/*
* Individual logging of AccessExclusiveLocks for use during LockAcquire()
*/
void
LogAccessExclusiveLock(Oid dbOid, Oid relOid)
{
xl_standby_lock xlrec;
/*
* Ensure that a TransactionId has been assigned to this transaction. We
* don't actually need the xid yet but if we don't do this then
* RecordTransactionCommit() and RecordTransactionAbort() will optimise
* away the transaction completion record which recovery relies upon to
* release locks. It's a hack, but for a corner case not worth adding code
* for into the main commit path.
*/
xlrec.xid = GetTopTransactionId();
/*
* Decode the locktag back to the original values, to avoid sending lots
* of empty bytes with every message. See lock.h to check how a locktag
* is defined for LOCKTAG_RELATION
*/
xlrec.dbOid = dbOid;
xlrec.relOid = relOid;
LogAccessExclusiveLocks(1, &xlrec);
}
/*
* txid_current() returns int8
*
* Return the current toplevel transaction ID as TXID
*/
Datum
txid_current(PG_FUNCTION_ARGS)
{
txid val;
TxidEpoch state;
load_xid_epoch(&state);
val = convert_xid(GetTopTransactionId(), &state);
PG_RETURN_INT64(val);
}
/*
* AtEOXact_Snapshot
* Snapshot manager's cleanup function for end of transaction
*/
void
AtEOXact_Snapshot(bool isCommit)
{
/*
* In transaction-snapshot mode we must release our privately-managed
* reference to the transaction snapshot. We must decrement
* RegisteredSnapshots to keep the check below happy. But we don't bother
* to do FreeSnapshot, for two reasons: the memory will go away with
* TopTransactionContext anyway, and if someone has left the snapshot
* stacked as active, we don't want the code below to be chasing through a
* dangling pointer.
*/
if (FirstXactSnapshot != NULL)
{
Assert(FirstXactSnapshot->regd_count > 0);
Assert(!pairingheap_is_empty(&RegisteredSnapshots));
pairingheap_remove(&RegisteredSnapshots, &FirstXactSnapshot->ph_node);
}
FirstXactSnapshot = NULL;
/*
* If we exported any snapshots, clean them up.
*/
if (exportedSnapshots != NIL)
{
TransactionId myxid = GetTopTransactionId();
int i;
char buf[MAXPGPATH];
ListCell *lc;
/*
* Get rid of the files. Unlink failure is only a WARNING because (1)
* it's too late to abort the transaction, and (2) leaving a leaked
* file around has little real consequence anyway.
*/
for (i = 1; i <= list_length(exportedSnapshots); i++)
{
XactExportFilePath(buf, myxid, i, "");
if (unlink(buf))
elog(WARNING, "could not unlink file \"%s\": %m", buf);
}
/*
* As with the FirstXactSnapshot, we needn't spend any effort on
* cleaning up the per-snapshot data structures, but we do need to
* unlink them from RegisteredSnapshots to prevent a warning below.
*/
foreach(lc, exportedSnapshots)
{
Snapshot snap = (Snapshot) lfirst(lc);
pairingheap_remove(&RegisteredSnapshots, &snap->ph_node);
}
/*
* xid_age - compute age of an XID (relative to current xact)
*/
Datum
xid_age(PG_FUNCTION_ARGS)
{
TransactionId xid = PG_GETARG_TRANSACTIONID(0);
TransactionId now = GetTopTransactionId();
/* Permanent XIDs are always infinitely old */
if (!TransactionIdIsNormal(xid))
PG_RETURN_INT32(INT_MAX);
PG_RETURN_INT32((int32) (now - xid));
}
/*
* LockGXact
* Locate the prepared transaction and mark it busy for COMMIT or PREPARE.
*/
static GlobalTransaction
LockGXact(const char *gid, Oid user)
{
int i;
LWLockAcquire(TwoPhaseStateLock, LW_EXCLUSIVE);
for (i = 0; i < TwoPhaseState->numPrepXacts; i++)
{
GlobalTransaction gxact = TwoPhaseState->prepXacts[i];
/* Ignore not-yet-valid GIDs */
if (!gxact->valid)
continue;
if (strcmp(gxact->gid, gid) != 0)
continue;
/* Found it, but has someone else got it locked? */
if (TransactionIdIsValid(gxact->locking_xid))
{
if (TransactionIdIsActive(gxact->locking_xid))
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("prepared transaction with identifier \"%s\" is busy",
gid)));
gxact->locking_xid = InvalidTransactionId;
}
if (user != gxact->owner && !superuser_arg(user))
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("permission denied to finish prepared transaction"),
errhint("Must be superuser or the user that prepared the transaction.")));
/* OK for me to lock it */
gxact->locking_xid = GetTopTransactionId();
LWLockRelease(TwoPhaseStateLock);
return gxact;
}
LWLockRelease(TwoPhaseStateLock);
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("prepared transaction with identifier \"%s\" does not exist",
gid)));
/* NOTREACHED */
return NULL;
}
/*
* Individual logging of AccessExclusiveLocks for use during LockAcquire()
*/
void
LogAccessExclusiveLock(Oid dbOid, Oid relOid)
{
xl_standby_lock xlrec;
xlrec.xid = GetTopTransactionId();
/*
* Decode the locktag back to the original values, to avoid sending lots
* of empty bytes with every message. See lock.h to check how a locktag
* is defined for LOCKTAG_RELATION
*/
xlrec.dbOid = dbOid;
xlrec.relOid = relOid;
LogAccessExclusiveLocks(1, &xlrec);
}
/*
* Prepare to log an AccessExclusiveLock, for use during LockAcquire()
*/
void
LogAccessExclusiveLockPrepare(void)
{
/*
* Ensure that a TransactionId has been assigned to this transaction, for
* two reasons, both related to lock release on the standby. First, we
* must assign an xid so that RecordTransactionCommit() and
* RecordTransactionAbort() do not optimise away the transaction
* completion record which recovery relies upon to release locks. It's a
* hack, but for a corner case not worth adding code for into the main
* commit path. Second, must must assign an xid before the lock is
* recorded in shared memory, otherwise a concurrently executing
* GetRunningTransactionLocks() might see a lock associated with an
* InvalidTransactionId which we later assert cannot happen.
*/
(void) GetTopTransactionId();
}
/*
* txid_current() returns int8
*
* Return the current toplevel transaction ID as TXID
* If the current transaction does not have one, one is assigned.
*/
Datum
txid_current(PG_FUNCTION_ARGS)
{
txid val;
TxidEpoch state;
/*
* Must prevent during recovery because if an xid is not assigned we try
* to assign one, which would fail. Programs already rely on this function
* to always return a valid current xid, so we should not change this to
* return NULL or similar invalid xid.
*/
PreventCommandDuringRecovery("txid_current()");
load_xid_epoch(&state);
val = convert_xid(GetTopTransactionId(), &state);
PG_RETURN_INT64(val);
}
/*
* XactLockTableWait
*
* Wait for the specified transaction to commit or abort.
*
* Note that this does the right thing for subtransactions: if we wait on a
* subtransaction, we will exit as soon as it aborts or its top parent commits.
* It takes some extra work to ensure this, because to save on shared memory
* the XID lock of a subtransaction is released when it ends, whether
* successfully or unsuccessfully. So we have to check if it's "still running"
* and if so wait for its parent.
*/
void
XactLockTableWait(TransactionId xid)
{
LOCKTAG tag;
for (;;)
{
Assert(TransactionIdIsValid(xid));
Assert(!TransactionIdEquals(xid, GetTopTransactionId()));
SET_LOCKTAG_TRANSACTION(tag, xid);
(void) LockAcquire(&tag, ShareLock, false, false);
LockRelease(&tag, ShareLock, false);
if (!TransactionIdIsInProgress(xid))
break;
xid = SubTransGetParent(xid);
}
}
struct GridPointDataListIterator * getExtractGridDataReturnValues(FunctionCallInfo fcinfo)
{
struct PlaceSpecification ps;
Datum placeSpec = PG_GETARG_DATUM(0);
extractPlaceSpecification( & ps, & placeSpec );
GEOSGeom location = NULL;
if ( ! PG_ARGISNULL(1) )
{
bytea * locationRaw = PG_GETARG_BYTEA_P(1);
location = GEOSGeomFromWKB_buf((unsigned char *) VARDATA(locationRaw), VARSIZE(locationRaw) - VARHDRSZ);
}
enum InterpolationType interpolation = (enum InterpolationType) PG_GETARG_INT32(2);
FileId dataId = PG_GETARG_INT64(3);
TransactionId xid = GetTopTransactionId();
CommandId cid = GetCurrentCommandId(true); // Incremented for each function call in the same transaction
// function takes ownership of location parameter
struct GridPointDataListIterator * ret = readPoints(& ps, location, interpolation, dataId, xid, cid);
return ret;
}
/*
* fetch insert plan from cache.
*/
static void *load_insert_plan(Datum qname, struct QueueState *state)
{
struct InsertCacheEntry *entry;
Oid queue_id = state->queue_id;
bool did_exist = false;
entry = hash_search(insert_cache, &queue_id, HASH_ENTER, &did_exist);
if (did_exist) {
if (entry->plan && state->cur_table == entry->cur_table)
goto valid_table;
if (entry->plan)
SPI_freeplan(entry->plan);
}
entry->cur_table = state->cur_table;
entry->last_xid = 0;
entry->plan = NULL;
/* this can fail, struct must be valid before */
entry->plan = make_plan(state);
valid_table:
if (state->per_tx_limit >= 0) {
TransactionId xid = GetTopTransactionId();
if (entry->last_xid != xid) {
entry->last_xid = xid;
entry->last_count = 0;
}
entry->last_count++;
if (entry->last_count > state->per_tx_limit)
elog(ERROR, "Queue '%s' allows max %d events from one TX",
TextDatumGetCString(qname), state->per_tx_limit);
}
return entry->plan;
}
/*
* ConditionalXactLockTableWait
*
* As above, but only lock if we can get the lock without blocking.
