bufmgr.c

PostgreSQL 8.1.4的源码 适用于Linux下的开源数据库系统

ShowBufferUsage(void)
{
	StringInfoData str;
	float		hitrate;
	float		localhitrate;

	initStringInfo(&str);

	if (ReadBufferCount == 0)
		hitrate = 0.0;
	else
		hitrate = (float) BufferHitCount *100.0 / ReadBufferCount;

	if (ReadLocalBufferCount == 0)
		localhitrate = 0.0;
	else
		localhitrate = (float) LocalBufferHitCount *100.0 / ReadLocalBufferCount;

	appendStringInfo(&str,
	"!\tShared blocks: %10ld read, %10ld written, buffer hit rate = %.2f%%\n",
				ReadBufferCount - BufferHitCount, BufferFlushCount, hitrate);
	appendStringInfo(&str,
	"!\tLocal  blocks: %10ld read, %10ld written, buffer hit rate = %.2f%%\n",
					 ReadLocalBufferCount - LocalBufferHitCount, LocalBufferFlushCount, localhitrate);
	appendStringInfo(&str,
					 "!\tDirect blocks: %10ld read, %10ld written\n",
					 NDirectFileRead, NDirectFileWrite);

	return str.data;
}

void
ResetBufferUsage(void)
{
	BufferHitCount = 0;
	ReadBufferCount = 0;
	BufferFlushCount = 0;
	LocalBufferHitCount = 0;
	ReadLocalBufferCount = 0;
	LocalBufferFlushCount = 0;
	NDirectFileRead = 0;
	NDirectFileWrite = 0;
}

/*
 *		AtEOXact_Buffers - clean up at end of transaction.
 *
 *		As of PostgreSQL 8.0, buffer pins should get released by the
 *		ResourceOwner mechanism.  This routine is just a debugging
 *		cross-check that no pins remain.
 */
void
AtEOXact_Buffers(bool isCommit)
{
#ifdef USE_ASSERT_CHECKING
	if (assert_enabled)
	{
		int			i;

		for (i = 0; i < NBuffers; i++)
		{
			Assert(PrivateRefCount[i] == 0);
		}
	}
#endif

	AtEOXact_LocalBuffers(isCommit);

	/* Make sure we reset the strategy hint in case VACUUM errored out */
	StrategyHintVacuum(false);
}

/*
 * InitBufferPoolBackend --- second-stage initialization of a new backend
 *
 * This is called after we have acquired a PGPROC and so can safely get
 * LWLocks.  We don't currently need to do anything at this stage ...
 * except register a shmem-exit callback.  AtProcExit_Buffers needs LWLock
 * access, and thereby has to be called at the corresponding phase of
 * backend shutdown.
 */
void
InitBufferPoolBackend(void)
{
	on_shmem_exit(AtProcExit_Buffers, 0);
}

/*
 * Ensure we have released all shared-buffer locks and pins during backend exit
 */
static void
AtProcExit_Buffers(int code, Datum arg)
{
	int			i;

	AbortBufferIO();
	UnlockBuffers();

	for (i = 0; i < NBuffers; i++)
	{
		if (PrivateRefCount[i] != 0)
		{
			volatile BufferDesc *buf = &(BufferDescriptors[i]);

			/*
			 * We don't worry about updating ResourceOwner; if we even got
			 * here, it suggests that ResourceOwners are messed up.
			 */
			PrivateRefCount[i] = 1;		/* make sure we release shared pin */
			UnpinBuffer(buf, false, false /* don't change freelist */ );
			Assert(PrivateRefCount[i] == 0);
		}
	}

	/* localbuf.c needs a chance too */
	AtProcExit_LocalBuffers();
}

/*
 * Helper routine to issue warnings when a buffer is unexpectedly pinned
 */
void
PrintBufferLeakWarning(Buffer buffer)
{
	volatile BufferDesc *buf;
	int32		loccount;

	Assert(BufferIsValid(buffer));
	if (BufferIsLocal(buffer))
	{
		buf = &LocalBufferDescriptors[-buffer - 1];
		loccount = LocalRefCount[-buffer - 1];
	}
	else
	{
		buf = &BufferDescriptors[buffer - 1];
		loccount = PrivateRefCount[buffer - 1];
	}

