
/* {iso(1) identified-organization(3) dod(6) internet(1) private(4) enterprises(1) gpdbMIB(31327) gpdbTraps(5) gpdbTrapsList(0) gpdbAlert(1)} */
/*
* We could specify this trap oid by name, rather than numeric oid, but then if the GPDB-MIB wasn't
* found, we'd get an error. Using the numeric oid means we can still work without the MIB loaded.
*/
snmp_add_var(pdu, objid_snmptrap,
sizeof(objid_snmptrap) / sizeof(oid),
'o', "1.3.6.1.4.1.31327.5.0.1"); // gpdbAlert
snmp_add_var(pdu, objid_gpdbAlertMsg,
sizeof(objid_gpdbAlertMsg) / sizeof(oid),
's', subject); // SnmpAdminString = UTF-8 text
snmp_add_var(pdu, objid_gpdbAlertSeverity,
sizeof(objid_gpdbAlertSeverity) / sizeof(oid),
'i', (char *)severity);
snmp_add_var(pdu, objid_gpdbAlertSqlstate,
sizeof(objid_gpdbAlertSqlstate) / sizeof(oid),
's', errorData->sql_state);
snmp_add_var(pdu, objid_gpdbAlertDetail,
sizeof(objid_gpdbAlertDetail) / sizeof(oid),
's', detail); // SnmpAdminString = UTF-8 text
snmp_add_var(pdu, objid_gpdbAlertSqlStmt,
sizeof(objid_gpdbAlertSqlStmt) / sizeof(oid),
's', sqlstmt); // SnmpAdminString = UTF-8 text
snmp_add_var(pdu, objid_gpdbAlertSystemName,
sizeof(objid_gpdbAlertSystemName) / sizeof(oid),
's', myhostname); // SnmpAdminString = UTF-8 text
elog(DEBUG2,"ready to send to %s",cur_gp_snmp_monitor_address);
if (inform)
status = snmp_synch_response(ss, pdu, &response);
else
status = snmp_send(ss, pdu) == 0;
elog(DEBUG2,"send, status %d",status);
if (status != STAT_SUCCESS)
{
/* Something went wrong */
if (ss)
{
char *err;
snmp_error(ss, NULL, NULL, &err);
elog(LOG, "sendalert failed to send %s: %s", inform ? "inform" : "trap", err);
free(err);
}
else
{
elog(LOG, "sendalert failed to send %s: %s", inform ? "inform" : "trap", "Something went wrong");
}
if (!inform)
snmp_free_pdu(pdu);
}
else if (inform)
snmp_free_pdu(response);
snmp_close(ss);
ss = NULL;
}
void PersistentRelation_MarkCreatePending(
RelFileNode *relFileNode,
ItemPointer persistentTid,
int64 *persistentSerialNum,
bool flushToXLog)
{
WRITE_PERSISTENT_STATE_ORDERED_LOCK_DECLARE;
PersistentFileSysObjName fsObjName;
RelationDirEntry relationDirEntry;
ItemPointerData previousFreeTid;
Datum values[Natts_gp_persistent_relation_node];
if (RelFileNode_IsEmpty(relFileNode))
{
elog(ERROR, "Invalid RelFileNode (0,0,0)");
}
MemSet(&previousFreeTid, 0, sizeof(ItemPointerData));
if (Persistent_BeforePersistenceWork())
{
if (Debug_persistent_print)
{
elog(Persistent_DebugPrintLevel(),
"Skipping persistent relation '%s' because we are before persistence work",
relpath(*relFileNode));
}
*persistentSerialNum = 0;
/*
* The initdb process will load the persistent table once we out
* of bootstrap mode.
*/
return;
}
PersistentRelation_VerifyInitScan();
PersistentFileSysObjName_SetRelationDir(
&fsObjName,
relFileNode,
is_tablespace_shared);
WRITE_PERSISTENT_STATE_ORDERED_LOCK;
relationDirEntry = PersistentRelation_CreateEntryUnderLock(relFileNode);
if (relationDirEntry == NULL)
{
/* If out of shared memory, no need to promote to PANIC. */
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
ereport(ERROR,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("Out of shared-memory for persistent relations"),
errhint("You may need to increase the gp_max_relations value"),
errOmitLocation(true)));
}
relationDirEntry->state = PersistentFileSysState_CreatePending;
GpPersistentRelationNode_SetDatumValues(
values,
relFileNode->spcNode,
relFileNode->dbNode,
relFileNode->relNode,
PersistentFileSysState_CreatePending,
/* reserved */ 0,
/* parentXid */ GetTopTransactionId(),
/* persistentSerialNum */ 0, // This will be set by PersistentFileSysObj_AddTuple.
&previousFreeTid,
is_tablespace_shared(relFileNode->spcNode));
PersistentFileSysObj_AddTuple(
PersistentFsObjType_RelationDir,
values,
flushToXLog,
&relationDirEntry->persistentTid,
&relationDirEntry->persistentSerialNum);
*persistentTid = relationDirEntry->persistentTid;
*persistentSerialNum = relationDirEntry->persistentSerialNum;
/*
* This XLOG must be generated under the persistent write-lock.
*/
#ifdef MASTER_MIRROR_SYNC
mmxlog_log_create_relation(
relFileNode->spcNode,
relFileNode->dbNode,
relFileNode->relNode,
persistentTid, persistentSerialNum);
#endif
/*
* MPP-18228
* To make adding 'Create Pending' entry to persistent table and
* adding to the PendingDelete list atomic
*/
PendingDelete_AddCreatePendingEntryWrapper(
&fsObjName,
persistentTid,
*persistentSerialNum);
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
}
/*
* CheckMyDatabase -- fetch information from the pg_database entry for our DB
*/
static void
CheckMyDatabase(const char *name, bool am_superuser)
{
HeapTuple tup;
Form_pg_database dbform;
/* Fetch our real pg_database row */
tup = SearchSysCache(DATABASEOID,
ObjectIdGetDatum(MyDatabaseId),
0, 0, 0);
if (!HeapTupleIsValid(tup))
elog(ERROR, "cache lookup failed for database %u", MyDatabaseId);
dbform = (Form_pg_database) GETSTRUCT(tup);
/* This recheck is strictly paranoia */
if (strcmp(name, NameStr(dbform->datname)) != 0)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" has disappeared from pg_database",
name),
errdetail("Database OID %u now seems to belong to \"%s\".",
MyDatabaseId, NameStr(dbform->datname))));
/*
* Check permissions to connect to the database.
*
* These checks are not enforced when in standalone mode, so that there is
* a way to recover from disabling all access to all databases, for
* example "UPDATE pg_database SET datallowconn = false;".
*
* We do not enforce them for the autovacuum worker processes either.
*/
if (IsUnderPostmaster && !IsAutoVacuumProcess())
{
/*
* Check that the database is currently allowing connections.
* (exception during upgrade_mode)
*/
if (gp_upgrade_mode && !dbform->datallowconn)
elog(INFO, "Connecting to no-connection db in upgrade mode.");
if (!dbform->datallowconn && !gp_upgrade_mode)
ereport(FATAL,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("database \"%s\" is not currently accepting connections",
name),
errOmitLocation(true),
errSendAlert(false)));
/*
* Check privilege to connect to the database. (The am_superuser test
* is redundant, but since we have the flag, might as well check it
* and save a few cycles.)
*/
if (!am_superuser &&
pg_database_aclcheck(MyDatabaseId, GetUserId(),
ACL_CONNECT) != ACLCHECK_OK)
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("permission denied for database \"%s\"", name),
errdetail("User does not have CONNECT privilege."),
errOmitLocation(true),
errSendAlert(false)));
/*
* Check connection limit for this database.
*
* There is a race condition here --- we create our PGPROC before
* checking for other PGPROCs. If two backends did this at about the
* same time, they might both think they were over the limit, while
* ideally one should succeed and one fail. Getting that to work
* exactly seems more trouble than it is worth, however; instead we
* just document that the connection limit is approximate.
*/
if (dbform->datconnlimit >= 0 &&
!am_superuser &&
CountDBBackends(MyDatabaseId) > dbform->datconnlimit)
ereport(FATAL,
(errcode(ERRCODE_TOO_MANY_CONNECTIONS),
errmsg("too many connections for database \"%s\"",
name),
errOmitLocation(true),
errSendAlert(false)));
}
/*
* OK, we're golden. Next to-do item is to save the encoding info out of
* the pg_database tuple.