* Returns TRUE if the lock was acquired.
*/
bool
ConditionalXactLockTableWait(TransactionId xid)
{
LOCKTAG tag;
for (;;)
{
Assert(TransactionIdIsValid(xid));
Assert(!TransactionIdEquals(xid, GetTopTransactionId()));
SET_LOCKTAG_TRANSACTION(tag, xid);
if (LockAcquire(&tag, ShareLock, false, true) == LOCKACQUIRE_NOT_AVAIL)
return false;
LockRelease(&tag, ShareLock, false);
if (!TransactionIdIsInProgress(xid))
break;
xid = SubTransGetParent(xid);
}
return true;
}
/*
* AtEOXact_Snapshot
* Snapshot manager's cleanup function for end of transaction
*/
void
AtEOXact_Snapshot(bool isCommit)
{
/*
* In transaction-snapshot mode we must release our privately-managed
* reference to the transaction snapshot. We must decrement
* RegisteredSnapshots to keep the check below happy. But we don't bother
* to do FreeSnapshot, for two reasons: the memory will go away with
* TopTransactionContext anyway, and if someone has left the snapshot
* stacked as active, we don't want the code below to be chasing through
* a dangling pointer.
*/
if (FirstXactSnapshot != NULL)
{
Assert(FirstXactSnapshot->regd_count > 0);
Assert(RegisteredSnapshots > 0);
RegisteredSnapshots--;
}
FirstXactSnapshot = NULL;
/*
* If we exported any snapshots, clean them up.
*/
if (exportedSnapshots != NIL)
{
TransactionId myxid = GetTopTransactionId();
int i;
char buf[MAXPGPATH];
/*
* Get rid of the files. Unlink failure is only a WARNING because
* (1) it's too late to abort the transaction, and (2) leaving a
* leaked file around has little real consequence anyway.
*/
for (i = 1; i <= list_length(exportedSnapshots); i++)
{
XactExportFilePath(buf, myxid, i, "");
if (unlink(buf))
elog(WARNING, "could not unlink file \"%s\": %m", buf);
}
/*
* As with the FirstXactSnapshot, we needn't spend any effort on
* cleaning up the per-snapshot data structures, but we do need to
* adjust the RegisteredSnapshots count to prevent a warning below.
*
* Note: you might be thinking "why do we have the exportedSnapshots
* list at all? All we need is a counter!". You're right, but we do
* it this way in case we ever feel like improving xmin management.
*/
Assert(RegisteredSnapshots >= list_length(exportedSnapshots));
RegisteredSnapshots -= list_length(exportedSnapshots);
exportedSnapshots = NIL;
}
/* On commit, complain about leftover snapshots */
if (isCommit)
{
ActiveSnapshotElt *active;
if (RegisteredSnapshots != 0)
elog(WARNING, "%d registered snapshots seem to remain after cleanup",
RegisteredSnapshots);
/* complain about unpopped active snapshots */
for (active = ActiveSnapshot; active != NULL; active = active->as_next)
elog(WARNING, "snapshot %p still active", active);
}
/*
* And reset our state. We don't need to free the memory explicitly --
* it'll go away with TopTransactionContext.
*/
ActiveSnapshot = NULL;
RegisteredSnapshots = 0;
CurrentSnapshot = NULL;
SecondarySnapshot = NULL;
FirstSnapshotSet = false;
SnapshotResetXmin();
}
/*
* 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));
}
/*
* StrategyBufferLookup
*
* Lookup a page request in the cache directory. A buffer is only
* returned for a T1 or T2 cache hit. B1 hits are just
* remembered here, to possibly affect the behaviour later.
*
* recheck indicates we are rechecking after I/O wait; do not change
* internal status in this case.
*
* *cdb_found_index is set to the index of the found CDB, or -1 if none.
* This is not intended to be used by the caller, except to pass to
* StrategyReplaceBuffer().
*/
BufferDesc *
StrategyBufferLookup(BufferTag *tagPtr, bool recheck,
int *cdb_found_index)
{
BufferStrategyCDB *cdb;
/* Optional stats printout */
if (DebugSharedBuffers > 0)
StrategyStatsDump();
/*
* Count lookups
*/
StrategyControl->num_lookup++;
/*
* Lookup the block in the shared hash table
*/
*cdb_found_index = BufTableLookup(tagPtr);
/*
* Done if complete CDB lookup miss
*/
if (*cdb_found_index < 0)
return NULL;
/*
* We found a CDB
*/
cdb = &StrategyCDB[*cdb_found_index];
/*
* Count hits
*/
StrategyControl->num_hit[cdb->list]++;
/*
* If this is a T2 hit, we simply move the CDB to the T2 MRU position
* and return the found buffer.
*
* A CDB in T2 cannot have t1_vacuum set, so we needn't check. However,
* if the current process is VACUUM then it doesn't promote to MRU.
*/
if (cdb->list == STRAT_LIST_T2)
{
if (!strategy_hint_vacuum)
{
STRAT_LIST_REMOVE(cdb);
STRAT_MRU_INSERT(cdb, STRAT_LIST_T2);
}
return &BufferDescriptors[cdb->buf_id];
}
/*
* If this is a T1 hit, we move the buffer to the T2 MRU only if
* another transaction had read it into T1, *and* neither transaction
* is a VACUUM. This is required because any UPDATE or DELETE in
* PostgreSQL does multiple ReadBuffer(), first during the scan, later
* during the heap_update() or heap_delete(). Otherwise move to T1
* MRU. VACUUM doesn't even get to make that happen.
*/
if (cdb->list == STRAT_LIST_T1)
{
if (!strategy_hint_vacuum)
{
if (!cdb->t1_vacuum &&
!TransactionIdEquals(cdb->t1_xid, GetTopTransactionId()))
{
STRAT_LIST_REMOVE(cdb);
STRAT_MRU_INSERT(cdb, STRAT_LIST_T2);
}
else
{
STRAT_LIST_REMOVE(cdb);
STRAT_MRU_INSERT(cdb, STRAT_LIST_T1);
/*
* If a non-VACUUM process references a page recently
* loaded by VACUUM, clear the stigma; the state will now
* be the same as if this process loaded it originally.
*/
if (cdb->t1_vacuum)
{
cdb->t1_xid = GetTopTransactionId();
cdb->t1_vacuum = false;
}
}
}
return &BufferDescriptors[cdb->buf_id];
}
/*
* Even though we had seen the block in the past, its data is not
* currently in memory ... cache miss to the bufmgr.
*/
Assert(cdb->list == STRAT_LIST_B1);
return NULL;
}
/*
* StrategyReplaceBuffer
*
* Called by the buffer manager to inform us that he flushed a buffer
* and is now about to replace the content. Prior to this call,
* the cache algorithm still reports the buffer as in the cache. After
* this call we report the new block, even if IO might still need to
* be done to bring in the new content.
*
* cdb_found_index and cdb_replace_index must be the auxiliary values
* returned by previous calls to StrategyBufferLookup and StrategyGetBuffer.
*/
void
StrategyReplaceBuffer(BufferDesc *buf, BufferTag *newTag,
int cdb_found_index, int cdb_replace_index)
{
BufferStrategyCDB *cdb_found;
BufferStrategyCDB *cdb_replace;
if (cdb_found_index >= 0)
{
/* This must have been a ghost buffer cache hit (B1 list) */
cdb_found = &StrategyCDB[cdb_found_index];
/* Assert that the buffer remembered in cdb_found is the one */
/* the buffer manager is currently faulting in */
Assert(BUFFERTAGS_EQUAL(cdb_found->buf_tag, *newTag));
if (cdb_replace_index >= 0)
{
/* We are satisfying it with an evicted T buffer */
cdb_replace = &StrategyCDB[cdb_replace_index];
/* Assert that the buffer remembered in cdb_replace is */
/* the one the buffer manager has just evicted */
Assert(cdb_replace->list == STRAT_LIST_T1 ||
cdb_replace->list == STRAT_LIST_T2);
Assert(cdb_replace->buf_id == buf->buf_id);
Assert(BUFFERTAGS_EQUAL(cdb_replace->buf_tag, buf->tag));
/*
* Under normal circumstances we move evicted T1 list entries
* to the B1 list. However, T1 entries that exist only because
* of VACUUM are just thrown into the unused list instead,
* since it's unlikely they'll be touched again soon. Similarly,
* evicted T2 entries are thrown away; the LRU T2 entry cannot
* have been touched recently.
*/
if (cdb_replace->t1_vacuum || cdb_replace->list == STRAT_LIST_T2)
{
BufTableDelete(&(cdb_replace->buf_tag));
STRAT_LIST_REMOVE(cdb_replace);
cdb_replace->next = StrategyControl->listUnusedCDB;
StrategyControl->listUnusedCDB = cdb_replace_index;
}
else
{
STRAT_LIST_REMOVE(cdb_replace);
STRAT_MRU_INSERT(cdb_replace, STRAT_LIST_B1);
}
/* And clear its block reference */
cdb_replace->buf_id = -1;
}
else
{
/* We are satisfying it with an unused buffer */
}
/* Now the found B1 CDB gets the buffer and is moved to T2 */
cdb_found->buf_id = buf->buf_id;
STRAT_LIST_REMOVE(cdb_found);
STRAT_MRU_INSERT(cdb_found, STRAT_LIST_T2);
}
else
{
/*
* This was a complete cache miss, so we need to create a new CDB.
