/*-------------------------------------------------------------------------
*
* varsup.c
- * postgres variable relation support routines
- *
- * Portions Copyright (c) 1996-2000, PostgreSQL, Inc
- * Portions Copyright (c) 1994, Regents of the University of California
+ * postgres OID & XID variables support routines
*
+ * Copyright (c) 2000-2019, PostgreSQL Global Development Group
*
* IDENTIFICATION
- * $Header: /cvsroot/pgsql/src/backend/access/transam/varsup.c,v 1.33 2000/11/20 16:47:30 petere Exp $
+ * src/backend/access/transam/varsup.c
*
*-------------------------------------------------------------------------
*/
-#include "postgres.h"
-#ifdef XLOG
-#include "xlog_varsup.c"
-#else
+#include "postgres.h"
-#include "access/heapam.h"
-#include "catalog/catname.h"
+#include "access/clog.h"
+#include "access/commit_ts.h"
+#include "access/subtrans.h"
+#include "access/transam.h"
+#include "access/xact.h"
+#include "access/xlog.h"
+#include "commands/dbcommands.h"
+#include "miscadmin.h"
+#include "postmaster/autovacuum.h"
+#include "storage/pmsignal.h"
#include "storage/proc.h"
+#include "utils/syscache.h"
-static void GetNewObjectIdBlock(Oid *oid_return, int oid_block_size);
-static void VariableRelationGetNextOid(Oid *oid_return);
-static void VariableRelationGetNextXid(TransactionId *xidP);
-static void VariableRelationPutNextOid(Oid oid);
-/* ---------------------
- * spin lock for oid generation
- * ---------------------
- */
-int OidGenLockId;
+/* Number of OIDs to prefetch (preallocate) per XLOG write */
+#define VAR_OID_PREFETCH 8192
-/* ---------------------
- * pointer to "variable cache" in shared memory (set up by shmem.c)
- * ---------------------
- */
+/* pointer to "variable cache" in shared memory (set up by shmem.c) */
VariableCache ShmemVariableCache = NULL;
-/* ----------------------------------------------------------------
- * variable relation query/update routines
- * ----------------------------------------------------------------
- */
-
-/* --------------------------------
- * VariableRelationGetNextXid
- * --------------------------------
+/*
+ * Allocate the next FullTransactionId for a new transaction or
+ * subtransaction.
+ *
+ * The new XID is also stored into MyPgXact before returning.
+ *
+ * Note: when this is called, we are actually already inside a valid
+ * transaction, since XIDs are now not allocated until the transaction
+ * does something. So it is safe to do a database lookup if we want to
+ * issue a warning about XID wrap.
*/
-static void
-VariableRelationGetNextXid(TransactionId *xidP)
+FullTransactionId
+GetNewTransactionId(bool isSubXact)
{
- Buffer buf;
- VariableRelationContents var;
+ FullTransactionId full_xid;
+ TransactionId xid;
- /* ----------------
- * We assume that a spinlock has been acquired to guarantee
- * exclusive access to the variable relation.
- * ----------------
+ /*
+ * Workers synchronize transaction state at the beginning of each parallel
+ * operation, so we can't account for new XIDs after that point.
*/
+ if (IsInParallelMode())
+ elog(ERROR, "cannot assign TransactionIds during a parallel operation");
- /* ----------------
- * do nothing before things are initialized
- * ----------------
- */
- if (!RelationIsValid(VariableRelation))
- return;
-
- /* ----------------
- * read the variable page, get the the nextXid field and
- * release the buffer
- * ----------------
+ /*
+ * During bootstrap initialization, we return the special bootstrap
+ * transaction id.
*/
- buf = ReadBuffer(VariableRelation, 0);
-
- if (!BufferIsValid(buf))
+ if (IsBootstrapProcessingMode())
{
- SpinRelease(OidGenLockId);
- elog(ERROR, "VariableRelationGetNextXid: ReadBuffer failed");
+ Assert(!isSubXact);
+ MyPgXact->xid = BootstrapTransactionId;
+ return FullTransactionIdFromEpochAndXid(0, BootstrapTransactionId);
}
- var = (VariableRelationContents) BufferGetBlock(buf);
-
- TransactionIdStore(var->nextXidData, xidP);
-
- ReleaseBuffer(buf);
-}
-
-/* --------------------------------
- * VariableRelationPutNextXid
- * --------------------------------
- */
-void
-VariableRelationPutNextXid(TransactionId xid)
-{
- Buffer buf;
- VariableRelationContents var;
+ /* safety check, we should never get this far in a HS standby */
+ if (RecoveryInProgress())
+ elog(ERROR, "cannot assign TransactionIds during recovery");
- /* ----------------
- * We assume that a spinlock has been acquired to guarantee
- * exclusive access to the variable relation.