	/* theoretically we should lock the bufhdr here */
	elog(WARNING,
		 "buffer refcount leak: [%03d] "
		 "(rel=%u/%u/%u, blockNum=%u, flags=0x%x, refcount=%u %d)",
		 buffer,
		 buf->tag.rnode.spcNode, buf->tag.rnode.dbNode,
		 buf->tag.rnode.relNode,
		 buf->tag.blockNum, buf->flags,
		 buf->refcount, loccount);
}

/*
 * FlushBufferPool
 *
 * Flush all dirty blocks in buffer pool to disk at the checkpoint time.
 * Local relations do not participate in checkpoints, so they don't need to be
 * flushed.
 */
void
FlushBufferPool(void)
{
	BufferSync();
	smgrsync();
}


/*
 * Do whatever is needed to prepare for commit at the bufmgr and smgr levels
 */
void
BufmgrCommit(void)
{
	/* Nothing to do in bufmgr anymore... */

	smgrcommit();
}

/*
 * BufferGetBlockNumber
 *		Returns the block number associated with a buffer.
 *
 * Note:
 *		Assumes that the buffer is valid and pinned, else the
 *		value may be obsolete immediately...
 */
BlockNumber
BufferGetBlockNumber(Buffer buffer)
{
	volatile BufferDesc *bufHdr;

	Assert(BufferIsPinned(buffer));

	if (BufferIsLocal(buffer))
		bufHdr = &(LocalBufferDescriptors[-buffer - 1]);
	else
		bufHdr = &BufferDescriptors[buffer - 1];

	/* pinned, so OK to read tag without spinlock */
	return bufHdr->tag.blockNum;
}

/*
 * BufferGetFileNode
 *		Returns the relation ID (RelFileNode) associated with a buffer.
 *
 * This should make the same checks as BufferGetBlockNumber, but since the
 * two are generally called together, we don't bother.
 */
RelFileNode
BufferGetFileNode(Buffer buffer)
{
	volatile BufferDesc *bufHdr;

	if (BufferIsLocal(buffer))
		bufHdr = &(LocalBufferDescriptors[-buffer - 1]);
	else
		bufHdr = &BufferDescriptors[buffer - 1];

	return bufHdr->tag.rnode;
}

/*
 * FlushBuffer
 *		Physically write out a shared buffer.
 *
 * NOTE: this actually just passes the buffer contents to the kernel; the
 * real write to disk won't happen until the kernel feels like it.  This
 * is okay from our point of view since we can redo the changes from WAL.
 * However, we will need to force the changes to disk via fsync before
 * we can checkpoint WAL.
 *
 * The caller must hold a pin on the buffer and have share-locked the
 * buffer contents.  (Note: a share-lock does not prevent updates of
 * hint bits in the buffer, so the page could change while the write
 * is in progress, but we assume that that will not invalidate the data
 * written.)
 *
 * If the caller has an smgr reference for the buffer's relation, pass it
 * as the second parameter.  If not, pass NULL.
 */
static void
FlushBuffer(volatile BufferDesc *buf, SMgrRelation reln)
{
	XLogRecPtr	recptr;
	ErrorContextCallback errcontext;

	/*
	 * Acquire the buffer's io_in_progress lock.  If StartBufferIO returns
	 * false, then someone else flushed the buffer before we could, so we need
	 * not do anything.
	 */
	if (!StartBufferIO(buf, false))
		return;

	/* Setup error traceback support for ereport() */
	errcontext.callback = buffer_write_error_callback;
	errcontext.arg = (void *) buf;
	errcontext.previous = error_context_stack;
	error_context_stack = &errcontext;

	/* Find smgr relation for buffer */
	if (reln == NULL)
		reln = smgropen(buf->tag.rnode);

	/*
	 * Force XLOG flush up to buffer's LSN.  This implements the basic WAL
	 * rule that log updates must hit disk before any of the data-file changes
	 * they describe do.
	 */
	recptr = BufferGetLSN(buf);
	XLogFlush(recptr);

	/*
	 * Now it's safe to write buffer to disk. Note that no one else should
	 * have been able to write it while we were busy with log flushing because
	 * we have the io_in_progress lock.
	 */