* We use a free one if available, else reclaim the tail end of B1.
*/
if (StrategyControl->listUnusedCDB >= 0)
{
cdb_found = &StrategyCDB[StrategyControl->listUnusedCDB];
StrategyControl->listUnusedCDB = cdb_found->next;
}
else
{
/* Can't fail because we have more CDBs than buffers... */
if (B1_LENGTH == 0)
elog(PANIC, "StrategyReplaceBuffer: out of CDBs");
cdb_found = &StrategyCDB[StrategyControl->listHead[STRAT_LIST_B1]];
BufTableDelete(&(cdb_found->buf_tag));
STRAT_LIST_REMOVE(cdb_found);
}
/* Set the CDB's buf_tag and insert it into the hash table */
cdb_found->buf_tag = *newTag;
BufTableInsert(&(cdb_found->buf_tag), (cdb_found - StrategyCDB));
if (cdb_replace_index >= 0)
{
/*
* The buffer was formerly in a T list, move its CDB to the
* appropriate list: B1 if T1, else discard it, as above
*/
cdb_replace = &StrategyCDB[cdb_replace_index];
Assert(cdb_replace->list == STRAT_LIST_T1 ||
cdb_replace->list == STRAT_LIST_T2);
Assert(cdb_replace->buf_id == buf->buf_id);
Assert(BUFFERTAGS_EQUAL(cdb_replace->buf_tag, buf->tag));
if (cdb_replace->list == STRAT_LIST_T1)
{
STRAT_LIST_REMOVE(cdb_replace);
STRAT_MRU_INSERT(cdb_replace, STRAT_LIST_B1);
}
else
{
BufTableDelete(&(cdb_replace->buf_tag));
STRAT_LIST_REMOVE(cdb_replace);
cdb_replace->next = StrategyControl->listUnusedCDB;
StrategyControl->listUnusedCDB = cdb_replace_index;
}
/* And clear its block reference */
cdb_replace->buf_id = -1;
}
else
{
/* We are satisfying it with an unused buffer */
}
/* Assign the buffer id to the new CDB */
cdb_found->buf_id = buf->buf_id;
/*
* Specialized VACUUM optimization. If this complete cache miss
* happened because vacuum needed the page, we place it at the LRU
* position of T1; normally it goes at the MRU position.
*/
if (strategy_hint_vacuum)
{
if (TransactionIdEquals(strategy_vacuum_xid,
GetTopTransactionId()))
STRAT_LRU_INSERT(cdb_found, STRAT_LIST_T1);
else
{
/* VACUUM must have been aborted by error, reset flag */
strategy_hint_vacuum = false;
STRAT_MRU_INSERT(cdb_found, STRAT_LIST_T1);
}
}
else
STRAT_MRU_INSERT(cdb_found, STRAT_LIST_T1);
/*
* Remember the Xid when this buffer went onto T1 to avoid a
* single UPDATE promoting a newcomer straight into T2. Also
* remember if it was loaded for VACUUM.
*/
cdb_found->t1_xid = GetTopTransactionId();
cdb_found->t1_vacuum = strategy_hint_vacuum;
}
}
/*
* 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;
}
/*
* AlterSequence
*
* Modify the definition of a sequence relation
*/
ObjectAddress
AlterSequence(AlterSeqStmt *stmt)
{
Oid relid;
SeqTable elm;
Relation seqrel;
Buffer buf;
HeapTupleData seqtuple;
Form_pg_sequence seq;
FormData_pg_sequence new___;
List *owned_by;
ObjectAddress address;
/* Open and lock sequence. */
relid = RangeVarGetRelid(stmt->sequence, AccessShareLock, stmt->missing_ok);
if (relid == InvalidOid)
{
ereport(NOTICE,
(errmsg("relation \"%s\" does not exist, skipping",
stmt->sequence->relname)));
return InvalidObjectAddress;
}
init_sequence(relid, &elm, &seqrel);
/* allow ALTER to sequence owner only */
if (!pg_class_ownercheck(relid, GetUserId()))
aclcheck_error(ACLCHECK_NOT_OWNER, ACL_KIND_CLASS,
stmt->sequence->relname);
/* lock page' buffer and read tuple into new___ sequence structure */
seq = read_seq_tuple(elm, seqrel, &buf, &seqtuple);
/* Copy old values of options into workspace */
memcpy(&new___, seq, sizeof(FormData_pg_sequence));
/* Check and set new___ values */
init_params(stmt->options, false, &new___, &owned_by);
/* Clear local cache so that we don't think we have cached numbers */
/* Note that we do not change the currval() state */
elm->cached = elm->last;
/* check the comment above nextval_internal()'s equivalent call. */
if (RelationNeedsWAL(seqrel))
GetTopTransactionId();
/* Now okay to update the on-disk tuple */
START_CRIT_SECTION();
memcpy(seq, &new___, sizeof(FormData_pg_sequence));
MarkBufferDirty(buf);
/* XLOG stuff */
if (RelationNeedsWAL(seqrel))
{
xl_seq_rec xlrec;
XLogRecPtr recptr;
Page page = BufferGetPage(buf);
XLogBeginInsert();
XLogRegisterBuffer(0, buf, REGBUF_WILL_INIT);
xlrec.node = seqrel->rd_node;
XLogRegisterData((char *) &xlrec, sizeof(xl_seq_rec));
XLogRegisterData((char *) seqtuple.t_data, seqtuple.t_len);
recptr = XLogInsert(RM_SEQ_ID, XLOG_SEQ_LOG);
PageSetLSN(page, recptr);
}
END_CRIT_SECTION();
UnlockReleaseBuffer(buf);
/* process OWNED BY if given */
if (owned_by)
process_owned_by(seqrel, owned_by);
InvokeObjectPostAlterHook(RelationRelationId, relid, 0);
ObjectAddressSet(address, RelationRelationId, relid);
relation_close(seqrel, NoLock);
return address;
}
/*
* Indicate we intend to create a tablespace 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 tablespace files that are going
* to be created, call ~_DoPendingCreates to do the actual file-system creates. (See its
* note on XLOG flushing).
*/
void
PersistentTablespace_MarkCreatePending(
Oid filespaceOid,
/* The filespace where the tablespace lives. */
Oid tablespaceOid,
/* The tablespace OID for the create. */
MirroredObjectExistenceState mirrorExistenceState,
ItemPointer persistentTid,
/* TID of the gp_persistent_rel_files tuple for the rel file */
int64 *persistentSerialNum,
bool flushToXLog)
/* When true, the XLOG record for this change will be flushed to disk. */
{
WRITE_PERSISTENT_STATE_ORDERED_LOCK_DECLARE;
PersistentFileSysObjName fsObjName;
TablespaceDirEntry tablespaceDirEntry;
TransactionId topXid;
if (Persistent_BeforePersistenceWork())
{
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Skipping persistent tablespace %u because we are before persistence work",
tablespaceOid);
return;
/*
* The initdb process will load the persistent table once we out of
* bootstrap mode.
*/
}
PersistentTablespace_VerifyInitScan();
PersistentFileSysObjName_SetTablespaceDir(&fsObjName, tablespaceOid);
topXid = GetTopTransactionId();
WRITE_PERSISTENT_STATE_ORDERED_LOCK;
PersistentTablespace_AddTuple(
filespaceOid,
tablespaceOid,
PersistentFileSysState_CreatePending,
/* createMirrorDataLossTrackingSessionNum */ 0,
mirrorExistenceState,
/* reserved */ 0,
/* parentXid */ topXid,
flushToXLog,
persistentTid,
persistentSerialNum);
WRITE_TABLESPACE_HASH_LOCK;
tablespaceDirEntry =
PersistentTablespace_CreateEntryUnderLock(filespaceOid, tablespaceOid);
Assert(tablespaceDirEntry != NULL);
tablespaceDirEntry->state = PersistentFileSysState_CreatePending;
ItemPointerCopy(persistentTid, &tablespaceDirEntry->persistentTid);
tablespaceDirEntry->persistentSerialNum = *persistentSerialNum;
WRITE_TABLESPACE_HASH_UNLOCK;
/*
* This XLOG must be generated under the persistent write-lock.
*/
#ifdef MASTER_MIRROR_SYNC
mmxlog_log_create_tablespace(
filespaceOid,
tablespaceOid);
#endif
SIMPLE_FAULT_INJECTOR(FaultBeforePendingDeleteTablespaceEntry);
/*
* 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 tablespace directory: Add '%s' in state 'Created', mirror existence state '%s', serial number " INT64_FORMAT " at TID %s",
PersistentFileSysObjName_ObjectName(&fsObjName),
MirroredObjectExistenceState_Name(mirrorExistenceState),
*persistentSerialNum,
ItemPointerToString(persistentTid));
}
/*
* StrategyHintVacuum -- tell us whether VACUUM is active
*/
void
StrategyHintVacuum(bool vacuum_active)
{
strategy_hint_vacuum = vacuum_active;
strategy_vacuum_xid = GetTopTransactionId();
}
/*
* Indicate we intend to create a relation file as part of the current transaction.
*
* This function adds an entry in 'gp_persistent_relation_node' for either a new table (segment file
* # 0) or a new segment file under AO table (segment file # > 0 for row/column-oriented AO) with a state
* 'Create Pending'. An XLOG IntentToCreate record is generated that will guard the subsequent file-system
* create in case the transaction aborts.
*
* Paramaters
* -----------
* relFileNode = The tablespace, database, and relation OIDs for the create
* segmentFileNum = As the name implies ( 0 for heap
* >= 0 for RO/CO AO as applicable)
* relStorageMgr = Persistent Relation storage Manager
* relBufpoolKind = Buffer pool type beneath corrosponding relation
* TODO bufferPollBulkLoad = ???