*/
if (GpIdentity.segindex == UNINITIALIZED_GP_IDENTITY_VALUE ||
GpIdentity.segindex == MASTER_CONTENT_ID)
SetDatabaseEncoding(dbform->encoding);
else
SetDatabaseEncoding(MyProcPort->encoding);
/* Record it as a GUC internal option, too */
SetConfigOption("server_encoding", GetDatabaseEncodingName(),
PGC_INTERNAL, PGC_S_OVERRIDE);
/* If we have no other source of client_encoding, use server encoding */
SetConfigOption("client_encoding", GetDatabaseEncodingName(),
PGC_BACKEND, PGC_S_DEFAULT);
/* Use the right encoding in translated messages */
#ifdef ENABLE_NLS
pg_bind_textdomain_codeset(textdomain(NULL));
#endif
/*
* Lastly, set up any database-specific configuration variables.
*/
if (IsUnderPostmaster)
{
Datum datum;
bool isnull;
datum = SysCacheGetAttr(DATABASEOID, tup, Anum_pg_database_datconfig,
&isnull);
if (!isnull)
{
ArrayType *a = DatumGetArrayTypeP(datum);
ProcessGUCArray(a, PGC_S_DATABASE);
}
}
ReleaseSysCache(tup);
}
/* --------------------------------
* InitPostgres
* Initialize POSTGRES.
*
* The database can be specified by name, using the in_dbname parameter, or by
* OID, using the dboid parameter. In the latter case, the computed database
* name is passed out to the caller as a palloc'ed string in out_dbname.
*
* In bootstrap mode no parameters are used.
*
* The return value indicates whether the userID is a superuser. (That
* can only be tested inside a transaction, so we want to do it during
* the startup transaction rather than doing a separate one in postgres.c.)
*
* As of PostgreSQL 8.2, we expect InitProcess() was already called, so we
* already have a PGPROC struct ... but it's not filled in yet.
*
* Note:
* Be very careful with the order of calls in the InitPostgres function.
* --------------------------------
*/
bool
InitPostgres(const char *in_dbname, Oid dboid, const char *username,
char **out_dbname)
{
bool bootstrap = IsBootstrapProcessingMode();
bool autovacuum = IsAutoVacuumProcess();
bool am_superuser;
char *fullpath;
char dbname[NAMEDATALEN];
/*
* Set up the global variables holding database id and path. But note we
* won't actually try to touch the database just yet.
*
* We take a shortcut in the bootstrap case, otherwise we have to look up
* the db name in pg_database.
*/
if (bootstrap)
{
MyDatabaseId = TemplateDbOid;
MyDatabaseTableSpace = DEFAULTTABLESPACE_OID;
}
else
{
/*
* Find tablespace of the database we're about to open. Since we're
* not yet up and running we have to use one of the hackish
* FindMyDatabase variants, which look in the flat-file copy of
* pg_database.
*
* If the in_dbname param is NULL, lookup database by OID.
*/
if (in_dbname == NULL)
{
if (!FindMyDatabaseByOid(dboid, dbname, &MyDatabaseTableSpace))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database %u does not exist", dboid),
errSendAlert(false)));
MyDatabaseId = dboid;
/* pass the database name to the caller */
*out_dbname = pstrdup(dbname);
}
else
{
if (!FindMyDatabase(in_dbname, &MyDatabaseId, &MyDatabaseTableSpace))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist",
in_dbname),
errOmitLocation(true),
errSendAlert(false)));
/* our database name is gotten from the caller */
strlcpy(dbname, in_dbname, NAMEDATALEN);
}
}
fullpath = GetDatabasePath(MyDatabaseId, MyDatabaseTableSpace);
SetDatabasePath(fullpath);
/*
* Finish filling in the PGPROC struct, and add it to the ProcArray. (We
* need to know MyDatabaseId before we can do this, since it's entered
* into the PGPROC struct.)
*
* Once I have done this, I am visible to other backends!
*/
InitProcessPhase2();
// get temporary directory for QD
if (!bootstrap && Gp_role == GP_ROLE_DISPATCH)
{
if (get_tmpdir_from_rm)
{
getLocalTmpDirFromMasterRM();
}
else
{
getLocalTmpDirFromMasterConfig(gp_session_id);
elog(LOG, "getLocalTmpDirFromMasterConfig session_id:%d tmpdir:%s", gp_session_id, LocalTempPath);
}
}
/* Initialize SessionState entry */
SessionState_Init();
/* Initialize memory protection */
GPMemoryProtect_Init();
/*
* Initialize my entry in the shared-invalidation manager's array of
* per-backend data.
*
* Sets up MyBackendId, a unique backend identifier.
*/
MyBackendId = InvalidBackendId;
SharedInvalBackendInit(false);
if (MyBackendId > MaxBackends || MyBackendId <= 0)
elog(FATAL, "bad backend id: %d", MyBackendId);
/* Now that we have a BackendId, we can participate in ProcSignal */
ProcSignalInit(MyBackendId);
/*
* bufmgr needs another initialization call too
*/
InitBufferPoolBackend();
/*
* Initialize local process's access to XLOG. In bootstrap case we may
* skip this since StartupXLOG() was run instead.
*/
if (!bootstrap)
InitXLOGAccess();
/*
* Initialize the relation cache and the system catalog caches. Note that
* no catalog access happens here; we only set up the hashtable structure.
* We must do this before starting a transaction because transaction abort
* would try to touch these hashtables.
*/
RelationCacheInitialize();
InitCatalogCache();
/* Initialize portal manager */
EnablePortalManager();
/* Initialize stats collection --- must happen before first xact */
if (!bootstrap)
pgstat_initialize();
/*
* Set up process-exit callback to do pre-shutdown cleanup. This has to
* be after we've initialized all the low-level modules like the buffer
* manager, because during shutdown this has to run before the low-level
* modules start to close down. On the other hand, we want it in place
* before we begin our first transaction --- if we fail during the
* initialization transaction, as is entirely possible, we need the
* AbortTransaction call to clean up.
*/
on_shmem_exit(ShutdownPostgres, 0);
/*
* Start a new transaction here before first access to db, and get a
* snapshot. We don't have a use for the snapshot itself, but we're
* interested in the secondary effect that it sets RecentGlobalXmin.
*/
if (!bootstrap)
{
StartTransactionCommand();
(void) GetTransactionSnapshot();
}
/*
* Now that we have a transaction, we can take locks. Take a writer's
* lock on the database we are trying to connect to. If there is a
* concurrently running DROP DATABASE on that database, this will block us
* until it finishes (and has updated the flat file copy of pg_database).
*
* Note that the lock is not held long, only until the end of this startup
* transaction. This is OK since we are already advertising our use of
* the database in the PGPROC array; anyone trying a DROP DATABASE after
* this point will see us there.
*
* Note: use of RowExclusiveLock here is reasonable because we envision
* our session as being a concurrent writer of the database. If we had a
* way of declaring a session as being guaranteed-read-only, we could use
* AccessShareLock for such sessions and thereby not conflict against
* CREATE DATABASE.
*/
if (!bootstrap)
{
if (MyDatabaseId == TemplateDbOid)
LockSharedObject(DatabaseRelationId, MyDatabaseId, 0,
AccessShareLock);
else
LockSharedObject(DatabaseRelationId, MyDatabaseId, 0,
RowExclusiveLock);
}
/*
* Recheck the flat file copy of pg_database to make sure the target
* database hasn't gone away. If there was a concurrent DROP DATABASE,
* this ensures we will die cleanly without creating a mess.
*/
if (!bootstrap)
{
Oid dbid2;
Oid tsid2;
if (!FindMyDatabase(dbname, &dbid2, &tsid2) ||
dbid2 != MyDatabaseId || tsid2 != MyDatabaseTableSpace)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist",
dbname),
errdetail("It seems to have just been dropped or renamed.")));
}
/*
* Now we should be able to access the database directory safely. Verify
* it's there and looks reasonable.
*/
if (!bootstrap)
{
if (access(fullpath, F_OK) == -1)
{
if (errno == ENOENT)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist",
dbname),
errdetail("The database subdirectory \"%s\" is missing.",
fullpath)));
else
ereport(FATAL,
(errcode_for_file_access(),
errmsg("could not access directory \"%s\": %m",
fullpath)));
}
ValidatePgVersion(fullpath);
}
/*
* It's now possible to do real access to the system catalogs.
*
* Load relcache entries for the system catalogs. This must create at
* least the minimum set of "nailed-in" cache entries.
*/
RelationCacheInitializePhase2();
/*
* Figure out our postgres user id, and see if we are a superuser.
*
* In standalone mode and in the autovacuum process, we use a fixed id,
* otherwise we figure it out from the authenticated user name.