- * ----------------
- */
+ LWLockAcquire(XidGenLock, LW_EXCLUSIVE);
- /* ----------------
- * do nothing before things are initialized
- * ----------------
- */
- if (!RelationIsValid(VariableRelation))
- return;
+ full_xid = ShmemVariableCache->nextFullXid;
+ xid = XidFromFullTransactionId(full_xid);
- /* ----------------
- * read the variable page, update the nextXid field and
- * write the page back out to disk (with immediate write).
- * ----------------
+ /*----------
+ * Check to see if it's safe to assign another XID. This protects against
+ * catastrophic data loss due to XID wraparound. The basic rules are:
+ *
+ * If we're past xidVacLimit, start trying to force autovacuum cycles.
+ * If we're past xidWarnLimit, start issuing warnings.
+ * If we're past xidStopLimit, refuse to execute transactions, unless
+ * we are running in single-user mode (which gives an escape hatch
+ * to the DBA who somehow got past the earlier defenses).
+ *
+ * Note that this coding also appears in GetNewMultiXactId.
+ *----------
*/
- buf = ReadBuffer(VariableRelation, 0);
-
- if (!BufferIsValid(buf))
+ if (TransactionIdFollowsOrEquals(xid, ShmemVariableCache->xidVacLimit))
{
- SpinRelease(OidGenLockId);
- elog(ERROR, "VariableRelationPutNextXid: ReadBuffer failed");
- }
-
- var = (VariableRelationContents) BufferGetBlock(buf);
+ /*
+ * For safety's sake, we release XidGenLock while sending signals,
+ * warnings, etc. This is not so much because we care about
+ * preserving concurrency in this situation, as to avoid any
+ * possibility of deadlock while doing get_database_name(). First,
+ * copy all the shared values we'll need in this path.
+ */
+ TransactionId xidWarnLimit = ShmemVariableCache->xidWarnLimit;
+ TransactionId xidStopLimit = ShmemVariableCache->xidStopLimit;
+ TransactionId xidWrapLimit = ShmemVariableCache->xidWrapLimit;
+ Oid oldest_datoid = ShmemVariableCache->oldestXidDB;
- TransactionIdStore(xid, &(var->nextXidData));
+ LWLockRelease(XidGenLock);
- FlushBuffer(buf, true, true);
-}
+ /*
+ * To avoid swamping the postmaster with signals, we issue the autovac
+ * request only once per 64K transaction starts. This still gives
+ * plenty of chances before we get into real trouble.
+ */
+ if (IsUnderPostmaster && (xid % 65536) == 0)
+ SendPostmasterSignal(PMSIGNAL_START_AUTOVAC_LAUNCHER);
+
+ if (IsUnderPostmaster &&
+ TransactionIdFollowsOrEquals(xid, xidStopLimit))
+ {
+ char *oldest_datname = get_database_name(oldest_datoid);
+
+ /* complain even if that DB has disappeared */
+ if (oldest_datname)
+ ereport(ERROR,
+ (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
+ errmsg("database is not accepting commands to avoid wraparound data loss in database \"%s\"",
+ oldest_datname),
+ errhint("Stop the postmaster and vacuum that database in single-user mode.\n"
+ "You might also need to commit or roll back old prepared transactions, or drop stale replication slots.")));
+ else
+ ereport(ERROR,
+ (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
+ errmsg("database is not accepting commands to avoid wraparound data loss in database with OID %u",
+ oldest_datoid),
+ errhint("Stop the postmaster and vacuum that database in single-user mode.\n"
+ "You might also need to commit or roll back old prepared transactions, or drop stale replication slots.")));
+ }
+ else if (TransactionIdFollowsOrEquals(xid, xidWarnLimit))
+ {
+ char *oldest_datname = get_database_name(oldest_datoid);
+
+ /* complain even if that DB has disappeared */
+ if (oldest_datname)
+ ereport(WARNING,
+ (errmsg("database \"%s\" must be vacuumed within %u transactions",
+ oldest_datname,
+ xidWrapLimit - xid),
+ errhint("To avoid a database shutdown, execute a database-wide VACUUM in that database.\n"
+ "You might also need to commit or roll back old prepared transactions, or drop stale replication slots.")));
+ else
+ ereport(WARNING,
+ (errmsg("database with OID %u must be vacuumed within %u transactions",
+ oldest_datoid,
+ xidWrapLimit - xid),
+ errhint("To avoid a database shutdown, execute a database-wide VACUUM in that database.\n"
+ "You might also need to commit or roll back old prepared transactions, or drop stale replication slots.")));
+ }
+
+ /* Re-acquire lock and start over */
+ LWLockAcquire(XidGenLock, LW_EXCLUSIVE);
+ full_xid = ShmemVariableCache->nextFullXid;
+ xid = XidFromFullTransactionId(full_xid);
+ }
-/* --------------------------------
- * VariableRelationGetNextOid
- * --------------------------------
- */
-static void
-VariableRelationGetNextOid(Oid *oid_return)
-{
- Buffer buf;
- VariableRelationContents var;
+ /*
+ * If we are allocating the first XID of a new page of the commit log,
+ * zero out that commit-log page before returning. We must do this while
+ * holding XidGenLock, else another xact could acquire and commit a later
+ * XID before we zero the page. Fortunately, a page of the commit log
+ * holds 32K or more transactions, so we don't have to do this very often.