* TODO mirrorExistenceState = ???
* TODO relDataSynchronizationState = ???
* flushToXlog = If true, the XLOG record for this change will be flushed to disk.
* TODO isLocalBuf = ???
*
* Return
* ------
* relationName = Name of the relation used for either debugging or to store in PendingDelete LL.
* persistentTid = Resulting TID of the gp_persistent_rel_files tuple for the relation
* serialNum = Resulting serial number for the relation. Distinquishes the uses of the tuple
*/
void PersistentRelation_AddCreatePending(
RelFileNode *relFileNode,
int32 segmentFileNum,
PersistentFileSysRelStorageMgr relStorageMgr,
PersistentFileSysRelBufpoolKind relBufpoolKind,
bool bufferPoolBulkLoad,
MirroredObjectExistenceState mirrorExistenceState,
MirroredRelDataSynchronizationState relDataSynchronizationState,
char *relationName,
ItemPointer persistentTid,
int64 *serialNum,
bool flushToXLog,
bool isLocalBuf)
{
WRITE_PERSISTENT_STATE_ORDERED_LOCK_DECLARE;
PersistentFileSysObjName fsObjName;
XLogRecPtr mirrorBufpoolResyncCkptLoc;
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));
MemSet(&mirrorBufpoolResyncCkptLoc, 0, sizeof(XLogRecPtr));
if (Persistent_BeforePersistenceWork())
{
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Skipping persistent relation '%s' because we are before persistence work",
relpath(*relFileNode));
MemSet(persistentTid, 0, sizeof(ItemPointerData));
*serialNum = 0;
return; // The initdb process will load the persistent table once we out of bootstrap mode.
}
/* Verify if the needed shared mem data structures for persistent tables are setup and inited */
PersistentRelation_VerifyInitScan();
/* Setup the file system object name */
PersistentFileSysObjName_SetRelationFile(
&fsObjName,
relFileNode,
segmentFileNum);
WRITE_PERSISTENT_STATE_ORDERED_LOCK;
/* Create a values array which will be used to create a 'gp_persistent_relation_node' tuple */
GpPersistentRelationNode_SetDatumValues(
values,
relFileNode->spcNode,
relFileNode->dbNode,
relFileNode->relNode,
segmentFileNum,
relStorageMgr,
(bufferPoolBulkLoad ?
PersistentFileSysState_BulkLoadCreatePending :
PersistentFileSysState_CreatePending),
/* createMirrorDataLossTrackingSessionNum */ 0,
mirrorExistenceState,
relDataSynchronizationState,
/* mirrorBufpoolMarkedForScanIncrementalResync */ false,
/* mirrorBufpoolResyncChangedPageCount */ 0,
&mirrorBufpoolResyncCkptLoc,
/* mirrorBufpoolResyncCkptBlockNum */ 0,
/* mirrorAppendOnlyLossEof */ 0,
/* mirrorAppendOnlyNewEof */ 0,
relBufpoolKind,
GetTopTransactionId(),
/* persistentSerialNum */ 0, // This will be set by PersistentFileSysObj_AddTuple.
&previousFreeTid);
/* Add a new tuple to 'gp_persistent_relation_node' table for the new relation/segment file
* we intend to create. This will also create and apply a new persistent serial number. */
PersistentFileSysObj_AddTuple(
PersistentFsObjType_RelationFile,
values,
flushToXLog,
persistentTid,
serialNum);
/*
* This XLOG must be generated under the persistent write-lock.
*/
#ifdef MASTER_MIRROR_SYNC
mmxlog_log_create_relfilenode(
relFileNode->spcNode,
relFileNode->dbNode,
relFileNode->relNode,
segmentFileNum);
#endif
#ifdef FAULT_INJECTOR
FaultInjector_InjectFaultIfSet(
FaultBeforePendingDeleteRelationEntry,
DDLNotSpecified,
"", // databaseName
""); // tableName
#endif
/* We'll add an entry to the PendingDelete LinkedList (LL) to remeber what we
* created in this transaction (or sub-transaction). If the transaction
* aborts then we can search for all such entries in this LL and get rid of (delete)
* such relations or segment files on the disk.
*
* MPP-18228
* To make adding 'Create Pending' entry to persistent table and adding
* to the PendingDelete list atomic
*/
PendingDelete_AddCreatePendingRelationEntry(
&fsObjName,
persistentTid,
serialNum,
relStorageMgr,
relationName,
isLocalBuf,
bufferPoolBulkLoad);
WRITE_PERSISTENT_STATE_ORDERED_UNLOCK;
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Persistent relation: Add '%s', relation name '%s' in state 'Create Pending', relation storage manager '%s', mirror existence state '%s', relation data resynchronization state '%s', serial number " INT64_FORMAT " at TID %s",
PersistentFileSysObjName_ObjectName(&fsObjName),
relationName,
PersistentFileSysRelStorageMgr_Name(relStorageMgr),
MirroredObjectExistenceState_Name(mirrorExistenceState),
MirroredRelDataSynchronizationState_Name(relDataSynchronizationState),
*serialNum,
ItemPointerToString(persistentTid));
}
Datum
_Slony_I_createEvent(PG_FUNCTION_ARGS)
{
TransactionId newXid = GetTopTransactionId();
Slony_I_ClusterStatus *cs;
char *ev_type_c;
Datum argv[9];
char nulls[10];
char *buf;
size_t buf_size;
int rc;
int i;
int64 retval;
bool isnull;
#ifdef HAVE_GETACTIVESNAPSHOT
if (GetActiveSnapshot() == NULL)
elog(ERROR, "Slony-I: ActiveSnapshot is NULL in createEvent()");
#else
if (SerializableSnapshot == NULL)
elog(ERROR, "Slony-I: SerializableSnapshot is NULL in createEvent()");
#endif
if ((rc = SPI_connect()) < 0)
elog(ERROR, "Slony-I: SPI_connect() failed in createEvent()");
/*
* Get or create the cluster status information and make sure it has the
* SPI plans that we need here.
*/
cs = getClusterStatus(PG_GETARG_NAME(0),
PLAN_INSERT_EVENT);
buf_size = 8192;
buf = palloc(buf_size);
/*
* Do the following only once per transaction.
*/
if (!TransactionIdEquals(cs->currentXid, newXid))
{
cs->currentXid = newXid;
}
/*
* Call the saved INSERT plan
*/
for (i = 1; i < 10; i++)
{
if (i >= PG_NARGS() || PG_ARGISNULL(i))
{
argv[i - 1] = (Datum) 0;
nulls[i - 1] = 'n';
}
else
{
argv[i - 1] = PG_GETARG_DATUM(i);
nulls[i - 1] = ' ';
}
}
nulls[9] = '\0';
if ((rc = SPI_execp(cs->plan_insert_event, argv, nulls, 0)) < 0)
elog(ERROR, "Slony-I: SPI_execp() failed for \"INSERT INTO sl_event ...\"");
/*
* The INSERT plan also contains a SELECT currval('sl_event_seq'), use the
* new sequence number as return value.
*/
if (SPI_processed != 1)
elog(ERROR, "Slony-I: INSERT plan did not return 1 result row");
retval = DatumGetInt64(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc, 1, &isnull));
/*
* For SYNC and ENABLE_SUBSCRIPTION events, we also remember all current
* sequence values.
*/
if (PG_NARGS() > 1 && !PG_ARGISNULL(1))
{
ev_type_c = DatumGetPointer(DirectFunctionCall1(
textout, PG_GETARG_DATUM(1)));
if (strcmp(ev_type_c, "SYNC") == 0 ||
strcmp(ev_type_c, "ENABLE_SUBSCRIPTION") == 0)
{
/*@-nullpass@*/
if ((rc = SPI_execp(cs->plan_record_sequences, NULL, NULL, 0)) < 0)
elog(ERROR, "Slony-I: SPI_execp() failed for \"INSERT INTO sl_seqlog ...\"");
/*@+nullpass@*/
}
}
(void) SPI_finish();
/*@-mustfreefresh@*/
PG_RETURN_INT64(retval);
}
/*
* Main internal procedure that handles 2 & 3 arg forms of SETVAL.
*
* Note that the 3 arg version (which sets the is_called flag) is
* only for use in pg_dump, and setting the is_called flag may not
* work if multiple users are attached to the database and referencing
* the sequence (unlikely if pg_dump is restoring it).
*
* It is necessary to have the 3 arg version so that pg_dump can
* restore the state of a sequence exactly during data-only restores -
* it is the only way to clear the is_called flag in an existing
* sequence.
*/
static void
do_setval(Oid relid, int64 next, bool iscalled)
{
SeqTable elm;
Relation seqrel;
Buffer buf;
HeapTupleData seqtuple;
Form_pg_sequence seq;
/* open and AccessShareLock sequence */
init_sequence(relid, &elm, &seqrel);
if (pg_class_aclcheck(elm->relid, GetUserId(), ACL_UPDATE) != ACLCHECK_OK)
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("permission denied for sequence %s",
RelationGetRelationName(seqrel))));
/* read-only transactions may only modify temp sequences */
if (!seqrel->rd_islocaltemp)
PreventCommandIfReadOnly("setval()");
/*
* Forbid this during parallel operation because, to make it work,
* the cooperating backends would need to share the backend-local cached
* sequence information. Currently, we don't support that.