	/* To check if block content changes while flushing. - vadim 01/17/97 */
	LockBufHdr_NoHoldoff(buf);
	buf->flags &= ~BM_JUST_DIRTIED;
	UnlockBufHdr_NoHoldoff(buf);

	smgrwrite(reln,
			  buf->tag.blockNum,
			  (char *) BufHdrGetBlock(buf),
			  false);

	BufferFlushCount++;

	/*
	 * Mark the buffer as clean (unless BM_JUST_DIRTIED has become set) and
	 * end the io_in_progress state.
	 */
	TerminateBufferIO(buf, true, 0);

	/* Pop the error context stack */
	error_context_stack = errcontext.previous;
}

/*
 * RelationGetNumberOfBlocks
 *		Determines the current number of pages in the relation.
 */
BlockNumber
RelationGetNumberOfBlocks(Relation relation)
{
	/* Open it at the smgr level if not already done */
	RelationOpenSmgr(relation);

	return smgrnblocks(relation->rd_smgr);
}

/*
 * RelationTruncate
 *		Physically truncate a relation to the specified number of blocks.
 *
 * As of Postgres 8.1, this includes getting rid of any buffers for the
 * blocks that are to be dropped; previously, callers had to do that.
 */
void
RelationTruncate(Relation rel, BlockNumber nblocks)
{
	/* Open it at the smgr level if not already done */
	RelationOpenSmgr(rel);

	/* Make sure rd_targblock isn't pointing somewhere past end */
	rel->rd_targblock = InvalidBlockNumber;

	/* Do the real work */
	smgrtruncate(rel->rd_smgr, nblocks, rel->rd_istemp);
}

/* ---------------------------------------------------------------------
 *		DropRelFileNodeBuffers
 *
 *		This function removes from the buffer pool all the pages of the
 *		specified relation that have block numbers >= firstDelBlock.
 *		(In particular, with firstDelBlock = 0, all pages are removed.)
 *		Dirty pages are simply dropped, without bothering to write them
 *		out first.	Therefore, this is NOT rollback-able, and so should be
 *		used only with extreme caution!
 *
 *		Currently, this is called only from smgr.c when the underlying file
 *		is about to be deleted or truncated (firstDelBlock is needed for
 *		the truncation case).  The data in the affected pages would therefore
 *		be deleted momentarily anyway, and there is no point in writing it.
 *		It is the responsibility of higher-level code to ensure that the
 *		deletion or truncation does not lose any data that could be needed
 *		later.	It is also the responsibility of higher-level code to ensure
 *		that no other process could be trying to load more pages of the
 *		relation into buffers.
 *
 *		XXX currently it sequentially searches the buffer pool, should be
 *		changed to more clever ways of searching.  However, this routine
 *		is used only in code paths that aren't very performance-critical,
 *		and we shouldn't slow down the hot paths to make it faster ...
 * --------------------------------------------------------------------
 */
void
DropRelFileNodeBuffers(RelFileNode rnode, bool istemp,
					   BlockNumber firstDelBlock)
{
	int			i;

	if (istemp)
	{
		DropRelFileNodeLocalBuffers(rnode, firstDelBlock);
		return;
	}

	for (i = 0; i < NBuffers; i++)
	{
		volatile BufferDesc *bufHdr = &BufferDescriptors[i];

		LockBufHdr(bufHdr);
		if (RelFileNodeEquals(bufHdr->tag.rnode, rnode) &&
			bufHdr->tag.blockNum >= firstDelBlock)
			InvalidateBuffer(bufHdr);	/* releases spinlock */
		else
			UnlockBufHdr(bufHdr);
	}
}

/* ---------------------------------------------------------------------
 *		DropBuffers
 *
 *		This function removes all the buffers in the buffer cache for a
 *		particular database.  Dirty pages are simply dropped, without
 *		bothering to write them out first.	This is used when we destroy a
 *		database, to avoid trying to flush data to disk when the directory
 *		tree no longer exists.	Implementation is pretty similar to
 *		DropRelFileNodeBuffers() which is for destroying just one relation.
 * --------------------------------------------------------------------
 */
void
DropBuffers(Oid dbid)
{
	int			i;
	volatile BufferDesc *bufHdr;