+ *
+ * Extend pg_subtrans and pg_commit_ts too.
+ */
+ ExtendCLOG(xid);
+ ExtendCommitTs(xid);
+ ExtendSUBTRANS(xid);
- /* ----------------
- * We assume that a spinlock has been acquired to guarantee
- * exclusive access to the variable relation.
- * ----------------
+ /*
+ * Now advance the nextFullXid counter. This must not happen until after
+ * we have successfully completed ExtendCLOG() --- if that routine fails,
+ * we want the next incoming transaction to try it again. We cannot
+ * assign more XIDs until there is CLOG space for them.
*/
+ FullTransactionIdAdvance(&ShmemVariableCache->nextFullXid);
- /* ----------------
- * if the variable relation is not initialized, then we
- * assume we are running at bootstrap time and so we return
- * an invalid object id (this path should never be taken, probably).
- * ----------------
+ /*
+ * We must store the new XID into the shared ProcArray before releasing
+ * XidGenLock. This ensures that every active XID older than
+ * latestCompletedXid is present in the ProcArray, which is essential for
+ * correct OldestXmin tracking; see src/backend/access/transam/README.
+ *
+ * Note that readers of PGXACT xid fields should be careful to fetch the
+ * value only once, rather than assume they can read a value multiple
+ * times and get the same answer each time. Note we are assuming that
+ * TransactionId and int fetch/store are atomic.
+ *
+ * The same comments apply to the subxact xid count and overflow fields.
+ *
+ * Use of a write barrier prevents dangerous code rearrangement in this
+ * function; other backends could otherwise e.g. be examining my subxids
+ * info concurrently, and we don't want them to see an invalid
+ * intermediate state, such as an incremented nxids before the array entry
+ * is filled.
+ *
+ * Other processes that read nxids should do so before reading xids
+ * elements with a pg_read_barrier() in between, so that they can be sure
+ * not to read an uninitialized array element; see
+ * src/backend/storage/lmgr/README.barrier.
+ *
+ * If there's no room to fit a subtransaction XID into PGPROC, set the
+ * cache-overflowed flag instead. This forces readers to look in
+ * pg_subtrans to map subtransaction XIDs up to top-level XIDs. There is a
+ * race-condition window, in that the new XID will not appear as running
+ * until its parent link has been placed into pg_subtrans. However, that
+ * will happen before anyone could possibly have a reason to inquire about
+ * the status of the XID, so it seems OK. (Snapshots taken during this
+ * window *will* include the parent XID, so they will deliver the correct
+ * answer later on when someone does have a reason to inquire.)
*/
- if (!RelationIsValid(VariableRelation))
+ if (!isSubXact)
+ MyPgXact->xid = xid; /* LWLockRelease acts as barrier */
+ else
{
- (*oid_return) = InvalidOid;
- return;
+ int nxids = MyPgXact->nxids;
+
+ if (nxids < PGPROC_MAX_CACHED_SUBXIDS)
+ {
+ MyProc->subxids.xids[nxids] = xid;
+ pg_write_barrier();
+ MyPgXact->nxids = nxids + 1;
+ }
+ else
+ MyPgXact->overflowed = true;
}
- /* ----------------
- * read the variable page, get the the nextOid field and
- * release the buffer
- * ----------------
- */
- buf = ReadBuffer(VariableRelation, 0);
+ LWLockRelease(XidGenLock);
- if (!BufferIsValid(buf))
- {
- SpinRelease(OidGenLockId);
- elog(ERROR, "VariableRelationGetNextOid: ReadBuffer failed");
- }
+ return full_xid;
+}
- var = (VariableRelationContents) BufferGetBlock(buf);
+/*
+ * Read nextFullXid but don't allocate it.