*/
PreventCommandIfParallelMode("setval()");
/* lock page' buffer and read tuple */
seq = read_seq_tuple(elm, seqrel, &buf, &seqtuple);
if ((next < seq->min_value) || (next > seq->max_value))
{
char bufv[100],
bufm[100],
bufx[100];
snprintf(bufv, sizeof(bufv), INT64_FORMAT, next);
snprintf(bufm, sizeof(bufm), INT64_FORMAT, seq->min_value);
snprintf(bufx, sizeof(bufx), INT64_FORMAT, seq->max_value);
ereport(ERROR,
(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
errmsg("setval: value %s is out of bounds for sequence \"%s\" (%s..%s)",
bufv, RelationGetRelationName(seqrel),
bufm, bufx)));
}
/* Set the currval() state only if iscalled = true */
if (iscalled)
{
elm->last = next; /* last returned number */
elm->last_valid = true;
}
/* In any case, forget any future cached numbers */
elm->cached = elm->last;
/* check the comment above nextval_internal()'s equivalent call. */
if (RelationNeedsWAL(seqrel))
GetTopTransactionId();
/* ready to change the on-disk (or really, in-buffer) tuple */
START_CRIT_SECTION();
seq->last_value = next; /* last fetched number */
seq->is_called = iscalled;
seq->log_cnt = 0;
MarkBufferDirty(buf);
/* XLOG stuff */
if (RelationNeedsWAL(seqrel))
{
xl_seq_rec xlrec;
XLogRecPtr recptr;
Page page = BufferGetPage(buf);
XLogBeginInsert();
XLogRegisterBuffer(0, buf, REGBUF_WILL_INIT);
xlrec.node = seqrel->rd_node;
XLogRegisterData((char *) &xlrec, sizeof(xl_seq_rec));
XLogRegisterData((char *) seqtuple.t_data, seqtuple.t_len);
recptr = XLogInsert(RM_SEQ_ID, XLOG_SEQ_LOG);
PageSetLSN(page, recptr);
}
END_CRIT_SECTION();
UnlockReleaseBuffer(buf);
relation_close(seqrel, NoLock);
}
static int64
nextval_internal(Oid relid)
{
SeqTable elm;
Relation seqrel;
Buffer buf;
Page page;
HeapTupleData seqtuple;
Form_pg_sequence seq;
int64 incby,
maxv,
minv,
cache,
log,
fetch,
last;
int64 result,
next,
rescnt = 0;
bool logit = false;
/* open and AccessShareLock sequence */
init_sequence(relid, &elm, &seqrel);
if (pg_class_aclcheck(elm->relid, GetUserId(),
ACL_USAGE | ACL_UPDATE) != ACLCHECK_OK)
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("permission denied for sequence %s",
RelationGetRelationName(seqrel))));
/* read-only transactions may only modify temp sequences */
if (!seqrel->rd_islocaltemp)
PreventCommandIfReadOnly("nextval()");
/*
* Forbid this during parallel operation because, to make it work,
* the cooperating backends would need to share the backend-local cached
* sequence information. Currently, we don't support that.
*/
PreventCommandIfParallelMode("nextval()");
if (elm->last != elm->cached) /* some numbers were cached */
{
Assert(elm->last_valid);
Assert(elm->increment != 0);
elm->last += elm->increment;
relation_close(seqrel, NoLock);
last_used_seq = elm;
return elm->last;
}
/* lock page' buffer and read tuple */
seq = read_seq_tuple(elm, seqrel, &buf, &seqtuple);
page = BufferGetPage(buf);
last = next = result = seq->last_value;
incby = seq->increment_by;
maxv = seq->max_value;
minv = seq->min_value;
fetch = cache = seq->cache_value;
log = seq->log_cnt;
if (!seq->is_called)
{
rescnt++; /* return last_value if not is_called */
fetch--;
}
/*
* Decide whether we should emit a WAL log record. If so, force up the
* fetch count to grab SEQ_LOG_VALS more values than we actually need to
* cache. (These will then be usable without logging.)
*
* If this is the first nextval after a checkpoint, we must force a new___
* WAL record to be written anyway, else replay starting from the
* checkpoint would fail to advance the sequence past the logged values.
* In this case we may as well fetch extra values.
*/
if (log < fetch || !seq->is_called)
{
/* forced log to satisfy local demand for values */
fetch = log = fetch + SEQ_LOG_VALS;
logit = true;
}
else
{
XLogRecPtr redoptr = GetRedoRecPtr();
if (PageGetLSN(page) <= redoptr)
{
/* last update of seq was before checkpoint */
fetch = log = fetch + SEQ_LOG_VALS;
logit = true;
}
}
while (fetch) /* try to fetch cache [+ log ] numbers */
{
/*
* Check MAXVALUE for ascending sequences and MINVALUE for descending
* sequences
*/
if (incby > 0)
{
/* ascending sequence */
if ((maxv >= 0 && next > maxv - incby) ||
(maxv < 0 && next + incby > maxv))
{
if (rescnt > 0)
break; /* stop fetching */
if (!seq->is_cycled)
{
char buf[100];
snprintf(buf, sizeof(buf), INT64_FORMAT, maxv);
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("nextval: reached maximum value of sequence \"%s\" (%s)",
RelationGetRelationName(seqrel), buf)));
}
next = minv;
}
else
next += incby;
}
else
{
/* descending sequence */
if ((minv < 0 && next < minv - incby) ||
(minv >= 0 && next + incby < minv))
{
if (rescnt > 0)
break; /* stop fetching */
if (!seq->is_cycled)
{
char buf[100];
snprintf(buf, sizeof(buf), INT64_FORMAT, minv);
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("nextval: reached minimum value of sequence \"%s\" (%s)",
RelationGetRelationName(seqrel), buf)));
}
next = maxv;
}
else
next += incby;
}
fetch--;
if (rescnt < cache)
{
log--;
rescnt++;
last = next;
if (rescnt == 1) /* if it's first result - */
result = next; /* it's what to return */
}
}
log -= fetch; /* adjust for any unfetched numbers */
Assert(log >= 0);
/* save info in local cache */
elm->last = result; /* last returned number */
elm->cached = last; /* last fetched number */
elm->last_valid = true;
last_used_seq = elm;
/*
* If something needs to be WAL logged, acquire an xid, so this
* transaction's commit will trigger a WAL flush and wait for
* syncrep. It's sufficient to ensure the toplevel transaction has an xid,
* no need to assign xids subxacts, that'll already trigger an appropriate
* wait. (Have to do that here, so we're outside the critical section)
*/
if (logit && RelationNeedsWAL(seqrel))
GetTopTransactionId();
/* ready to change the on-disk (or really, in-buffer) tuple */
START_CRIT_SECTION();
/*
* We must mark the buffer dirty before doing XLogInsert(); see notes in
* SyncOneBuffer(). However, we don't apply the desired changes just yet.
* This looks like a violation of the buffer update protocol, but it is in
* fact safe because we hold exclusive lock on the buffer. Any other
* process, including a checkpoint, that tries to examine the buffer
* contents will block until we release the lock, and then will see the
* final state that we install below.
*/
MarkBufferDirty(buf);
/* XLOG stuff */
if (logit && RelationNeedsWAL(seqrel))
{
xl_seq_rec xlrec;
XLogRecPtr recptr;
/*
* We don't log the current state of the tuple, but rather the state
* as it would appear after "log" more fetches. This lets us skip
* that many future WAL records, at the cost that we lose those
* sequence values if we crash.
*/
XLogBeginInsert();
XLogRegisterBuffer(0, buf, REGBUF_WILL_INIT);
/* set values that will be saved in xlog */
seq->last_value = next;
seq->is_called = true;
seq->log_cnt = 0;
xlrec.node = seqrel->rd_node;
XLogRegisterData((char *) &xlrec, sizeof(xl_seq_rec));
XLogRegisterData((char *) seqtuple.t_data, seqtuple.t_len);
recptr = XLogInsert(RM_SEQ_ID, XLOG_SEQ_LOG);
PageSetLSN(page, recptr);
}
/* Now update sequence tuple to the intended final state */
seq->last_value = last; /* last fetched number */
seq->is_called = true;
seq->log_cnt = log; /* how much is logged */
END_CRIT_SECTION();
UnlockReleaseBuffer(buf);
relation_close(seqrel, NoLock);
return result;
}
/*
* ExportSnapshot
* Export the snapshot to a file so that other backends can import it.
* Returns the token (the file name) that can be used to import this
* snapshot.
*/
static char *
ExportSnapshot(Snapshot snapshot)
{
TransactionId topXid;
TransactionId *children;
int nchildren;
int addTopXid;
StringInfoData buf;
FILE *f;
int i;
MemoryContext oldcxt;
char path[MAXPGPATH];
char pathtmp[MAXPGPATH];
/*
* It's tempting to call RequireTransactionChain here, since it's not
* very useful to export a snapshot that will disappear immediately
* afterwards. However, we haven't got enough information to do that,
* since we don't know if we're at top level or not. For example, we
* could be inside a plpgsql function that is going to fire off other
* transactions via dblink. Rather than disallow perfectly legitimate
* usages, don't make a check.
*
* Also note that we don't make any restriction on the transaction's
* isolation level; however, importers must check the level if they
* are serializable.
*/
/*
* This will assign a transaction ID if we do not yet have one.
*/
topXid = GetTopTransactionId();
/*
* We cannot export a snapshot from a subtransaction because there's no
* easy way for importers to verify that the same subtransaction is still
* running.
*/
if (IsSubTransaction())
ereport(ERROR,
(errcode(ERRCODE_ACTIVE_SQL_TRANSACTION),
errmsg("cannot export a snapshot from a subtransaction")));
/*
* We do however allow previous committed subtransactions to exist.
* Importers of the snapshot must see them as still running, so get their
* XIDs to add them to the snapshot.