*/
if (bootstrap || autovacuum)
{
InitializeSessionUserIdStandalone();
am_superuser = true;
}
else if (!IsUnderPostmaster)
{
InitializeSessionUserIdStandalone();
am_superuser = true;
if (!ThereIsAtLeastOneRole())
ereport(WARNING,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("no roles are defined in this database system"),
errhint("You should immediately run CREATE USER \"%s\" CREATEUSER;.",
username)));
}
else
{
/* normal multiuser case */
InitializeSessionUserId(username);
am_superuser = superuser();
}
/* set up ACL framework (so CheckMyDatabase can check permissions) */
initialize_acl();
/*
* Read the real pg_database row for our database, check permissions and
* set up database-specific GUC settings. We can't do this until all the
* database-access infrastructure is up. (Also, it wants to know if the
* user is a superuser, so the above stuff has to happen first.)
*/
if (!bootstrap)
CheckMyDatabase(dbname, am_superuser);
/*
* Check a normal user hasn't connected to a superuser reserved slot.
*/
if (!am_superuser &&
ReservedBackends > 0 &&
!HaveNFreeProcs(ReservedBackends))
ereport(FATAL,
(errcode(ERRCODE_TOO_MANY_CONNECTIONS),
errmsg("connection limit exceeded for non-superusers"),
errOmitLocation(true),
errSendAlert(true)));
/*
* Initialize various default states that can't be set up until we've
* selected the active user and gotten the right GUC settings.
*/
/* set default namespace search path */
InitializeSearchPath();
/* initialize client encoding */
InitializeClientEncoding();
/* report this backend in the PgBackendStatus array */
if (!bootstrap)
pgstat_bestart();
/*
* MPP package setup
*
* Primary function is to establish connctions to the qExecs.
* This is SKIPPED when the database is in bootstrap mode or
* Is not UnderPostmaster.
*/
if (!bootstrap && IsUnderPostmaster)
{
cdb_setup();
on_proc_exit( cdb_cleanup, 0 );
}
/* close the transaction we started above */
if (!bootstrap)
CommitTransactionCommand();
return am_superuser;
}
/* ----------------------------------------------------------------
* ExecInsert
*
* INSERTs have to add the tuple into
* the base relation and insert appropriate tuples into the
* index relations.
* Insert can be part of an update operation when
* there is a preceding SplitUpdate node.
* ----------------------------------------------------------------
*/
void
ExecInsert(TupleTableSlot *slot,
DestReceiver *dest,
EState *estate,
PlanGenerator planGen,
bool isUpdate)
{
void *tuple = NULL;
ResultRelInfo *resultRelInfo = NULL;
Relation resultRelationDesc = NULL;
Oid newId = InvalidOid;
TupleTableSlot *partslot = NULL;
AOTupleId aoTupleId = AOTUPLEID_INIT;
bool rel_is_heap = false;
bool rel_is_aorows = false;
bool rel_is_external = false;
bool rel_is_parquet = false;
/*
* get information on the (current) result relation
*/
if (estate->es_result_partitions)
{
resultRelInfo = slot_get_partition(slot, estate);
estate->es_result_relation_info = resultRelInfo;
if (NULL != resultRelInfo->ri_parquetSendBack)
{
/*
* The Parquet part we are about to insert into
* has sendBack information. This means we're inserting into the
* part twice, which is not supported. Error out (GPSQL-2291)
*/
Assert(gp_parquet_insert_sort);
ereport(ERROR, (errcode(ERRCODE_CDB_FEATURE_NOT_YET),
errmsg("Cannot insert out-of-order tuples in parquet partitions"),
errhint("Sort the data on the partitioning key(s) before inserting"),
errOmitLocation(true)));
}
/*
* Check if we need to close the last parquet partition we
* inserted into (GPSQL-2291).
*/
Oid new_part_oid = resultRelInfo->ri_RelationDesc->rd_id;
if (gp_parquet_insert_sort &&
PLANGEN_OPTIMIZER == planGen &&
InvalidOid != estate->es_last_parq_part &&
new_part_oid != estate->es_last_parq_part)
{
Assert(NULL != estate->es_partition_state->result_partition_hash);
ResultPartHashEntry *entry = hash_search(estate->es_partition_state->result_partition_hash,
&estate->es_last_parq_part,
HASH_FIND,
NULL /* found */);
Assert(NULL != entry);
Assert(entry->offset < estate->es_num_result_relations);
ResultRelInfo *oldResultRelInfo = & estate->es_result_relations[entry->offset];
elog(DEBUG1, "Switching from old part oid=%d name=[%s] to new part oid=%d name=[%s]",
estate->es_last_parq_part,
oldResultRelInfo->ri_RelationDesc->rd_rel->relname.data,
new_part_oid,
resultRelInfo->ri_RelationDesc->rd_rel->relname.data);
/*
* We are opening a new partition, and the last partition we
* inserted into was a Parquet part. Let's close the old
* parquet insert descriptor to free the memory before
* opening the new one.
*/
ParquetInsertDescData *oldInsertDesc = oldResultRelInfo->ri_parquetInsertDesc;
/*
* We need to preserve the "sendback" information that needs to be
* sent back to the QD process from this part.
* Compute it here, and store it for later use.
*/
QueryContextDispatchingSendBack sendback =
CreateQueryContextDispatchingSendBack(1);
sendback->relid = RelationGetRelid(oldResultRelInfo->ri_RelationDesc);
oldInsertDesc->sendback = sendback;
parquet_insert_finish(oldInsertDesc);
/* Store the sendback information in the resultRelInfo for this part */
oldResultRelInfo->ri_parquetSendBack = sendback;
/* Record in the resultRelInfo that we closed the parquet insert descriptor */
oldResultRelInfo->ri_parquetInsertDesc = NULL;
/* Reset the last parquet part Oid, it's now closed */
estate->es_last_parq_part = InvalidOid;
}
}
else
{
resultRelInfo = estate->es_result_relation_info;
}
Assert (!resultRelInfo->ri_projectReturning);
resultRelationDesc = resultRelInfo->ri_RelationDesc;
rel_is_heap = RelationIsHeap(resultRelationDesc);
rel_is_aorows = RelationIsAoRows(resultRelationDesc);
rel_is_external = RelationIsExternal(resultRelationDesc);
rel_is_parquet = RelationIsParquet(resultRelationDesc);
/* Validate that insert is not part of an non-allowed update operation. */
if (isUpdate && (rel_is_aorows || rel_is_parquet))
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("Append-only tables are not updatable. Operation not permitted."),
errOmitLocation(true)));
}
partslot = reconstructMatchingTupleSlot(slot, resultRelInfo);
if (rel_is_heap || rel_is_external)
{
tuple = ExecFetchSlotHeapTuple(partslot);
}
else if (rel_is_aorows)
{
tuple = ExecFetchSlotMemTuple(partslot, false);
}
else if (rel_is_parquet)
{
tuple = NULL;
}
Assert( partslot != NULL );
Assert( rel_is_parquet || (tuple != NULL));
/* Execute triggers in Planner-generated plans */
if (planGen == PLANGEN_PLANNER)
{
/* BEFORE ROW INSERT Triggers */
if (resultRelInfo->ri_TrigDesc &&
resultRelInfo->ri_TrigDesc->n_before_row[TRIGGER_EVENT_INSERT] > 0)
{
HeapTuple newtuple;
/* NYI */
if(rel_is_parquet)
elog(ERROR, "triggers are not supported on tables that use column-oriented storage");
newtuple = ExecBRInsertTriggers(estate, resultRelInfo, tuple);
if (newtuple == NULL) /* "do nothing" */
{
return;
}
if (newtuple != tuple) /* modified by Trigger(s) */
{
/*
* Put the modified tuple into a slot for convenience of routines
* below. We assume the tuple was allocated in per-tuple memory
* context, and therefore will go away by itself. The tuple table
* slot should not try to clear it.
*/
TupleTableSlot *newslot = estate->es_trig_tuple_slot;
if (newslot->tts_tupleDescriptor != partslot->tts_tupleDescriptor)
ExecSetSlotDescriptor(newslot, partslot->tts_tupleDescriptor);
ExecStoreGenericTuple(newtuple, newslot, false);
newslot->tts_tableOid = partslot->tts_tableOid; /* for constraints */
tuple = newtuple;
partslot = newslot;
}
}
}
/*
* Check the constraints of the tuple
*/
if (resultRelationDesc->rd_att->constr &&
planGen == PLANGEN_PLANNER)
{
ExecConstraints(resultRelInfo, partslot, estate);
}
/*
* insert the tuple
*
* Note: heap_insert returns the tid (location) of the new tuple in the
* t_self field.