	/*
	 * We needn't consider local buffers, since by assumption the target
	 * database isn't our own.
	 */

	for (i = 0; i < NBuffers; i++)
	{
		bufHdr = &BufferDescriptors[i];
		LockBufHdr(bufHdr);
		if (bufHdr->tag.rnode.dbNode == dbid)
			InvalidateBuffer(bufHdr);	/* releases spinlock */
		else
			UnlockBufHdr(bufHdr);
	}
}

/* -----------------------------------------------------------------
 *		PrintBufferDescs
 *
 *		this function prints all the buffer descriptors, for debugging
 *		use only.
 * -----------------------------------------------------------------
 */
#ifdef NOT_USED
void
PrintBufferDescs(void)
{
	int			i;
	volatile BufferDesc *buf = BufferDescriptors;

	for (i = 0; i < NBuffers; ++i, ++buf)
	{
		/* theoretically we should lock the bufhdr here */
		elog(LOG,
			 "[%02d] (freeNext=%d, rel=%u/%u/%u, "
			 "blockNum=%u, flags=0x%x, refcount=%u %d)",
			 i, buf->freeNext,
			 buf->tag.rnode.spcNode, buf->tag.rnode.dbNode,
			 buf->tag.rnode.relNode,
			 buf->tag.blockNum, buf->flags,
			 buf->refcount, PrivateRefCount[i]);
	}
}
#endif

#ifdef NOT_USED
void
PrintPinnedBufs(void)
{
	int			i;
	volatile BufferDesc *buf = BufferDescriptors;

	for (i = 0; i < NBuffers; ++i, ++buf)
	{
		if (PrivateRefCount[i] > 0)
		{
			/* theoretically we should lock the bufhdr here */
			elog(LOG,
				 "[%02d] (freeNext=%d, rel=%u/%u/%u, "
				 "blockNum=%u, flags=0x%x, refcount=%u %d)",
				 i, buf->freeNext,
				 buf->tag.rnode.spcNode, buf->tag.rnode.dbNode,
				 buf->tag.rnode.relNode,
				 buf->tag.blockNum, buf->flags,
				 buf->refcount, PrivateRefCount[i]);
		}
	}
}
#endif

/* ---------------------------------------------------------------------
 *		FlushRelationBuffers
 *
 *		This function writes all dirty pages of a relation out to disk
 *		(or more accurately, out to kernel disk buffers), ensuring that the
 *		kernel has an up-to-date view of the relation.
 *
 *		Generally, the caller should be holding AccessExclusiveLock on the
 *		target relation to ensure that no other backend is busy dirtying
 *		more blocks of the relation; the effects can't be expected to last
 *		after the lock is released.
 *
 *		XXX currently it sequentially searches the buffer pool, should be
 *		changed to more clever ways of searching.  This routine is not
 *		used in any performance-critical code paths, so it's not worth
 *		adding additional overhead to normal paths to make it go faster;
 *		but see also DropRelFileNodeBuffers.
 * --------------------------------------------------------------------
 */
void
FlushRelationBuffers(Relation rel)
{
	int			i;
	volatile BufferDesc *bufHdr;

	/* Open rel at the smgr level if not already done */
	RelationOpenSmgr(rel);

	if (rel->rd_istemp)
	{
		for (i = 0; i < NLocBuffer; i++)
		{
			bufHdr = &LocalBufferDescriptors[i];
			if (RelFileNodeEquals(bufHdr->tag.rnode, rel->rd_node) &&
				(bufHdr->flags & BM_VALID) && (bufHdr->flags & BM_DIRTY))
			{
				ErrorContextCallback errcontext;

				/* Setup error traceback support for ereport() */
				errcontext.callback = buffer_write_error_callback;
				errcontext.arg = (void *) bufHdr;
				errcontext.previous = error_context_stack;
				error_context_stack = &errcontext;

				smgrwrite(rel->rd_smgr,
						  bufHdr->tag.blockNum,
						  (char *) LocalBufHdrGetBlock(bufHdr),
						  true);

				bufHdr->flags &= ~(BM_DIRTY | BM_JUST_DIRTIED);

				/* Pop the error context stack */
				error_context_stack = errcontext.previous;
			}
		}

		return;
	}