*/
nchildren = xactGetCommittedChildren(&children);
/*
* Copy the snapshot into TopTransactionContext, add it to the
* exportedSnapshots list, and mark it pseudo-registered. We do this to
* ensure that the snapshot's xmin is honored for the rest of the
* transaction. (Right now, because SnapshotResetXmin is so stupid, this
* is overkill; but later we might make that routine smarter.)
*/
snapshot = CopySnapshot(snapshot);
oldcxt = MemoryContextSwitchTo(TopTransactionContext);
exportedSnapshots = lappend(exportedSnapshots, snapshot);
MemoryContextSwitchTo(oldcxt);
snapshot->regd_count++;
RegisteredSnapshots++;
/*
* Fill buf with a text serialization of the snapshot, plus identification
* data about this transaction. The format expected by ImportSnapshot
* is pretty rigid: each line must be fieldname:value.
*/
initStringInfo(&buf);
appendStringInfo(&buf, "xid:%u\n", topXid);
appendStringInfo(&buf, "dbid:%u\n", MyDatabaseId);
appendStringInfo(&buf, "iso:%d\n", XactIsoLevel);
appendStringInfo(&buf, "ro:%d\n", XactReadOnly);
appendStringInfo(&buf, "xmin:%u\n", snapshot->xmin);
appendStringInfo(&buf, "xmax:%u\n", snapshot->xmax);
/*
* We must include our own top transaction ID in the top-xid data, since
* by definition we will still be running when the importing transaction
* adopts the snapshot, but GetSnapshotData never includes our own XID in
* the snapshot. (There must, therefore, be enough room to add it.)
*
* However, it could be that our topXid is after the xmax, in which case
* we shouldn't include it because xip[] members are expected to be before
* xmax. (We need not make the same check for subxip[] members, see
* snapshot.h.)
*/
addTopXid = TransactionIdPrecedes(topXid, snapshot->xmax) ? 1 : 0;
appendStringInfo(&buf, "xcnt:%d\n", snapshot->xcnt + addTopXid);
for (i = 0; i < snapshot->xcnt; i++)
appendStringInfo(&buf, "xip:%u\n", snapshot->xip[i]);
if (addTopXid)
appendStringInfo(&buf, "xip:%u\n", topXid);
/*
* Similarly, we add our subcommitted child XIDs to the subxid data.
* Here, we have to cope with possible overflow.
*/
if (snapshot->suboverflowed ||
snapshot->subxcnt + nchildren > GetMaxSnapshotSubxidCount())
appendStringInfoString(&buf, "sof:1\n");
else
{
appendStringInfoString(&buf, "sof:0\n");
appendStringInfo(&buf, "sxcnt:%d\n", snapshot->subxcnt + nchildren);
for (i = 0; i < snapshot->subxcnt; i++)
appendStringInfo(&buf, "sxp:%u\n", snapshot->subxip[i]);
for (i = 0; i < nchildren; i++)
appendStringInfo(&buf, "sxp:%u\n", children[i]);
}
appendStringInfo(&buf, "rec:%u\n", snapshot->takenDuringRecovery);
/*
* Now write the text representation into a file. We first write to a
* ".tmp" filename, and rename to final filename if no error. This
* ensures that no other backend can read an incomplete file
* (ImportSnapshot won't allow it because of its valid-characters check).
*/
XactExportFilePath(pathtmp, topXid, list_length(exportedSnapshots), ".tmp");
if (!(f = AllocateFile(pathtmp, PG_BINARY_W)))
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create file \"%s\": %m", pathtmp)));
if (fwrite(buf.data, buf.len, 1, f) != 1)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not write to file \"%s\": %m", pathtmp)));
/* no fsync() since file need not survive a system crash */
if (FreeFile(f))
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not write to file \"%s\": %m", pathtmp)));
/*
* Now that we have written everything into a .tmp file, rename the file
* to remove the .tmp suffix.
*/
XactExportFilePath(path, topXid, list_length(exportedSnapshots), "");
if (rename(pathtmp, path) < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not rename file \"%s\" to \"%s\": %m",
pathtmp, path)));
/*
* The basename of the file is what we return from pg_export_snapshot().
* It's already in path in a textual format and we know that the path
* starts with SNAPSHOT_EXPORT_DIR. Skip over the prefix and the slash
* and pstrdup it so as not to return the address of a local variable.
*/
return pstrdup(path + strlen(SNAPSHOT_EXPORT_DIR) + 1);
}
Datum
_Slony_I_logTrigger(PG_FUNCTION_ARGS)
{
TransactionId newXid = GetTopTransactionId();
Slony_I_ClusterStatus *cs;
TriggerData *tg;
Datum argv[4];
text *cmdtype = NULL;
int rc;
Name cluster_name;
int32 tab_id;
char *attkind;
int attkind_idx;
int cmddata_need;
/*
* Don't do any logging if the current session role isn't Origin.
*/
if (SessionReplicationRole != SESSION_REPLICATION_ROLE_ORIGIN)
return PointerGetDatum(NULL);
/*
* Get the trigger call context
*/
if (!CALLED_AS_TRIGGER(fcinfo))
elog(ERROR, "Slony-I: logTrigger() not called as trigger");
tg = (TriggerData *) (fcinfo->context);
/*
* Check all logTrigger() calling conventions
*/
if (!TRIGGER_FIRED_AFTER(tg->tg_event))
elog(ERROR, "Slony-I: logTrigger() must be fired AFTER");
if (!TRIGGER_FIRED_FOR_ROW(tg->tg_event))
elog(ERROR, "Slony-I: logTrigger() must be fired FOR EACH ROW");
if (tg->tg_trigger->tgnargs != 3)
elog(ERROR, "Slony-I: logTrigger() must be defined with 3 args");
/*
* Connect to the SPI manager
*/
if ((rc = SPI_connect()) < 0)
elog(ERROR, "Slony-I: SPI_connect() failed in createEvent()");
/*
* Get all the trigger arguments
*/
cluster_name = DatumGetName(DirectFunctionCall1(namein,
CStringGetDatum(tg->tg_trigger->tgargs[0])));
tab_id = strtol(tg->tg_trigger->tgargs[1], NULL, 10);
attkind = tg->tg_trigger->tgargs[2];
/*
* Get or create the cluster status information and make sure it has the
* SPI plans that we need here.
*/
cs = getClusterStatus(cluster_name, PLAN_INSERT_LOG);
/*
* Do the following only once per transaction.
*/
if (!TransactionIdEquals(cs->currentXid, newXid))
{
int32 log_status;
bool isnull;
/*
* Determine the currently active log table
*/
if (SPI_execp(cs->plan_get_logstatus, NULL, NULL, 0) < 0)
elog(ERROR, "Slony-I: cannot determine log status");
if (SPI_processed != 1)
elog(ERROR, "Slony-I: cannot determine log status");
log_status = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc, 1, &isnull));
SPI_freetuptable(SPI_tuptable);
switch (log_status)
{
case 0:
case 2:
cs->plan_active_log = cs->plan_insert_log_1;
break;
case 1:
case 3:
cs->plan_active_log = cs->plan_insert_log_2;
break;
default:
elog(ERROR, "Slony-I: illegal log status %d", log_status);
break;
}
cs->currentXid = newXid;
}
/*
* Determine cmdtype and cmddata depending on the command type
*/
if (TRIGGER_FIRED_BY_INSERT(tg->tg_event))
{
HeapTuple new_row = tg->tg_trigtuple;
TupleDesc tupdesc = tg->tg_relation->rd_att;
char *col_ident;
char *col_value;
int len_ident;
int len_value;
int i;
int need_comma = false;
char *OldDateStyle;
char *cp = VARDATA(cs->cmddata_buf);
/*
* INSERT
*
* cmdtype = 'I' cmddata = ("col" [, ...]) values ('value' [, ...])
*/
cmdtype = cs->cmdtype_I;
/*
* Specify all the columns
*/
*cp++ = '(';
for (i = 0; i < tg->tg_relation->rd_att->natts; i++)
{
/*
* Skip dropped columns
*/
if (tupdesc->attrs[i]->attisdropped)
continue;
col_ident = (char *) slon_quote_identifier(SPI_fname(tupdesc, i + 1));
cmddata_need = (cp - (char *) (cs->cmddata_buf)) + 16 +
(len_ident = strlen(col_ident));
if (cs->cmddata_size < cmddata_need)
{
int have = (cp - (char *) (cs->cmddata_buf));
while (cs->cmddata_size < cmddata_need)
cs->cmddata_size *= 2;
cs->cmddata_buf = realloc(cs->cmddata_buf, cs->cmddata_size);
cp = (char *) (cs->cmddata_buf) + have;
}
if (need_comma)
*cp++ = ',';
else
need_comma = true;
memcpy(cp, col_ident, len_ident);
cp += len_ident;
}
/*
* Append the string ") values ("
*/
*cp++ = ')';
*cp++ = ' ';
*cp++ = 'v';
*cp++ = 'a';
*cp++ = 'l';
*cp++ = 'u';
*cp++ = 'e';
*cp++ = 's';
*cp++ = ' ';
*cp++ = '(';
/*
* Append the values
*/
need_comma = false;
OldDateStyle = GetConfigOptionByName("DateStyle", NULL);
if (!strstr(OldDateStyle, "ISO"))
set_config_option("DateStyle", "ISO", PGC_USERSET, PGC_S_SESSION, true, true);
for (i = 0; i < tg->tg_relation->rd_att->natts; i++)
{
/*
* Skip dropped columns
*/
if (tupdesc->attrs[i]->attisdropped)
continue;
if ((col_value = SPI_getvalue(new_row, tupdesc, i + 1)) == NULL)
{
col_value = "NULL";
}
else
{
col_value = slon_quote_literal(col_value);
}
cmddata_need = (cp - (char *) (cs->cmddata_buf)) + 16 +
(len_value = strlen(col_value));
if (cs->cmddata_size < cmddata_need)
{
int have = (cp - (char *) (cs->cmddata_buf));
while (cs->cmddata_size < cmddata_need)
cs->cmddata_size *= 2;
cs->cmddata_buf = realloc(cs->cmddata_buf, cs->cmddata_size);
cp = (char *) (cs->cmddata_buf) + have;
}
if (need_comma)
*cp++ = ',';
else
need_comma = true;
memcpy(cp, col_value, len_value);
cp += len_value;
}
if (!strstr(OldDateStyle, "ISO"))
set_config_option("DateStyle", OldDateStyle, PGC_USERSET, PGC_S_SESSION, true, true);
/*
* Terminate and done
*/
*cp++ = ')';
*cp = '\0';
SET_VARSIZE(cs->cmddata_buf,
VARHDRSZ + (cp - VARDATA(cs->cmddata_buf)));
}
else if (TRIGGER_FIRED_BY_UPDATE(tg->tg_event))
{
HeapTuple old_row = tg->tg_trigtuple;
HeapTuple new_row = tg->tg_newtuple;
TupleDesc tupdesc = tg->tg_relation->rd_att;
Datum old_value;
Datum new_value;
bool old_isnull;
bool new_isnull;
char *col_ident;
char *col_value;
int len_ident;
int len_value;
int i;
int need_comma = false;
int need_and = false;
char *OldDateStyle;
char *cp = VARDATA(cs->cmddata_buf);
/*
* UPDATE
*
* cmdtype = 'U' cmddata = "col_ident"='value' [, ...] where
* "pk_ident" = 'value' [ and ...]