*
* NOTE: for append-only relations we use the append-only access methods.
*/
if (rel_is_aorows)
{
if (resultRelInfo->ri_aoInsertDesc == NULL)
{
ResultRelSegFileInfo *segfileinfo = NULL;
/* Set the pre-assigned fileseg number to insert into */
ResultRelInfoSetSegFileInfo(resultRelInfo, estate->es_result_segfileinfos);
segfileinfo = (ResultRelSegFileInfo *)list_nth(resultRelInfo->ri_aosegfileinfos, GetQEIndex());
resultRelInfo->ri_aoInsertDesc =
appendonly_insert_init(resultRelationDesc,
segfileinfo);
}
appendonly_insert(resultRelInfo->ri_aoInsertDesc, tuple, &newId, &aoTupleId);
}
else if (rel_is_external)
{
/* Writable external table */
if (resultRelInfo->ri_extInsertDesc == NULL)
resultRelInfo->ri_extInsertDesc = external_insert_init(
resultRelationDesc, 0);
newId = external_insert(resultRelInfo->ri_extInsertDesc, tuple);
}
else if(rel_is_parquet)
{
/* If there is no parquet insert descriptor, create it now. */
if (resultRelInfo->ri_parquetInsertDesc == NULL)
{
ResultRelSegFileInfo *segfileinfo = NULL;
ResultRelInfoSetSegFileInfo(resultRelInfo, estate->es_result_segfileinfos);
segfileinfo = (ResultRelSegFileInfo *)list_nth(resultRelInfo->ri_aosegfileinfos, GetQEIndex());
resultRelInfo->ri_parquetInsertDesc = parquet_insert_init(resultRelationDesc, segfileinfo);
/*
* Just opened a new parquet partition for insert. Save the Oid
* in estate, so that we can close it when switching to a
* new partition (GPSQL-2291)
*/
elog(DEBUG1, "Saving es_last_parq_part. Old=%d, new=%d", estate->es_last_parq_part, resultRelationDesc->rd_id);
estate->es_last_parq_part = resultRelationDesc->rd_id;
}
newId = parquet_insert(resultRelInfo->ri_parquetInsertDesc, partslot);
}
else
{
Insist(rel_is_heap);
newId = heap_insert(resultRelationDesc,
tuple,
estate->es_snapshot->curcid,
true, true, GetCurrentTransactionId());
}
IncrAppended();
(estate->es_processed)++;
(resultRelInfo->ri_aoprocessed)++;
estate->es_lastoid = newId;
partslot->tts_tableOid = RelationGetRelid(resultRelationDesc);
if (rel_is_aorows || rel_is_parquet)
{
/* NOTE: Current version does not support index upon parquet table. */
/*
* insert index entries for AO Row-Store tuple
*/
if (resultRelInfo->ri_NumIndices > 0 && !rel_is_parquet)
ExecInsertIndexTuples(partslot, (ItemPointer)&aoTupleId, estate, false);
}
else
{
/* Use parttuple for index update in case this is an indexed heap table. */
TupleTableSlot *xslot = partslot;
void *xtuple = tuple;
setLastTid(&(((HeapTuple) xtuple)->t_self));
/*
* insert index entries for tuple
*/
if (resultRelInfo->ri_NumIndices > 0)
ExecInsertIndexTuples(xslot, &(((HeapTuple) xtuple)->t_self), estate, false);
}
if (planGen == PLANGEN_PLANNER)
{
/* AFTER ROW INSERT Triggers */
ExecARInsertTriggers(estate, resultRelInfo, tuple);
}
}
// -----------------------------------------------------------------------------
// Rebuild filespace persistent table 'gp_persistent_filespace_node'
// -----------------------------------------------------------------------------
void PersistentFilespace_AddCreated(
Oid filespaceOid,
/* The filespace OID to be added. */
bool flushToXLog)
/* When true, the XLOG record for this change will be flushed to disk. */
{
WRITE_PERSISTENT_STATE_ORDERED_LOCK_DECLARE;
PersistentFileSysObjName fsObjName;
ItemPointerData persistentTid;
int64 persistentSerialNum;
FilespaceDirEntry filespaceDirEntry;
/*if (Persistent_BeforePersistenceWork())
{
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Skipping persistent filespace %u because we are before persistence work",
filespaceOid);
return; // The initdb process will load the persistent table once we out of bootstrap mode.
}*/
PersistentFilespace_VerifyInitScan();
PersistentFileSysObjName_SetFilespaceDir(&fsObjName,filespaceOid,is_filespace_shared);
WRITE_PERSISTENT_STATE_ORDERED_LOCK;
filespaceDirEntry = PersistentFilespace_CreateDirUnderLock(filespaceOid);
if (filespaceDirEntry == NULL)
{
/* If out of shared memory, no need to promote to PANIC. */
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
ereport(ERROR,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("Out of shared-memory for persistent filespaces"),
errhint("You may need to increase the gp_max_filespaces value"),
errOmitLocation(true)));
}
// if it is a new generated one, we need to set info from pg_filespace_entry
if(filespaceDirEntry->persistentSerialNum==0 || strlen(filespaceDirEntry->locationBlankPadded1)==0)
{
Relation pg_fs_entry_rel;
HeapScanDesc scandesc;
HeapTuple tuple;
ScanKeyData entry[1];
bool isNull;
Datum locDatum;
char *loc;
/* Lookup the information for the current pg_filespace_entry */
pg_fs_entry_rel = heap_open(FileSpaceEntryRelationId, AccessShareLock);
ScanKeyInit(&entry[0],
Anum_pg_filespace_entry_fsefsoid,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(filespaceOid));
scandesc = heap_beginscan(pg_fs_entry_rel, SnapshotNow, 1, entry);
tuple = heap_getnext(scandesc, ForwardScanDirection);
/* We assume that there can be at most one matching tuple */
if (!HeapTupleIsValid(tuple))
{
elog(ERROR, "filespace %u could not be found in pg_filespace_entry", filespaceOid);
}
locDatum = heap_getattr(tuple,
Anum_pg_filespace_entry_fselocation,
pg_fs_entry_rel->rd_att,
&isNull);
loc = TextDatumGetCString(locDatum);
//convert location with blank padded
memset(filespaceDirEntry->locationBlankPadded1, ' ', FilespaceLocationBlankPaddedWithNullTermLen);
filespaceDirEntry->locationBlankPadded1[FilespaceLocationBlankPaddedWithNullTermLen-1]='\0';
memcpy(filespaceDirEntry->locationBlankPadded1, loc, strlen(loc));
if(isNull)
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("internal error: filespace '%u' has no name defined",
filespaceOid)));
heap_endscan(scandesc);
heap_close(pg_fs_entry_rel, AccessShareLock);
}
filespaceDirEntry->state = PersistentFileSysState_Created;
PersistentFilespace_AddTuple(
filespaceDirEntry,
/* createMirrorDataLossTrackingSessionNum */ 0,
/* reserved */ 0,
/* parentXid */ InvalidTransactionId,
flushToXLog);
persistentTid = filespaceDirEntry->persistentTid;
persistentSerialNum = filespaceDirEntry->persistentSerialNum;
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Persistent filespace directory: Add '%s' in state 'Created', serial number " INT64_FORMAT " at TID '%s' ",
PersistentFileSysObjName_ObjectName(&fsObjName),
persistentSerialNum,
ItemPointerToString(&persistentTid));
}
void
abstime2tm(AbsoluteTime _time, int *tzp, struct pg_tm * tm, char **tzn)
{
pg_time_t time = (pg_time_t) _time;
struct pg_tm *tx;
/*
* If HasCTZSet is true then we have a brute force time zone specified. Go
* ahead and rotate to the local time zone since we will later bypass any
* calls which adjust the tm fields.
*/
if (HasCTZSet && (tzp != NULL))
time -= CTimeZone;
if (!HasCTZSet && tzp != NULL)
tx = pg_localtime(&time, session_timezone);
else
tx = pg_gmtime(&time);
tm->tm_year = tx->tm_year + 1900;
tm->tm_mon = tx->tm_mon + 1;
tm->tm_mday = tx->tm_mday;
tm->tm_hour = tx->tm_hour;
tm->tm_min = tx->tm_min;
tm->tm_sec = tx->tm_sec;
tm->tm_isdst = tx->tm_isdst;
tm->tm_gmtoff = tx->tm_gmtoff;
tm->tm_zone = tx->tm_zone;
if (tzp != NULL)
{
/*
* We have a brute force time zone per SQL99? Then use it without
* change since we have already rotated to the time zone.