+ */
+FullTransactionId
+ReadNextFullTransactionId(void)
+{
+ FullTransactionId fullXid;
- (*oid_return) = var->nextOid;
+ LWLockAcquire(XidGenLock, LW_SHARED);
+ fullXid = ShmemVariableCache->nextFullXid;
+ LWLockRelease(XidGenLock);
- ReleaseBuffer(buf);
+ return fullXid;
}
-/* --------------------------------
- * VariableRelationPutNextOid
- * --------------------------------
+/*
+ * Advance nextFullXid to the value after a given xid. The epoch is inferred.
+ * This must only be called during recovery or from two-phase start-up code.
*/
-static void
-VariableRelationPutNextOid(Oid oid)
+void
+AdvanceNextFullTransactionIdPastXid(TransactionId xid)
{
- Buffer buf;
- VariableRelationContents var;
+ FullTransactionId newNextFullXid;
+ TransactionId next_xid;
+ uint32 epoch;
- /* ----------------
- * We assume that a spinlock has been acquired to guarantee
- * exclusive access to the variable relation.
- * ----------------
+ /*
+ * It is safe to read nextFullXid without a lock, because this is only
+ * called from the startup process or single-process mode, meaning that no
+ * other process can modify it.
*/
+ Assert(AmStartupProcess() || !IsUnderPostmaster);
- /* ----------------
- * do nothing before things are initialized
- * ----------------
- */
- if (!RelationIsValid(VariableRelation))
+ /* Fast return if this isn't an xid high enough to move the needle. */
+ next_xid = XidFromFullTransactionId(ShmemVariableCache->nextFullXid);
+ if (!TransactionIdFollowsOrEquals(xid, next_xid))
return;
- /* ----------------
- * read the variable page, update the nextXid field and
- * write the page back out to disk.
- * ----------------
+ /*
+ * Compute the FullTransactionId that comes after the given xid. To do
+ * this, we preserve the existing epoch, but detect when we've wrapped
+ * into a new epoch. This is necessary because WAL records and 2PC state
+ * currently contain 32 bit xids. The wrap logic is safe in those cases
+ * because the span of active xids cannot exceed one epoch at any given
+ * point in the WAL stream.
*/
- buf = ReadBuffer(VariableRelation, 0);
-
- if (!BufferIsValid(buf))
- {
- SpinRelease(OidGenLockId);
- elog(ERROR, "VariableRelationPutNextOid: ReadBuffer failed");
- }
-
- var = (VariableRelationContents) BufferGetBlock(buf);
-
- var->nextOid = oid;
+ TransactionIdAdvance(xid);
+ epoch = EpochFromFullTransactionId(ShmemVariableCache->nextFullXid);
+ if (unlikely(xid < next_xid))
+ ++epoch;
+ newNextFullXid = FullTransactionIdFromEpochAndXid(epoch, xid);
- WriteBuffer(buf);
+ /*
+ * We still need to take a lock to modify the value when there are
+ * concurrent readers.
+ */
+ LWLockAcquire(XidGenLock, LW_EXCLUSIVE);
+ ShmemVariableCache->nextFullXid = newNextFullXid;
+ LWLockRelease(XidGenLock);
}
-/* ----------------------------------------------------------------
- * transaction id generation support
- * ----------------------------------------------------------------
- */
-
-/* ----------------
- * GetNewTransactionId
- *
- * Transaction IDs are allocated via a cache in shared memory.
- * Each time we need more IDs, we advance the "next XID" value
- * in pg_variable by VAR_XID_PREFETCH and set the cache to
- * show that many XIDs as available. Then, allocating those XIDs
- * requires just a spinlock and not a buffer read/write cycle.
- *
- * Since the cache is shared across all backends, cached but unused
- * XIDs are not lost when a backend exits, only when the postmaster
- * quits or forces shared memory reinit. So we can afford to have
- * a pretty big value of VAR_XID_PREFETCH.
+/*
+ * Advance the cluster-wide value for the oldest valid clog entry.
*
- * This code does not worry about initializing the transaction counter
- * (see transam.c's InitializeTransactionLog() for that). We also
- * ignore the possibility that the counter could someday wrap around.
- * ----------------
+ * We must acquire CLogTruncationLock to advance the oldestClogXid. It's not
+ * necessary to hold the lock during the actual clog truncation, only when we
+ * advance the limit, as code looking up arbitrary xids is required to hold
+ * CLogTruncationLock from when it tests oldestClogXid through to when it
+ * completes the clog lookup.
*/
-
-#define VAR_XID_PREFETCH 1024
-
void
-GetNewTransactionId(TransactionId *xid)
+AdvanceOldestClogXid(TransactionId oldest_datfrozenxid)
{
-
- /* ----------------
- * during bootstrap initialization, we return the special
- * bootstrap transaction id.