*/
cmdtype = cs->cmdtype_U;
for (i = 0; i < tg->tg_relation->rd_att->natts; i++)
{
/*
* Ignore dropped columns
*/
if (tupdesc->attrs[i]->attisdropped)
continue;
old_value = SPI_getbinval(old_row, tupdesc, i + 1, &old_isnull);
new_value = SPI_getbinval(new_row, tupdesc, i + 1, &new_isnull);
/*
* If old and new value are NULL, the column is unchanged
*/
if (old_isnull && new_isnull)
continue;
/*
* If both are NOT NULL, we need to compare the values and skip
* setting the column if equal
*/
if (!old_isnull && !new_isnull)
{
Oid opr_oid;
FmgrInfo *opr_finfo_p;
/*
* Lookup the equal operators function call info using the
* typecache if available
*/
#ifdef HAVE_TYPCACHE
TypeCacheEntry *type_cache;
type_cache = lookup_type_cache(
SPI_gettypeid(tupdesc, i + 1),
TYPECACHE_EQ_OPR | TYPECACHE_EQ_OPR_FINFO);
opr_oid = type_cache->eq_opr;
if (opr_oid == ARRAY_EQ_OP)
opr_oid = InvalidOid;
else
opr_finfo_p = &(type_cache->eq_opr_finfo);
#else
FmgrInfo opr_finfo;
opr_oid = compatible_oper_funcid(makeList1(makeString("=")),
SPI_gettypeid(tupdesc, i + 1),
SPI_gettypeid(tupdesc, i + 1), true);
if (OidIsValid(opr_oid))
{
fmgr_info(opr_oid, &opr_finfo);
opr_finfo_p = &opr_finfo;
}
#endif
/*
* If we have an equal operator, use that to do binary
* comparision. Else get the string representation of both
* attributes and do string comparision.
*/
if (OidIsValid(opr_oid))
{
if (DatumGetBool(FunctionCall2(opr_finfo_p,
old_value, new_value)))
continue;
}
else
{
char *old_strval = SPI_getvalue(old_row, tupdesc, i + 1);
char *new_strval = SPI_getvalue(new_row, tupdesc, i + 1);
if (strcmp(old_strval, new_strval) == 0)
continue;
}
}
if (need_comma)
*cp++ = ',';
else
need_comma = true;
col_ident = (char *) slon_quote_identifier(SPI_fname(tupdesc, i + 1));
if (new_isnull)
col_value = "NULL";
else
{
OldDateStyle = GetConfigOptionByName("DateStyle", NULL);
if (!strstr(OldDateStyle, "ISO"))
set_config_option("DateStyle", "ISO", PGC_USERSET, PGC_S_SESSION, true, true);
col_value = slon_quote_literal(SPI_getvalue(new_row, tupdesc, i + 1));
if (!strstr(OldDateStyle, "ISO"))
set_config_option("DateStyle", OldDateStyle, PGC_USERSET, PGC_S_SESSION, true, true);
}
cmddata_need = (cp - (char *) (cs->cmddata_buf)) + 16 +
(len_ident = strlen(col_ident)) +
(len_value = strlen(col_value));
if (cs->cmddata_size < cmddata_need)
{
int have = (cp - (char *) (cs->cmddata_buf));
while (cs->cmddata_size < cmddata_need)
cs->cmddata_size *= 2;
cs->cmddata_buf = realloc(cs->cmddata_buf, cs->cmddata_size);
cp = (char *) (cs->cmddata_buf) + have;
}
memcpy(cp, col_ident, len_ident);
cp += len_ident;
*cp++ = '=';
memcpy(cp, col_value, len_value);
cp += len_value;
}
/*
* It can happen that the only UPDATE an application does is to set a
* column to the same value again. In that case, we'd end up here with
* no columns in the SET clause yet. We add the first key column here
* with it's old value to simulate the same for the replication
* engine.
*/
if (!need_comma)
{
for (i = 0, attkind_idx = -1; i < tg->tg_relation->rd_att->natts; i++)
{
if (tupdesc->attrs[i]->attisdropped)
continue;
attkind_idx++;
if (!attkind[attkind_idx])
elog(ERROR, "Slony-I: no key columns found in logTrigger() attkind parameter");
if (attkind[attkind_idx] == 'k')
break;
}
col_ident = (char *) slon_quote_identifier(SPI_fname(tupdesc, i + 1));
col_value = slon_quote_literal(SPI_getvalue(old_row, tupdesc, i + 1));
cmddata_need = (cp - (char *) (cs->cmddata_buf)) + 16 +
(len_ident = strlen(col_ident)) +
(len_value = strlen(col_value));
if (cs->cmddata_size < cmddata_need)
{
int have = (cp - (char *) (cs->cmddata_buf));
while (cs->cmddata_size < cmddata_need)
cs->cmddata_size *= 2;
cs->cmddata_buf = realloc(cs->cmddata_buf, cs->cmddata_size);
cp = (char *) (cs->cmddata_buf) + have;
}
memcpy(cp, col_ident, len_ident);
cp += len_ident;
*cp++ = '=';
memcpy(cp, col_value, len_value);
cp += len_value;
}
*cp++ = ' ';
*cp++ = 'w';
*cp++ = 'h';
*cp++ = 'e';
*cp++ = 'r';
*cp++ = 'e';
*cp++ = ' ';
for (i = 0, attkind_idx = -1; i < tg->tg_relation->rd_att->natts; i++)
{
/*
* Ignore dropped columns
*/
if (tupdesc->attrs[i]->attisdropped)
continue;
attkind_idx++;
if (!attkind[attkind_idx])
break;
if (attkind[attkind_idx] != 'k')
continue;
col_ident = (char *) slon_quote_identifier(SPI_fname(tupdesc, i + 1));
col_value = slon_quote_literal(SPI_getvalue(old_row, tupdesc, i + 1));
if (col_value == NULL)
elog(ERROR, "Slony-I: old key column %s.%s IS NULL on UPDATE",
NameStr(tg->tg_relation->rd_rel->relname), col_ident);
cmddata_need = (cp - (char *) (cs->cmddata_buf)) + 16 +
(len_ident = strlen(col_ident)) +
(len_value = strlen(col_value));
if (cs->cmddata_size < cmddata_need)
{
int have = (cp - (char *) (cs->cmddata_buf));
while (cs->cmddata_size < cmddata_need)
cs->cmddata_size *= 2;
cs->cmddata_buf = realloc(cs->cmddata_buf, cs->cmddata_size);
cp = (char *) (cs->cmddata_buf) + have;
}
if (need_and)
{
*cp++ = ' ';
*cp++ = 'a';
*cp++ = 'n';
*cp++ = 'd';
*cp++ = ' ';
}
else
need_and = true;
memcpy(cp, col_ident, len_ident);
cp += len_ident;
*cp++ = '=';
memcpy(cp, col_value, len_value);
cp += len_value;
}
*cp = '\0';
SET_VARSIZE(cs->cmddata_buf,
VARHDRSZ + (cp - VARDATA(cs->cmddata_buf)));
}
else if (TRIGGER_FIRED_BY_DELETE(tg->tg_event))
{
HeapTuple old_row = tg->tg_trigtuple;
TupleDesc tupdesc = tg->tg_relation->rd_att;
char *col_ident;
char *col_value;
int len_ident;
int len_value;
int i;
int need_and = false;
char *cp = VARDATA(cs->cmddata_buf);
/*
* DELETE
*
* cmdtype = 'D' cmddata = "pk_ident"='value' [and ...]