*/
if (HasCTZSet)
{
*tzp = CTimeZone;
tm->tm_gmtoff = CTimeZone;
tm->tm_isdst = 0;
tm->tm_zone = NULL;
if (tzn != NULL)
*tzn = NULL;
}
else
{
*tzp = -tm->tm_gmtoff; /* tm_gmtoff is Sun/DEC-ism */
/*
* XXX FreeBSD man pages indicate that this should work - tgl
* 97/04/23
*/
if (tzn != NULL)
{
/*
* Copy no more than MAXTZLEN bytes of timezone to tzn, in
* case it contains an error message, which doesn't fit in the
* buffer
*/
StrNCpy(*tzn, tm->tm_zone, MAXTZLEN + 1);
if (strlen(tm->tm_zone) > MAXTZLEN)
ereport(WARNING,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("invalid time zone name: \"%s\"",
tm->tm_zone),
errOmitLocation(true)));
}
}
}
else
tm->tm_isdst = -1;
}
/*
* Import data into GPDB.
*/
Datum
demoprot_import(PG_FUNCTION_ARGS)
{
extprotocol_t *myData;
char *data;
int datlen;
size_t nread = 0;
/* Must be called via the external table format manager */
if (!CALLED_AS_EXTPROTOCOL(fcinfo))
elog(ERROR, "extprotocol_import: not called by external protocol manager");
/* Get our internal description of the protocol */
myData = (extprotocol_t *) EXTPROTOCOL_GET_USER_CTX(fcinfo);
if(EXTPROTOCOL_IS_LAST_CALL(fcinfo))
{
/* we're done receiving data. close our connection */
if(myData && myData->file)
if(fclose(myData->file))
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not close file \"%s\": %m",
myData->filename)));
PG_RETURN_INT32(0);
}
if (myData == NULL)
{
/* first call. do any desired init */
const char *p_name = "demoprot";
DemoUri *parsed_url;
char *url = EXTPROTOCOL_GET_URL(fcinfo);
myData = palloc(sizeof(extprotocol_t));
myData->url = pstrdup(url);
parsed_url = ParseDemoUri(myData->url);
myData->filename = pstrdup(parsed_url->path);
if(strcasecmp(parsed_url->protocol, p_name) != 0)
elog(ERROR, "internal error: demoprot called with a different protocol (%s)",
parsed_url->protocol);
FreeDemoUri(parsed_url);
/* open the destination file (or connect to remote server in other cases) */
myData->file = fopen(myData->filename, "r");
if (myData->file == NULL)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("demoprot_import: could not open file \"%s\" for reading: %m",
myData->filename),
errOmitLocation(true)));
EXTPROTOCOL_SET_USER_CTX(fcinfo, myData);
}
/* =======================================================================
* DO THE IMPORT
* ======================================================================= */
data = EXTPROTOCOL_GET_DATABUF(fcinfo);
datlen = EXTPROTOCOL_GET_DATALEN(fcinfo);
if(datlen > 0)
{
nread = fread(data, 1, datlen, myData->file);
if (ferror(myData->file))
ereport(ERROR,
(errcode_for_file_access(),
errmsg("demoprot_import: could not write to file \"%s\": %m",
myData->filename)));
}
PG_RETURN_INT32((int)nread);
}
/*
* Indicate we intend to create a filespace file as part of the current transaction.
*
* An XLOG IntentToCreate record is generated that will guard the subsequent file-system
* create in case the transaction aborts.
*
* After 1 or more calls to this routine to mark intention about filespace files that are going
* to be created, call ~_DoPendingCreates to do the actual file-system creates. (See its
* note on XLOG flushing).
*/
void PersistentFilespace_MarkCreatePending(
Oid filespaceOid,
char *filespaceLocation,
ItemPointer persistentTid,
int64 *persistentSerialNum,
bool flushToXLog)
{
WRITE_PERSISTENT_STATE_ORDERED_LOCK_DECLARE;
PersistentFileSysObjName fsObjName;
FilespaceDirEntry filespaceDirEntry;
if (Persistent_BeforePersistenceWork())
{
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Skipping persistent filespace %u because we are before persistence work",
filespaceOid);
return; // The initdb process will load the persistent table once we out of bootstrap mode.
}
PersistentFilespace_VerifyInitScan();
PersistentFileSysObjName_SetFilespaceDir(&fsObjName,filespaceOid,is_filespace_shared);
WRITE_PERSISTENT_STATE_ORDERED_LOCK;
filespaceDirEntry =
PersistentFilespace_CreateDirUnderLock(filespaceOid);
if (filespaceDirEntry == NULL)
{
/* If out of shared memory, no need to promote to PANIC. */
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
ereport(ERROR,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("Out of shared-memory for persistent filespaces"),
errhint("You may need to increase the gp_max_filespaces value"),
errOmitLocation(true)));
}
PersistentFilespace_BlankPadCopyLocation(
filespaceDirEntry->locationBlankPadded1,
filespaceLocation);
filespaceDirEntry->state = PersistentFileSysState_CreatePending;
PersistentFilespace_AddTuple(
filespaceDirEntry,
/* createMirrorDataLossTrackingSessionNum */ 0,
/* reserved */ 0,
/* parentXid */ GetTopTransactionId(),
flushToXLog);
*persistentTid = filespaceDirEntry->persistentTid;
*persistentSerialNum = filespaceDirEntry->persistentSerialNum;
/*
* This XLOG must be generated under the persistent write-lock.
*/
#ifdef MASTER_MIRROR_SYNC
mmxlog_log_create_filespace(filespaceOid);
#endif
#ifdef FAULT_INJECTOR
FaultInjector_InjectFaultIfSet(
FaultBeforePendingDeleteFilespaceEntry,
DDLNotSpecified,
"", // databaseName
""); // tableName
#endif
/*
* MPP-18228
* To make adding 'Create Pending' entry to persistent table and adding
* to the PendingDelete list atomic
*/
PendingDelete_AddCreatePendingEntryWrapper(
&fsObjName,
persistentTid,
*persistentSerialNum);
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Persistent filespace directory: Add '%s' in state 'Created', serial number " INT64_FORMAT " at TID %s",
PersistentFileSysObjName_ObjectName(&fsObjName),
*persistentSerialNum,
ItemPointerToString(persistentTid));
}
/*
* Implements the 'EXECUTE' utility statement.
*
* Note: this is one of very few places in the code that needs to deal with
* two query strings at once. The passed-in queryString is that of the
* EXECUTE, which we might need for error reporting while processing the
* parameter expressions. The query_string that we copy from the plan
* source is that of the original PREPARE.
*/
void
ExecuteQuery(ExecuteStmt *stmt, const char *queryString,
ParamListInfo params,
DestReceiver *dest, char *completionTag)
{
PreparedStatement *entry;
List *stmt_list;
MemoryContext qcontext;
ParamListInfo paramLI = NULL;
EState *estate = NULL;
Portal portal;
/* Look it up in the hash table */
entry = FetchPreparedStatement(stmt->name, true);
qcontext = entry->context;
/* Evaluate parameters, if any */
if (entry->argtype_list != NIL)
{
/*
* Need an EState to evaluate parameters; must not delete it till end
* of query, in case parameters are pass-by-reference.
*/
estate = CreateExecutorState();
estate->es_param_list_info = params;
paramLI = EvaluateParams(estate, stmt->params, entry->argtype_list);
}
/* Create a new portal to run the query in */
portal = CreateNewPortal();
/* Don't display the portal in pg_cursors, it is for internal use only */
portal->visible = false;
/* Plan the query. If this is a CTAS, copy the "into" information into
* the query so that we construct the plan correctly. Else the table
* might not be created on the segments. (MPP-8135) */
{
List *query_list = copyObject(entry->query_list); /* planner scribbles on query tree :( */
if ( stmt->into )
{
Query *query = (Query*)linitial(query_list);
Assert(IsA(query, Query) && query->intoClause == NULL);
query->intoClause = copyObject(stmt->into);
}
stmt_list = pg_plan_queries(query_list, paramLI, false, QRL_ONCE);
}
/*
* For CREATE TABLE / AS EXECUTE, make a copy of the stored query so that
* we can modify its destination (yech, but this has always been ugly).
* For regular EXECUTE we can just use the stored query where it sits,
* since the executor is read-only.