- * ----------------
- */
- if (AMI_OVERRIDE)
+ LWLockAcquire(CLogTruncationLock, LW_EXCLUSIVE);
+ if (TransactionIdPrecedes(ShmemVariableCache->oldestClogXid,
+ oldest_datfrozenxid))
{
- TransactionIdStore(AmiTransactionId, xid);
- return;
- }
-
- SpinAcquire(OidGenLockId); /* not good for concurrency... */
-
- if (ShmemVariableCache->xid_count == 0)
- {
- TransactionId nextid;
-
- VariableRelationGetNextXid(&nextid);
- TransactionIdStore(nextid, &(ShmemVariableCache->nextXid));
- ShmemVariableCache->xid_count = VAR_XID_PREFETCH;
- TransactionIdAdd(&nextid, VAR_XID_PREFETCH);
- VariableRelationPutNextXid(nextid);
+ ShmemVariableCache->oldestClogXid = oldest_datfrozenxid;
}
-
- TransactionIdStore(ShmemVariableCache->nextXid, xid);
- TransactionIdAdd(&(ShmemVariableCache->nextXid), 1);
- (ShmemVariableCache->xid_count)--;
-
- if (MyProc != (PROC *) NULL)
- MyProc->xid = *xid;
-
- SpinRelease(OidGenLockId);
+ LWLockRelease(CLogTruncationLock);
}
/*
- * Like GetNewTransactionId reads nextXid but don't fetch it.
+ * Determine the last safe XID to allocate using the currently oldest
+ * datfrozenxid (ie, the oldest XID that might exist in any database
+ * of our cluster), and the OID of the (or a) database with that value.
*/
void
-ReadNewTransactionId(TransactionId *xid)
+SetTransactionIdLimit(TransactionId oldest_datfrozenxid, Oid oldest_datoid)
{
+ TransactionId xidVacLimit;
+ TransactionId xidWarnLimit;
+ TransactionId xidStopLimit;
+ TransactionId xidWrapLimit;
+ TransactionId curXid;
- /* ----------------
- * during bootstrap initialization, we return the special
- * bootstrap transaction id.
- * ----------------
- */
- if (AMI_OVERRIDE)
- {
- TransactionIdStore(AmiTransactionId, xid);
- return;
- }
-
- SpinAcquire(OidGenLockId); /* not good for concurrency... */
+ Assert(TransactionIdIsNormal(oldest_datfrozenxid));
/*
- * Note that we don't check is ShmemVariableCache->xid_count equal to
- * 0 or not. This will work as long as we don't call
- * ReadNewTransactionId() before GetNewTransactionId().
+ * The place where we actually get into deep trouble is halfway around
+ * from the oldest potentially-existing XID. (This calculation is
+ * probably off by one or two counts, because the special XIDs reduce the
+ * size of the loop a little bit. But we throw in plenty of slop below,
+ * so it doesn't matter.)
*/
- if (ShmemVariableCache->nextXid == 0)
- elog(ERROR, "ReadNewTransactionId: ShmemVariableCache->nextXid is not initialized");
-
- TransactionIdStore(ShmemVariableCache->nextXid, xid);
-
- SpinRelease(OidGenLockId);
-}
-
-/* ----------------------------------------------------------------
- * object id generation support
- * ----------------------------------------------------------------
- */
+ xidWrapLimit = oldest_datfrozenxid + (MaxTransactionId >> 1);
+ if (xidWrapLimit < FirstNormalTransactionId)
+ xidWrapLimit += FirstNormalTransactionId;
-/* ----------------
- * GetNewObjectIdBlock
- *
- * This support function is used to allocate a block of object ids
- * of the given size.
- * ----------------
- */
-static void
-GetNewObjectIdBlock(Oid *oid_return, /* place to return the first new
- * object id */
- int oid_block_size) /* number of oids desired */
-{
- Oid firstfreeoid;
- Oid nextoid;
-
- /* ----------------
- * Obtain exclusive access to the variable relation page
- * ----------------
+ /*
+ * We'll refuse to continue assigning XIDs in interactive mode once we get
+ * within 1M transactions of data loss. This leaves lots of room for the
+ * DBA to fool around fixing things in a standalone backend, while not
+ * being significant compared to total XID space. (Note that since
+ * vacuuming requires one transaction per table cleaned, we had better be
+ * sure there's lots of XIDs left...)
*/
- SpinAcquire(OidGenLockId);
+ xidStopLimit = xidWrapLimit - 1000000;
+ if (xidStopLimit < FirstNormalTransactionId)
+ xidStopLimit -= FirstNormalTransactionId;
- /* ----------------
- * get the "next" oid from the variable relation
- * ----------------
+ /*
+ * We'll start complaining loudly when we get within 10M transactions of
+ * the stop point. This is kind of arbitrary, but if you let your gas
+ * gauge get down to 1% of full, would you be looking for the next gas
+ * station? We need to be fairly liberal about this number because there
+ * are lots of scenarios where most transactions are done by automatic
+ * clients that won't pay attention to warnings. (No, we're not gonna make
+ * this configurable. If you know enough to configure it, you know enough
+ * to not get in this kind of trouble in the first place.)