*/
cmdtype = cs->cmdtype_D;
for (i = 0, attkind_idx = -1; i < tg->tg_relation->rd_att->natts; i++)
{
if (tupdesc->attrs[i]->attisdropped)
continue;
attkind_idx++;
if (!attkind[attkind_idx])
break;
if (attkind[attkind_idx] != 'k')
continue;
col_ident = (char *) slon_quote_identifier(SPI_fname(tupdesc, i + 1));
col_value = slon_quote_literal(SPI_getvalue(old_row, tupdesc, i + 1));
if (col_value == NULL)
elog(ERROR, "Slony-I: old key column %s.%s IS NULL on DELETE",
NameStr(tg->tg_relation->rd_rel->relname), col_ident);
cmddata_need = (cp - (char *) (cs->cmddata_buf)) + 16 +
(len_ident = strlen(col_ident)) +
(len_value = strlen(col_value));
if (cs->cmddata_size < cmddata_need)
{
int have = (cp - (char *) (cs->cmddata_buf));
while (cs->cmddata_size < cmddata_need)
cs->cmddata_size *= 2;
cs->cmddata_buf = realloc(cs->cmddata_buf, cs->cmddata_size);
cp = (char *) (cs->cmddata_buf) + have;
}
if (need_and)
{
*cp++ = ' ';
*cp++ = 'a';
*cp++ = 'n';
*cp++ = 'd';
*cp++ = ' ';
}
else
need_and = true;
memcpy(cp, col_ident, len_ident);
cp += len_ident;
*cp++ = '=';
memcpy(cp, col_value, len_value);
cp += len_value;
}
*cp = '\0';
SET_VARSIZE(cs->cmddata_buf,
VARHDRSZ + (cp - VARDATA(cs->cmddata_buf)));
}
else
elog(ERROR, "Slony-I: logTrigger() fired for unhandled event");
/*
* Construct the parameter array and insert the log row.
*/
argv[0] = Int32GetDatum(tab_id);
argv[1] = PointerGetDatum(cmdtype);
argv[2] = PointerGetDatum(cs->cmddata_buf);
SPI_execp(cs->plan_active_log, argv, NULL, 0);
SPI_finish();
return PointerGetDatum(NULL);
}
/*
* LockGXact
* Locate the prepared transaction and mark it busy for COMMIT or PREPARE.
*/
static GlobalTransaction
LockGXact(const char *gid, Oid user)
{
int i;
LWLockAcquire(TwoPhaseStateLock, LW_EXCLUSIVE);
for (i = 0; i < TwoPhaseState->numPrepXacts; i++)
{
GlobalTransaction gxact = TwoPhaseState->prepXacts[i];
/* Ignore not-yet-valid GIDs */
if (!gxact->valid)
continue;
if (strcmp(gxact->gid, gid) != 0)
continue;
/* Found it, but has someone else got it locked? */
if (TransactionIdIsValid(gxact->locking_xid))
{
if (TransactionIdIsActive(gxact->locking_xid))
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("prepared transaction with identifier \"%s\" is busy",
gid)));
gxact->locking_xid = InvalidTransactionId;
}
if (user != gxact->owner && !superuser_arg(user))
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("permission denied to finish prepared transaction"),
errhint("Must be superuser or the user that prepared the transaction.")));
/*
* Note: it probably would be possible to allow committing from
* another database; but at the moment NOTIFY is known not to work and
* there may be some other issues as well. Hence disallow until
* someone gets motivated to make it work.
*/
if (MyDatabaseId != gxact->proc.databaseId)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("prepared transaction belongs to another database"),
errhint("Connect to the database where the transaction was prepared to finish it.")));
/* OK for me to lock it */
gxact->locking_xid = GetTopTransactionId();
LWLockRelease(TwoPhaseStateLock);
return gxact;
}
LWLockRelease(TwoPhaseStateLock);
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("prepared transaction with identifier \"%s\" does not exist",
gid)));
/* NOTREACHED */
return 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;
}
/*
* 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,
/* The filespace where the filespace lives. */
int16 primaryDbId,
char *primaryFilespaceLocation,
/*
* The primary filespace directory path. NOT Blank padded.
* Just a NULL terminated string.
*/
int16 mirrorDbId,
char *mirrorFilespaceLocation,
MirroredObjectExistenceState mirrorExistenceState,
ItemPointer persistentTid,
/* TID of the gp_persistent_rel_files tuple for the rel file */
int64 *persistentSerialNum,
bool flushToXLog)
/* When true, the XLOG record for this change will be flushed to disk. */
{
WRITE_PERSISTENT_STATE_ORDERED_LOCK_DECLARE;
PersistentFileSysObjName fsObjName;
FilespaceDirEntry filespaceDirEntry;
TransactionId topXid;
Datum values[Natts_gp_persistent_filespace_node];
char mirrorFilespaceLocationBlankPadded[FilespaceLocationBlankPaddedWithNullTermLen];
char primaryFilespaceLocationBlankPadded[FilespaceLocationBlankPaddedWithNullTermLen];
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);
topXid = GetTopTransactionId();
WRITE_PERSISTENT_STATE_ORDERED_LOCK;
PersistentFilespace_BlankPadCopyLocation(
primaryFilespaceLocationBlankPadded,
primaryFilespaceLocation);
PersistentFilespace_BlankPadCopyLocation(
mirrorFilespaceLocationBlankPadded,
mirrorFilespaceLocation);
GpPersistentFilespaceNode_SetDatumValues(
values,
filespaceOid,
primaryDbId,
primaryFilespaceLocationBlankPadded,
mirrorDbId,
mirrorFilespaceLocationBlankPadded,
PersistentFileSysState_CreatePending,
/* createMirrorDataLossTrackingSessionNum */ 0,
mirrorExistenceState,
/* reserved */ 0,
/* parentXid */ topXid,
/* persistentSerialNum */ 0); // This will be set by PersistentFileSysObj_AddTuple.
PersistentFileSysObj_AddTuple(
PersistentFsObjType_FilespaceDir,
values,
flushToXLog,
persistentTid,
persistentSerialNum);
WRITE_FILESPACE_HASH_LOCK;
filespaceDirEntry = PersistentFilespace_CreateDirUnderLock(filespaceOid);
Assert(filespaceDirEntry != NULL);
filespaceDirEntry->dbId1 = primaryDbId;
memcpy(filespaceDirEntry->locationBlankPadded1, primaryFilespaceLocationBlankPadded,
FilespaceLocationBlankPaddedWithNullTermLen);
filespaceDirEntry->dbId2 = mirrorDbId;
memcpy(filespaceDirEntry->locationBlankPadded2, mirrorFilespaceLocationBlankPadded,
FilespaceLocationBlankPaddedWithNullTermLen);
filespaceDirEntry->state = PersistentFileSysState_CreatePending;
ItemPointerCopy(persistentTid, &filespaceDirEntry->persistentTid);
filespaceDirEntry->persistentSerialNum = *persistentSerialNum;
WRITE_FILESPACE_HASH_UNLOCK;
/*
* This XLOG must be generated under the persistent write-lock.
*/
#ifdef MASTER_MIRROR_SYNC
mmxlog_log_create_filespace(filespaceOid);
#endif
SIMPLE_FAULT_INJECTOR(FaultBeforePendingDeleteFilespaceEntry);
/*
* 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', mirror existence state '%s', serial number " INT64_FORMAT " at TID %s",
PersistentFileSysObjName_ObjectName(&fsObjName),
MirroredObjectExistenceState_Name(mirrorExistenceState),
*persistentSerialNum,
ItemPointerToString(persistentTid));
}
/*
* Initialize a sequence's relation with the specified tuple as content
*/
static void
fill_seq_with_data(Relation rel, HeapTuple tuple)
{
Buffer buf;
Page page;
sequence_magic *sm;
OffsetNumber offnum;
/* Initialize first page of relation with special magic number */
buf = ReadBuffer(rel, P_NEW);
Assert(BufferGetBlockNumber(buf) == 0);
page = BufferGetPage(buf);
PageInit(page, BufferGetPageSize(buf), sizeof(sequence_magic));
sm = (sequence_magic *) PageGetSpecialPointer(page);
sm->magic = SEQ_MAGIC;
/* Now insert sequence tuple */
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
/*
* Since VACUUM does not process sequences, we have to force the tuple to
* have xmin = FrozenTransactionId now. Otherwise it would become
* invisible to SELECTs after 2G transactions. It is okay to do this
* because if the current transaction aborts, no other xact will ever
* examine the sequence tuple anyway.
*/
HeapTupleHeaderSetXmin(tuple->t_data, FrozenTransactionId);
HeapTupleHeaderSetXminFrozen(tuple->t_data);
HeapTupleHeaderSetCmin(tuple->t_data, FirstCommandId);
HeapTupleHeaderSetXmax(tuple->t_data, InvalidTransactionId);
tuple->t_data->t_infomask |= HEAP_XMAX_INVALID;
ItemPointerSet(&tuple->t_data->t_ctid, 0, FirstOffsetNumber);
/* check the comment above nextval_internal()'s equivalent call. */
if (RelationNeedsWAL(rel))
GetTopTransactionId();
START_CRIT_SECTION();
MarkBufferDirty(buf);
offnum = PageAddItem(page, (Item) tuple->t_data, tuple->t_len,
InvalidOffsetNumber, false, false);
if (offnum != FirstOffsetNumber)
elog(ERROR, "failed to add sequence tuple to page");
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
xl_seq_rec xlrec;
XLogRecPtr recptr;
XLogBeginInsert();
XLogRegisterBuffer(0, buf, REGBUF_WILL_INIT);
xlrec.node = rel->rd_node;
XLogRegisterData((char *) &xlrec, sizeof(xl_seq_rec));
XLogRegisterData((char *) tuple->t_data, tuple->t_len);
recptr = XLogInsert(RM_SEQ_ID, XLOG_SEQ_LOG);
PageSetLSN(page, recptr);
}
END_CRIT_SECTION();
UnlockReleaseBuffer(buf);
}