*/
if (stmt->into)
{
MemoryContext oldContext;
PlannedStmt *pstmt;
if (list_length(stmt_list) != 1)
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("prepared statement is not a SELECT")));
oldContext = MemoryContextSwitchTo(PortalGetHeapMemory(portal));
stmt_list = copyObject(stmt_list);
qcontext = PortalGetHeapMemory(portal);
pstmt = (PlannedStmt *) linitial(stmt_list);
pstmt->qdContext = qcontext;
if (pstmt->commandType != CMD_SELECT)
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("prepared statement is not a SELECT"),
errOmitLocation(true)));
pstmt->intoClause = copyObject(stmt->into);
/* XXX Is it legitimate to assign a constant default policy without
* even checking the relation?
*/
pstmt->intoPolicy = palloc0(sizeof(GpPolicy)- sizeof(pstmt->intoPolicy->attrs)
+ 255 * sizeof(pstmt->intoPolicy->attrs[0]));
pstmt->intoPolicy->nattrs = 0;
pstmt->intoPolicy->ptype = POLICYTYPE_PARTITIONED;
pstmt->intoPolicy->bucketnum = GetRelOpt_bucket_num_fromRangeVar(stmt->into->rel, GetRandomDistPartitionNum());
MemoryContextSwitchTo(oldContext);
}
/* Copy the plan's saved query string into the portal's memory */
Assert(entry->query_string != NULL);
char *query_string = MemoryContextStrdup(PortalGetHeapMemory(portal),
entry->query_string);
PortalDefineQuery(portal,
NULL,
query_string,
entry->sourceTag,
entry->commandTag,
stmt_list,
qcontext);
create_filesystem_credentials(portal);
/*
* Run the portal to completion.
*/
PortalStart(portal, paramLI, ActiveSnapshot,
savedSeqServerHost, savedSeqServerPort);
(void) PortalRun(portal,
FETCH_ALL,
true, /* Effectively always top level. */
dest,
dest,
completionTag);
PortalDrop(portal, false);
if (estate)
FreeExecutorState(estate);
/* No need to pfree other memory, MemoryContext will be reset */
}
/*
* make_scalar_array_op()
* Build expression tree for "scalar op ANY/ALL (array)" construct.
*/
Expr *
make_scalar_array_op(ParseState *pstate, List *opname,
bool useOr,
Node *ltree, Node *rtree,
int location)
{
Oid ltypeId,
rtypeId,
atypeId,
res_atypeId;
Operator tup;
Form_pg_operator opform;
Oid actual_arg_types[2];
Oid declared_arg_types[2];
List *args;
Oid rettype;
ScalarArrayOpExpr *result;
ltypeId = exprType(ltree);
atypeId = exprType(rtree);
/*
* The right-hand input of the operator will be the element type of the
* array. However, if we currently have just an untyped literal on the
* right, stay with that and hope we can resolve the operator.
*/
if (atypeId == UNKNOWNOID)
rtypeId = UNKNOWNOID;
else
{
rtypeId = get_element_type(atypeId);
if (!OidIsValid(rtypeId))
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("op ANY/ALL (array) requires array on right side"),
errOmitLocation(true),
parser_errposition(pstate, location)));
}
/* Now resolve the operator */
tup = oper(pstate, opname, ltypeId, rtypeId, false, location);
opform = (Form_pg_operator) GETSTRUCT(tup);
args = list_make2(ltree, rtree);
actual_arg_types[0] = ltypeId;
actual_arg_types[1] = rtypeId;
declared_arg_types[0] = opform->oprleft;
declared_arg_types[1] = opform->oprright;
/*
* enforce consistency with ANYARRAY and ANYELEMENT argument and return
* types, possibly adjusting return type or declared_arg_types (which will
* be used as the cast destination by make_fn_arguments)
*/
rettype = enforce_generic_type_consistency(actual_arg_types,
declared_arg_types,
2,
opform->oprresult);
/*
* Check that operator result is boolean
*/
if (rettype != BOOLOID)
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("op ANY/ALL (array) requires operator to yield boolean"),
errOmitLocation(true),
parser_errposition(pstate, location)));
if (get_func_retset(opform->oprcode))
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("op ANY/ALL (array) requires operator not to return a set"),
errOmitLocation(true),
parser_errposition(pstate, location)));
/*
* Now switch back to the array type on the right, arranging for any
* needed cast to be applied.
*/
res_atypeId = get_array_type(declared_arg_types[1]);
if (!OidIsValid(res_atypeId))
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("could not find array type for data type %s",
format_type_be(declared_arg_types[1])),
errOmitLocation(true),
parser_errposition(pstate, location)));
actual_arg_types[1] = atypeId;
declared_arg_types[1] = res_atypeId;
/* perform the necessary typecasting of arguments */
make_fn_arguments(pstate, args, actual_arg_types, declared_arg_types);
/* and build the expression node */
result = makeNode(ScalarArrayOpExpr);
result->opno = oprid(tup);
result->opfuncid = InvalidOid;
result->useOr = useOr;
result->args = args;
ReleaseOperator(tup);
return (Expr *) result;
}
/*
* Indicate we intend to create a relation file as part of the current transaction.
*
* An XLOG IntentToCreate record is generated that will guard the subsequent file-system
* create in case the transaction aborts.
*
* After 1 or more calls to this routine to mark intention about relation files that are going
* to be created, call ~_DoPendingCreates to do the actual file-system creates. (See its
* note on XLOG flushing).
*/
void PersistentDatabase_MarkCreatePending(
DbDirNode *dbDirNode,
ItemPointer persistentTid,
int64 *persistentSerialNum,
bool flushToXLog)
{
WRITE_PERSISTENT_STATE_ORDERED_LOCK_DECLARE;
DatabaseDirEntry databaseDirEntry;
SharedOidSearchAddResult addResult;
PersistentFileSysObjName fsObjName;
if (Persistent_BeforePersistenceWork())
{
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Skipping persistent database '%s' because we are before persistence work",
GetDatabasePath(
dbDirNode->database,
dbDirNode->tablespace));
/*
* The initdb process will load the persistent table once we
* out of bootstrap mode.
*/
return;
}
PersistentDatabase_VerifyInitScan();
PersistentFileSysObjName_SetDatabaseDir(
&fsObjName,
dbDirNode->tablespace,
dbDirNode->database,
is_tablespace_shared);
WRITE_PERSISTENT_STATE_ORDERED_LOCK;
databaseDirEntry =
(DatabaseDirEntry)
SharedOidSearch_Find(
&persistentDatabaseSharedData->databaseDirSearchTable,
dbDirNode->database,
dbDirNode->tablespace);
if (databaseDirEntry != NULL)
elog(ERROR, "Persistent database entry '%s' already exists in state '%s'",
GetDatabasePath(
dbDirNode->database,
dbDirNode->tablespace),
PersistentFileSysObjState_Name(databaseDirEntry->state));
addResult =
SharedOidSearch_Add(
&persistentDatabaseSharedData->databaseDirSearchTable,
dbDirNode->database,
dbDirNode->tablespace,
(SharedOidSearchObjHeader**)&databaseDirEntry);
if (addResult == SharedOidSearchAddResult_NoMemory)
{
/* If out of shared memory, no need to promote to PANIC. */
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
ereport(ERROR,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("Out of shared-memory for persistent databases"),
errhint("You may need to increase the gp_max_databases and "
"gp_max_tablespaces value"),
errOmitLocation(true)));
}
else if (addResult == SharedOidSearchAddResult_Exists)
elog(PANIC, "Persistent database entry '%s' already exists in state '%s'",
GetDatabasePath(
dbDirNode->database,
dbDirNode->tablespace),
PersistentFileSysObjState_Name(databaseDirEntry->state));
else
Assert(addResult == SharedOidSearchAddResult_Ok);
databaseDirEntry->state = PersistentFileSysState_CreatePending;
databaseDirEntry->iteratorRefCount = 0;
PersistentDatabase_AddTuple(
databaseDirEntry,
/* reserved */ 0,
/* parentXid */ GetTopTransactionId(),
flushToXLog);
*persistentTid = databaseDirEntry->persistentTid;
*persistentSerialNum = databaseDirEntry->persistentSerialNum;
/*
* This XLOG must be generated under the persistent write-lock.
*/
#ifdef MASTER_MIRROR_SYNC
mmxlog_log_create_database(dbDirNode->tablespace, dbDirNode->database);
#endif
#ifdef FAULT_INJECTOR
FaultInjector_InjectFaultIfSet(
FaultBeforePendingDeleteDatabaseEntry,
DDLNotSpecified,
"", // databaseName
""); // tableName
#endif
/*
* MPP-18228
* To make adding 'Create Pending' entry to persistent table and adding
* to the PendingDelete list atomic
*/
PendingDelete_AddCreatePendingEntryWrapper(
&fsObjName,
persistentTid,
*persistentSerialNum);
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
}
/* ----------------------------------------------------------------
* ExecUpdate
*
* note: we can't run UPDATE queries with transactions
* off because UPDATEs are actually INSERTs and our
* scan will mistakenly loop forever, updating the tuple
* it just inserted.. This should be fixed but until it
* is, we don't want to get stuck in an infinite loop
* which corrupts your database..