*/
- VariableRelationGetNextOid(&firstfreeoid);
+ xidWarnLimit = xidStopLimit - 10000000;
+ if (xidWarnLimit < FirstNormalTransactionId)
+ xidWarnLimit -= FirstNormalTransactionId;
- /* ----------------
- * Allocate the range of OIDs to be returned to the caller.
- *
- * There are two things going on here.
- *
- * One: in a virgin database pg_variable will initially contain zeroes,
- * so we will read out firstfreeoid = InvalidOid. We want to start
- * allocating OIDs at BootstrapObjectIdData instead (OIDs below that
- * are reserved for static assignment in the initial catalog data).
+ /*
+ * We'll start trying to force autovacuums when oldest_datfrozenxid gets
+ * to be more than autovacuum_freeze_max_age transactions old.
*
- * Two: if a database is run long enough, the OID counter will wrap
- * around. We must not generate an invalid OID when that happens,
- * and it seems wise not to generate anything in the reserved range.
- * Therefore we advance to BootstrapObjectIdData in this case too.
+ * Note: guc.c ensures that autovacuum_freeze_max_age is in a sane range,
+ * so that xidVacLimit will be well before xidWarnLimit.
*
- * The comparison here assumes that Oid is an unsigned type.
+ * Note: autovacuum_freeze_max_age is a PGC_POSTMASTER parameter so that
+ * we don't have to worry about dealing with on-the-fly changes in its
+ * value. It doesn't look practical to update shared state from a GUC
+ * assign hook (too many processes would try to execute the hook,
+ * resulting in race conditions as well as crashes of those not connected
+ * to shared memory). Perhaps this can be improved someday. See also
+ * SetMultiXactIdLimit.
*/
- nextoid = firstfreeoid + oid_block_size;
-
- if (! OidIsValid(firstfreeoid) || nextoid < firstfreeoid)
- {
- /* Initialization or wraparound time, force it up to safe range */
- firstfreeoid = BootstrapObjectIdData;
- nextoid = firstfreeoid + oid_block_size;
- }
-
- (*oid_return) = firstfreeoid;
-
- /* ----------------
- * Update the variable relation to show the block range as used.
- * ----------------
- */
- VariableRelationPutNextOid(nextoid);
-
- /* ----------------
- * Relinquish our lock on the variable relation page
- * ----------------
- */
- SpinRelease(OidGenLockId);
-}
-
-/* ----------------
- * GetNewObjectId
- *
- * This function allocates and parses out object ids. Like
- * GetNewTransactionId(), it "prefetches" 32 object ids by
- * incrementing the nextOid stored in the var relation by 32 and then
- * returning these id's one at a time until they are exhausted.
- * This means we reduce the number of accesses to the variable
- * relation by 32 for each backend.
- *
- * Note: 32 has no special significance. We don't want the
- * number to be too large because when the backend
- * terminates, we lose the oids we cached.
- *
- * Question: couldn't we use a shared-memory cache just like XIDs?
- * That would allow a larger interval between pg_variable updates
- * without cache losses. Note, however, that we can assign an OID
- * without even a spinlock from the backend-local OID cache.
- * Maybe two levels of caching would be good.
- * ----------------
- */
-
-#define VAR_OID_PREFETCH 32
+ xidVacLimit = oldest_datfrozenxid + autovacuum_freeze_max_age;
+ if (xidVacLimit < FirstNormalTransactionId)
+ xidVacLimit += FirstNormalTransactionId;
+
+ /* Grab lock for just long enough to set the new limit values */
+ LWLockAcquire(XidGenLock, LW_EXCLUSIVE);
+ ShmemVariableCache->oldestXid = oldest_datfrozenxid;
+ ShmemVariableCache->xidVacLimit = xidVacLimit;
+ ShmemVariableCache->xidWarnLimit = xidWarnLimit;
+ ShmemVariableCache->xidStopLimit = xidStopLimit;
+ ShmemVariableCache->xidWrapLimit = xidWrapLimit;
+ ShmemVariableCache->oldestXidDB = oldest_datoid;
+ curXid = XidFromFullTransactionId(ShmemVariableCache->nextFullXid);
+ LWLockRelease(XidGenLock);
+
+ /* Log the info */
+ ereport(DEBUG1,
+ (errmsg("transaction ID wrap limit is %u, limited by database with OID %u",
+ xidWrapLimit, oldest_datoid)));
-static int prefetched_oid_count = 0;
-static Oid next_prefetched_oid;
-
-void
-GetNewObjectId(Oid *oid_return) /* place to return the new object id */
-{
- /* ----------------
- * if we run out of prefetched oids, then we get some
- * more before handing them out to the caller.