* ----------------------------------------------------------------
*/
void
ExecUpdate(TupleTableSlot *slot,
ItemPointer tupleid,
TupleTableSlot *planSlot,
DestReceiver *dest,
EState *estate)
{
HeapTuple tuple;
ResultRelInfo *resultRelInfo;
Relation resultRelationDesc;
HTSU_Result result;
ItemPointerData update_ctid;
TransactionId update_xmax;
/*
* abort the operation if not running transactions
*/
if (IsBootstrapProcessingMode())
elog(ERROR, "cannot UPDATE during bootstrap");
/*
* get the heap tuple out of the tuple table slot, making sure we have a
* writable copy
*/
tuple = ExecFetchSlotHeapTuple(slot);
/*
* get information on the (current) result relation
*/
resultRelInfo = estate->es_result_relation_info;
resultRelationDesc = resultRelInfo->ri_RelationDesc;
/* see if this update would move the tuple to a different partition */
if (estate->es_result_partitions)
{
AttrNumber max_attr;
Datum *values;
bool *nulls;
Oid targetid;
Assert(estate->es_partition_state != NULL &&
estate->es_partition_state->accessMethods != NULL);
if (!estate->es_partition_state->accessMethods->part_cxt)
estate->es_partition_state->accessMethods->part_cxt =
GetPerTupleExprContext(estate)->ecxt_per_tuple_memory;
Assert(PointerIsValid(estate->es_result_partitions));
max_attr = estate->es_partition_state->max_partition_attr;
slot_getsomeattrs(slot, max_attr);
values = slot_get_values(slot);
nulls = slot_get_isnull(slot);
targetid = selectPartition(estate->es_result_partitions, values,
nulls, slot->tts_tupleDescriptor,
estate->es_partition_state->accessMethods);
if (!OidIsValid(targetid))
ereport(ERROR,
(errcode(ERRCODE_NO_PARTITION_FOR_PARTITIONING_KEY),
errmsg("no partition for partitioning key")));
if (RelationGetRelid(resultRelationDesc) != targetid)
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("moving tuple from partition \"%s\" to "
"partition \"%s\" not supported",
get_rel_name(RelationGetRelid(resultRelationDesc)),
get_rel_name(targetid)),
errOmitLocation(true)));
}
}
/* BEFORE ROW UPDATE Triggers */
if (resultRelInfo->ri_TrigDesc &&
resultRelInfo->ri_TrigDesc->n_before_row[TRIGGER_EVENT_UPDATE] > 0)
{
HeapTuple newtuple;
newtuple = ExecBRUpdateTriggers(estate, resultRelInfo,
tupleid, tuple,
estate->es_snapshot->curcid);
if (newtuple == NULL) /* "do nothing" */
return;
if (newtuple != tuple) /* modified by Trigger(s) */
{
/*
* Put the modified tuple into a slot for convenience of routines
* below. We assume the tuple was allocated in per-tuple memory
* context, and therefore will go away by itself. The tuple table
* slot should not try to clear it.
*/
TupleTableSlot *newslot = estate->es_trig_tuple_slot;
if (newslot->tts_tupleDescriptor != slot->tts_tupleDescriptor)
ExecSetSlotDescriptor(newslot, slot->tts_tupleDescriptor);
ExecStoreGenericTuple(newtuple, newslot, false);
newslot->tts_tableOid = slot->tts_tableOid; /* for constraints */
slot = newslot;
tuple = newtuple;
}
}
/*
* Check the constraints of the tuple
*
* If we generate a new candidate tuple after EvalPlanQual testing, we
* must loop back here and recheck constraints. (We don't need to redo
* triggers, however. If there are any BEFORE triggers then trigger.c
* will have done heap_lock_tuple to lock the correct tuple, so there's no
* need to do them again.)
*/
lreplace:;
if (resultRelationDesc->rd_att->constr)
ExecConstraints(resultRelInfo, slot, estate);
if (!GpPersistent_IsPersistentRelation(resultRelationDesc->rd_id))
{
/*
* Normal UPDATE path.
*/
/*
* replace the heap tuple
*
* Note: if es_crosscheck_snapshot isn't InvalidSnapshot, we check that
* the row to be updated is visible to that snapshot, and throw a can't-
* serialize error if not. This is a special-case behavior needed for
* referential integrity updates in serializable transactions.
*/
result = heap_update(resultRelationDesc, tupleid, tuple,
&update_ctid, &update_xmax,
estate->es_snapshot->curcid,
estate->es_crosscheck_snapshot,
true /* wait for commit */ );
switch (result)
{
case HeapTupleSelfUpdated:
/* already deleted by self; nothing to do */
return;
case HeapTupleMayBeUpdated:
break;
case HeapTupleUpdated:
if (IsXactIsoLevelSerializable)
ereport(ERROR,
(errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
errmsg("could not serialize access due to concurrent update")));
else if (!ItemPointerEquals(tupleid, &update_ctid))
{
TupleTableSlot *epqslot;
epqslot = EvalPlanQual(estate,
resultRelInfo->ri_RangeTableIndex,
&update_ctid,
update_xmax,
estate->es_snapshot->curcid);
if (!TupIsNull(epqslot))
{
*tupleid = update_ctid;
slot = ExecFilterJunk(estate->es_junkFilter, epqslot);
tuple = ExecFetchSlotHeapTuple(slot);
goto lreplace;
}
}
/* tuple already deleted; nothing to do */
return;
default:
elog(ERROR, "unrecognized heap_update status: %u", result);
return;
}
}
else
{
HeapTuple persistentTuple;
/*
* Persistent metadata path.
*/
persistentTuple = heap_copytuple(tuple);
persistentTuple->t_self = *tupleid;
frozen_heap_inplace_update(resultRelationDesc, persistentTuple);
heap_freetuple(persistentTuple);
}
IncrReplaced();
(estate->es_processed)++;
/*
* Note: instead of having to update the old index tuples associated with
* the heap tuple, all we do is form and insert new index tuples. This is
* because UPDATEs are actually DELETEs and INSERTs, and index tuple
* deletion is done later by VACUUM (see notes in ExecDelete). All we do
* here is insert new index tuples. -cim 9/27/89
*/
/*
* insert index entries for tuple
*
* Note: heap_update returns the tid (location) of the new tuple in the
* t_self field.
*/
if (resultRelInfo->ri_NumIndices > 0)
ExecInsertIndexTuples(slot, &(tuple->t_self), estate, false);
/* AFTER ROW UPDATE Triggers */
ExecARUpdateTriggers(estate, resultRelInfo, tupleid, tuple);
}
void PersistentDatabase_AddCreated(
DbDirNode *dbDirNode,
/* The tablespace and database OIDs for the create. */
ItemPointer persistentTid,
/* TID of the gp_persistent_rel_files tuple for the rel file */
bool flushToXLog)
/* When true, the XLOG record for this change will be flushed to disk. */
{
WRITE_PERSISTENT_STATE_ORDERED_LOCK_DECLARE;
PersistentFileSysObjName fsObjName;
DatabaseDirEntry databaseDirEntry;
SharedOidSearchAddResult addResult;
int64 persistentSerialNum;
if (!Persistent_BeforePersistenceWork())
elog(ERROR, "We can only add to persistent meta-data when special states");
// Verify PersistentFileSysObj_BuildInitScan has been called.
PersistentDatabase_VerifyInitScan();
PersistentFileSysObjName_SetDatabaseDir(&fsObjName,dbDirNode->tablespace,dbDirNode->database,is_tablespace_shared);
WRITE_PERSISTENT_STATE_ORDERED_LOCK;
/*
* GpIdentity.segindex
* In the initdb, GpIdentity.segindex is set to -10000. It will update this
* value to the correct GpIdentity.segindex.
*/
databaseDirEntry =
(DatabaseDirEntry)
SharedOidSearch_Find(
&persistentDatabaseSharedData->databaseDirSearchTable,
dbDirNode->database,
dbDirNode->tablespace);
if (databaseDirEntry != NULL)
elog(ERROR, "Persistent database entry '%s' already exists in state '%s'",
GetDatabasePath(
dbDirNode->database,
dbDirNode->tablespace),
PersistentFileSysObjState_Name(databaseDirEntry->state));
addResult =
SharedOidSearch_Add(
&persistentDatabaseSharedData->databaseDirSearchTable,
dbDirNode->database,
dbDirNode->tablespace,
(SharedOidSearchObjHeader**)&databaseDirEntry);
if (addResult == SharedOidSearchAddResult_NoMemory)
{
/* If out of shared memory, no need to promote to PANIC. */
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
ereport(ERROR,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("Out of shared-memory for persistent databases"),
errhint("You may need to increase the gp_max_databases and "
"gp_max_tablespaces value"),
errOmitLocation(true)));
}
else if (addResult == SharedOidSearchAddResult_Exists)
elog(PANIC, "Persistent database entry '%s' already exists in state '%s'",
GetDatabasePath(
dbDirNode->database,
dbDirNode->tablespace),
PersistentFileSysObjState_Name(databaseDirEntry->state));
else
Assert(addResult == SharedOidSearchAddResult_Ok);
databaseDirEntry->state = PersistentFileSysState_Created;
databaseDirEntry->iteratorRefCount = 0;
PersistentDatabase_AddTuple(
databaseDirEntry,
/* reserved */ 0,
InvalidTransactionId,
flushToXLog);
*persistentTid = databaseDirEntry->persistentTid;
persistentSerialNum = databaseDirEntry->persistentSerialNum;
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Persistent database directory: Add '%s' in state 'Created', serial number " INT64_FORMAT " at TID %s",
PersistentFileSysObjName_ObjectName(&fsObjName),
persistentSerialNum,
ItemPointerToString(persistentTid));
}
/* ----------------------------------------------------------------
* cql1()
* The underlying function for the cql() macro.
* find the bind parameters in a caql string and build a key list
* ----------------------------------------------------------------
*/
cq_list *
cql1(const char* caqlStr, const char* filename, int lineno, ...)
{
int maxkeys = 0;
int badbind = 0;
cq_list *pcql = (cq_list *) palloc0(sizeof(cq_list));
const char* pc = caqlStr;
if ((NULL == caqlStr) ||
(0 == strlen(caqlStr)))
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg(
"Invalid or undefined CaQL string"
),
errOmitLocation(true)));
pcql->bGood = true;
pcql->caqlStr = caqlStr;
pcql->filename = filename;
pcql->lineno = lineno;
while (*pc && (maxkeys < 5))
{
if (*pc != ':')
{
pc++;
continue;
}
/* look for numeric bind parameter of the form ":1" to ":5" */
pc++;
if (!*pc)
break;
switch (*pc)
{
case '1':
if (maxkeys != 0)
{
badbind = 1;
goto L_wrong_args;
}
maxkeys++;
break;
case '2':
if (maxkeys != 1)
{
badbind = 2;
goto L_wrong_args;
}
maxkeys++;
break;
case '3':
if (maxkeys != 2)
{
badbind = 3;
goto L_wrong_args;
}
maxkeys++;
break;
case '4':
if (maxkeys != 3)
{
badbind = 4;
goto L_wrong_args;
}
maxkeys++;
break;
case '5':
if (maxkeys != 4)
{
badbind = 5;
goto L_wrong_args;
}
maxkeys++;
break;
case '6':
case '7':
case '8':
case '9':
case '0':
{
badbind = 6;
goto L_wrong_args;
}
default:
break;
} /* end switch */
} /* end while */
pcql->maxkeys = maxkeys;
if (maxkeys)
{
va_list ap;
va_start(ap, lineno);
for (int ii = 0; ii < maxkeys; ii++)
{
pcql->cqlKeys[ii] = va_arg(ap, Datum);
}
va_end(ap);
}
return (pcql);
L_wrong_args:
if (badbind != 5)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg(
"bind parameter out of range (1-5)"
),
errOmitLocation(true)));
else
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg(
"missing bind parameter before :%d, or out of sequence",
badbind
),
errOmitLocation(true)));
return NULL;
} /* end cql1 */
/*
* Implements the 'PREPARE' utility statement.
*/
void
PrepareQuery(PrepareStmt *stmt, const char *queryString)
{
const char *commandTag = NULL;
Query *query = NULL;
List *query_list = NIL;
List *plan_list = NIL;
List *query_list_copy = NIL;
NodeTag srctag; /* GPDB */
/*
* Disallow empty-string statement name (conflicts with protocol-level
* unnamed statement).
*/
if (!stmt->name || stmt->name[0] == '\0')
ereport(ERROR,
(errcode(ERRCODE_INVALID_PSTATEMENT_DEFINITION),
errmsg("invalid statement name: must not be empty")));
switch (stmt->query->commandType)
{
case CMD_SELECT:
commandTag = "SELECT";
srctag = T_SelectStmt;
break;
case CMD_INSERT:
commandTag = "INSERT";
srctag = T_InsertStmt;
break;
case CMD_UPDATE:
commandTag = "UPDATE";
srctag = T_UpdateStmt;
break;
case CMD_DELETE:
commandTag = "DELETE";
srctag = T_DeleteStmt;
break;
default:
ereport(ERROR,
(errcode(ERRCODE_INVALID_PSTATEMENT_DEFINITION),
errmsg("utility statements cannot be prepared"),
errOmitLocation(true)));
commandTag = NULL; /* keep compiler quiet */
srctag = T_Query;
break;
}
/*
* Parse analysis is already done, but we must still rewrite and plan the
* query.
*/
/*
* Because the planner is not cool about not scribbling on its input, we
* make a preliminary copy of the source querytree. This prevents
* problems in the case that the PREPARE is in a portal or plpgsql
* function and is executed repeatedly. (See also the same hack in
* DECLARE CURSOR and EXPLAIN.) XXX the planner really shouldn't modify
* its input ... FIXME someday.
*/
query = copyObject(stmt->query);
/* Rewrite the query. The result could be 0, 1, or many queries. */
AcquireRewriteLocks(query);
query_list = QueryRewrite(query);
query_list_copy = copyObject(query_list); /* planner scribbles on query tree */
/* Generate plans for queries. Snapshot is already set. */
plan_list = pg_plan_queries(query_list, NULL, false, QRL_NONE);
/*
* Save the results. We don't have the query string for this PREPARE, but
* we do have the string we got from the client, so use that.
*/
StorePreparedStatement(stmt->name,
queryString, /* WAS global debug_query_string, */
srctag,
commandTag,
query_list_copy,
stmt->argtype_oids,
true);
}
/*
* Wait for all QEs to finish, then report any errors from the given
* CdbDispatchResults objects and free them. If not all QEs in the
* associated gang(s) executed the command successfully, throws an
* error and does not return. No-op if both CdbDispatchResults ptrs are NULL.
* This is a convenience function; callers with unusual requirements may
* instead call CdbCheckDispatchResult(), etc., directly.
*/
void
cdbdisp_finishCommand(struct CdbDispatcherState *ds,
void (*handle_results_callback) (CdbDispatchResults *
primaryResults,
void *ctx), void *ctx)
{
StringInfoData buf;
int errorcode = 0;
/*
* If cdbdisp_dispatchToGang() wasn't called, don't wait.
*/
if (!ds || !ds->primaryResults)
return;
/*
* If we are called in the dying sequence, don't touch QE connections.
* Anything below could cause ERROR in which case we would miss a chance
* to clean up shared memory as this is from AbortTransaction.
* QE may stay a bit longer, but since we can consider QD as libpq
* client to QE, they will notice that we as a client do not
* appear anymore and will finish soon. Also ERROR report doesn't
* go to the client anyway since we are in proc_exit.
*/
if (proc_exit_inprogress)
return;
/*
* Wait for all QEs to finish. Don't cancel them.
*/
CdbCheckDispatchResult(ds, DISPATCH_WAIT_NONE);
/*
* If no errors, free the CdbDispatchResults objects and return.
*/
if (ds->primaryResults)
errorcode = ds->primaryResults->errcode;
if (!errorcode)
{
/*
* Call the callback function to handle the results
*/
if (handle_results_callback != NULL)
handle_results_callback(ds->primaryResults, ctx);
cdbdisp_destroyDispatcherState(ds);
return;
}
/*
* Format error messages from the primary gang.
*/
initStringInfo(&buf);
cdbdisp_dumpDispatchResults(ds->primaryResults, &buf);
cdbdisp_destroyDispatcherState(ds);
/*
* Too bad, our gang got an error.
*/
PG_TRY();
{
ereport(ERROR, (errcode(errorcode),
errOmitLocation(true), errmsg("%s", buf.data)));
}
PG_CATCH();
{
pfree(buf.data);
PG_RE_THROW();
}
PG_END_TRY();
}