- * ----------------
+ /*
+ * If past the autovacuum force point, immediately signal an autovac
+ * request. The reason for this is that autovac only processes one
+ * database per invocation. Once it's finished cleaning up the oldest
+ * database, it'll call here, and we'll signal the postmaster to start
+ * another iteration immediately if there are still any old databases.
*/
+ if (TransactionIdFollowsOrEquals(curXid, xidVacLimit) &&
+ IsUnderPostmaster && !InRecovery)
+ SendPostmasterSignal(PMSIGNAL_START_AUTOVAC_LAUNCHER);
- if (prefetched_oid_count == 0)
+ /* Give an immediate warning if past the wrap warn point */
+ if (TransactionIdFollowsOrEquals(curXid, xidWarnLimit) && !InRecovery)
{
- int oid_block_size = VAR_OID_PREFETCH;
-
- /* ----------------
- * Make sure pg_variable is open.
- * ----------------
+ char *oldest_datname;
+
+ /*
+ * We can be called when not inside a transaction, for example during
+ * StartupXLOG(). In such a case we cannot do database access, so we
+ * must just report the oldest DB's OID.
+ *
+ * Note: it's also possible that get_database_name fails and returns
+ * NULL, for example because the database just got dropped. We'll
+ * still warn, even though the warning might now be unnecessary.
*/
- if (!RelationIsValid(VariableRelation))
- VariableRelation = heap_openr(VariableRelationName, NoLock);
-
- /* ----------------
- * get a new block of prefetched object ids.
- * ----------------
- */
- GetNewObjectIdBlock(&next_prefetched_oid, oid_block_size);
-
- /* ----------------
- * now reset the prefetched_oid_count.
- * ----------------
- */
- prefetched_oid_count = oid_block_size;
+ if (IsTransactionState())
+ oldest_datname = get_database_name(oldest_datoid);
+ else
+ oldest_datname = NULL;
+
+ if (oldest_datname)
+ ereport(WARNING,
+ (errmsg("database \"%s\" must be vacuumed within %u transactions",
+ oldest_datname,
+ xidWrapLimit - curXid),
+ errhint("To avoid a database shutdown, execute a database-wide VACUUM in that database.\n"
+ "You might also need to commit or roll back old prepared transactions, or drop stale replication slots.")));
+ else
+ ereport(WARNING,
+ (errmsg("database with OID %u must be vacuumed within %u transactions",
+ oldest_datoid,
+ xidWrapLimit - curXid),
+ errhint("To avoid a database shutdown, execute a database-wide VACUUM in that database.\n"
+ "You might also need to commit or roll back old prepared transactions, or drop stale replication slots.")));
}
+}
- /* ----------------
- * return the next prefetched oid in the pointer passed by
- * the user and decrement the prefetch count.
- * ----------------
- */
- if (PointerIsValid(oid_return))
- (*oid_return) = next_prefetched_oid;
- next_prefetched_oid++;
- prefetched_oid_count--;
+/*
+ * ForceTransactionIdLimitUpdate -- does the XID wrap-limit data need updating?
+ *
+ * We primarily check whether oldestXidDB is valid. The cases we have in
+ * mind are that that database was dropped, or the field was reset to zero
+ * by pg_resetwal. In either case we should force recalculation of the
+ * wrap limit. Also do it if oldestXid is old enough to be forcing
+ * autovacuums or other actions; this ensures we update our state as soon
+ * as possible once extra overhead is being incurred.
+ */
+bool
+ForceTransactionIdLimitUpdate(void)
+{
+ TransactionId nextXid;
+ TransactionId xidVacLimit;
+ TransactionId oldestXid;
+ Oid oldestXidDB;
+
+ /* Locking is probably not really necessary, but let's be careful */
+ LWLockAcquire(XidGenLock, LW_SHARED);
+ nextXid = XidFromFullTransactionId(ShmemVariableCache->nextFullXid);
+ xidVacLimit = ShmemVariableCache->xidVacLimit;
+ oldestXid = ShmemVariableCache->oldestXid;
+ oldestXidDB = ShmemVariableCache->oldestXidDB;
+ LWLockRelease(XidGenLock);
+
+ if (!TransactionIdIsNormal(oldestXid))
+ return true; /* shouldn't happen, but just in case */
+ if (!TransactionIdIsValid(xidVacLimit))
+ return true; /* this shouldn't happen anymore either */
+ if (TransactionIdFollowsOrEquals(nextXid, xidVacLimit))
+ return true; /* past xidVacLimit, don't delay updating */
+ if (!SearchSysCacheExists1(DATABASEOID, ObjectIdGetDatum(oldestXidDB)))
+ return true; /* could happen, per comments above */
+ return false;
}
-void
-CheckMaxObjectId(Oid assigned_oid)
-{
- Oid temp_oid;
- if (prefetched_oid_count == 0) /* make sure next/max is set, or
- * reload */
- GetNewObjectId(&temp_oid);
+/*
+ * GetNewObjectId -- allocate a new OID
+ *
+ * OIDs are generated by a cluster-wide counter. Since they are only 32 bits
+ * wide, counter wraparound will occur eventually, and therefore it is unwise
+ * to assume they are unique unless precautions are taken to make them so.
+ * Hence, this routine should generally not be used directly. The only direct
+ * callers should be GetNewOidWithIndex() and GetNewRelFileNode() in
+ * catalog/catalog.c.
+ */
+Oid
+GetNewObjectId(void)
+{
+ Oid result;
- /* ----------------
- * If we are below prefetched limits, do nothing
- * ----------------
- */
+ /* safety check, we should never get this far in a HS standby */
+ if (RecoveryInProgress())
+ elog(ERROR, "cannot assign OIDs during recovery");
- if (assigned_oid < next_prefetched_oid)
- return;
+ LWLockAcquire(OidGenLock, LW_EXCLUSIVE);
- /* ----------------
- * If we are here, we are coming from a 'copy from' with oid's
+ /*
+ * Check for wraparound of the OID counter. We *must* not return 0
+ * (InvalidOid), and in normal operation we mustn't return anything below
+ * FirstNormalObjectId since that range is reserved for initdb (see
+ * IsCatalogRelationOid()). Note we are relying on unsigned comparison.
*
- * If we are in the prefetched oid range, just bump it up
- * ----------------
+ * During initdb, we start the OID generator at FirstBootstrapObjectId, so
+ * we only wrap if before that point when in bootstrap or standalone mode.
+ * The first time through this routine after normal postmaster start, the
+ * counter will be forced up to FirstNormalObjectId. This mechanism
+ * leaves the OIDs between FirstBootstrapObjectId and FirstNormalObjectId
+ * available for automatic assignment during initdb, while ensuring they
+ * will never conflict with user-assigned OIDs.
*/
+ if (ShmemVariableCache->nextOid < ((Oid) FirstNormalObjectId))
+ {
+ if (IsPostmasterEnvironment)
+ {
+ /* wraparound, or first post-initdb assignment, in normal mode */
+ ShmemVariableCache->nextOid = FirstNormalObjectId;
+ ShmemVariableCache->oidCount = 0;
+ }
+ else
+ {
+ /* we may be bootstrapping, so don't enforce the full range */
+ if (ShmemVariableCache->nextOid < ((Oid) FirstBootstrapObjectId))
+ {
+ /* wraparound in standalone mode (unlikely but possible) */
+ ShmemVariableCache->nextOid = FirstNormalObjectId;
+ ShmemVariableCache->oidCount = 0;
+ }
+ }
+ }
- if (assigned_oid <= next_prefetched_oid + prefetched_oid_count - 1)
+ /* If we run out of logged for use oids then we must log more */
+ if (ShmemVariableCache->oidCount == 0)
{
- prefetched_oid_count -= assigned_oid - next_prefetched_oid + 1;
- next_prefetched_oid = assigned_oid + 1;
- return;
+ XLogPutNextOid(ShmemVariableCache->nextOid + VAR_OID_PREFETCH);
+ ShmemVariableCache->oidCount = VAR_OID_PREFETCH;
}
- /* ----------------
- * We have exceeded the prefetch oid range
- *
- * We should lock the database and kill all other backends
- * but we are loading oid's that we can not guarantee are unique
- * anyway, so we must rely on the user
- *
- * We now:
- * set the variable relation with the new max oid
- * force the backend to reload its oid cache
- *
- * By reloading the oid cache, we don't have to update the variable
- * relation every time when sequential OIDs are being loaded by COPY.
- * ----------------
- */
+ result = ShmemVariableCache->nextOid;
- SpinAcquire(OidGenLockId);
- VariableRelationPutNextOid(assigned_oid);
- SpinRelease(OidGenLockId);
+ (ShmemVariableCache->nextOid)++;
+ (ShmemVariableCache->oidCount)--;
- prefetched_oid_count = 0; /* force reload */
- GetNewObjectId(&temp_oid); /* cause target OID to be allocated */
-}
+ LWLockRelease(OidGenLock);
-#endif /* !XLOG */
+ return result;
+}
projects / postgresql / blobdiff