bufmgr.c

postgresql8.3.4源码,开源数据库

 *
 *		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);
	}
}

/* ---------------------------------------------------------------------
 *		DropDatabaseBuffers
 *
 *		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
DropDatabaseBuffers(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;
	}

	/* Make sure we can handle the pin inside the loop */
	ResourceOwnerEnlargeBuffers(CurrentResourceOwner);

	for (i = 0; i < NBuffers; i++)
	{
		bufHdr = &BufferDescriptors[i];
		LockBufHdr(bufHdr);
		if (RelFileNodeEquals(bufHdr->tag.rnode, rel->rd_node) &&
			(bufHdr->flags & BM_VALID) && (bufHdr->flags & BM_DIRTY))
		{
			PinBuffer_Locked(bufHdr);
			LWLockAcquire(bufHdr->content_lock, LW_SHARED);
			FlushBuffer(bufHdr, rel->rd_smgr);
			LWLockRelease(bufHdr->content_lock);
			UnpinBuffer(bufHdr, true);
		}
		else
			UnlockBufHdr(bufHdr);
	}
}

/* ---------------------------------------------------------------------
 *		FlushDatabaseBuffers
 *
 *		This function writes all dirty pages of a database out to disk
 *		(or more accurately, out to kernel disk buffers), ensuring that the
 *		kernel has an up-to-date view of the database.
 *
 *		Generally, the caller should be holding an appropriate lock to ensure
 *		no other backend is active in the target database; otherwise more
 *		pages could get dirtied.
 *
 *		Note we don't worry about flushing any pages of temporary relations.
 *		It's assumed these wouldn't be interesting.
 * --------------------------------------------------------------------
 */
void
FlushDatabaseBuffers(Oid dbid)
{
	int			i;
	volatile BufferDesc *bufHdr;

	/* Make sure we can handle the pin inside the loop */
	ResourceOwnerEnlargeBuffers(CurrentResourceOwner);

	for (i = 0; i < NBuffers; i++)
	{
		bufHdr = &BufferDescriptors[i];
		LockBufHdr(bufHdr);
		if (bufHdr->tag.rnode.dbNode == dbid &&
			(bufHdr->flags & BM_VALID) && (bufHdr->flags & BM_DIRTY))
		{
			PinBuffer_Locked(bufHdr);
			LWLockAcquire(bufHdr->content_lock, LW_SHARED);
			FlushBuffer(bufHdr, NULL);
			LWLockRelease(bufHdr->content_lock);
			UnpinBuffer(bufHdr, true);
		}
		else
			UnlockBufHdr(bufHdr);
	}
}

/*
 * ReleaseBuffer -- release the pin on a buffer
 */
void
ReleaseBuffer(Buffer buffer)
{
	volatile BufferDesc *bufHdr;

	if (!BufferIsValid(buffer))
		elog(ERROR, "bad buffer id: %d", buffer);

	ResourceOwnerForgetBuffer(CurrentResourceOwner, buffer);

	if (BufferIsLocal(buffer))
	{
		Assert(LocalRefCount[-buffer - 1] > 0);
		LocalRefCount[-buffer - 1]--;
		return;
	}

	bufHdr = &BufferDescriptors[buffer - 1];

	Assert(PrivateRefCount[buffer - 1] > 0);

	if (PrivateRefCount[buffer - 1] > 1)
		PrivateRefCount[buffer - 1]--;
	else
		UnpinBuffer(bufHdr, false);
}

/*
 * UnlockReleaseBuffer -- release the content lock and pin on a buffer
 *
 * This is just a shorthand for a common combination.
 */
void
UnlockReleaseBuffer(Buffer buffer)
{
	LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
	ReleaseBuffer(buffer);
}

/*
 * IncrBufferRefCount
 *		Increment the pin count on a buffer that we have *already* pinned
 *		at least once.
 *
 *		This function cannot be used on a buffer we do not have pinned,
 *		because it doesn't change the shared buffer state.
 */
void
IncrBufferRefCount(Buffer buffer)
{
	Assert(BufferIsPinned(buffer));
	ResourceOwnerEnlargeBuffers(CurrentResourceOwner);
	ResourceOwnerRememberBuffer(CurrentResourceOwner, buffer);
	if (BufferIsLocal(buffer))
		LocalRefCount[-buffer - 1]++;
	else
		PrivateRefCount[buffer - 1]++;
}

/*
 * SetBufferCommitInfoNeedsSave
 *
 *	Mark a buffer dirty when we have updated tuple commit-status bits in it.
 *
 * This is essentially the same as MarkBufferDirty, except that the caller
 * might have only share-lock instead of exclusive-lock on the buffer's
 * content lock.  We preserve the distinction mainly as a way of documenting
 * that the caller has not made a critical data change --- the status-bit
 * update could be redone by someone else just as easily.  Therefore, no WAL
 * log record need be generated, whereas calls to MarkBufferDirty really ought
 * to be associated with a WAL-entry-creating action.
 */
void
SetBufferCommitInfoNeedsSave(Buffer buffer)
{
	volatile BufferDesc *bufHdr;

	if (!BufferIsValid(buffer))
		elog(ERROR, "bad buffer id: %d", buffer);

	if (BufferIsLocal(buffer))
	{
		MarkLocalBufferDirty(buffer);
		return;
	}

	bufHdr = &BufferDescriptors[buffer - 1];

	Assert(PrivateRefCount[buffer - 1] > 0);
	/* here, either share or exclusive lock is OK */
	Assert(LWLockHeldByMe(bufHdr->content_lock));

	/*
	 * This routine might get called many times on the same page, if we are
	 * making the first scan after commit of an xact that added/deleted many
	 * tuples.	So, be as quick as we can if the buffer is already dirty.  We
	 * do this by not acquiring spinlock if it looks like the status bits are
	 * already OK.	(Note it is okay if someone else clears BM_JUST_DIRTIED
	 * immediately after we look, because the buffer content update is already
	 * done and will be reflected in the I/O.)
	 */
	if ((bufHdr->flags & (BM_DIRTY | BM_JUST_DIRTIED)) !=
		(BM_DIRTY | BM_JUST_DIRTIED))
	{
		LockBufHdr(bufHdr);
		Assert(bufHdr->refcount > 0);
		if (!(bufHdr->flags & BM_DIRTY) && VacuumCostActive)
			VacuumCostBalance += VacuumCostPageDirty;
		bufHdr->flags |= (BM_DIRTY | BM_JUST_DIRTIED);
		UnlockBufHdr(bufHdr);
	}
}

/*
 * Release buffer content locks for shared buffers.
 *
 * Used to clean up after errors.
 *
 * Currently, we can expect that lwlock.c's LWLockReleaseAll() took care
 * of releasing buffer content locks per se; the only thing we need to deal
 * with here is clearing any PIN_COUNT request that was in progress.
 */
void
UnlockBuffers(void)
{
	volatile BufferDesc *buf = PinCountWaitBuf;

	if (buf)
	{
		LockBufHdr(buf);

		/*
		 * Don't complain if flag bit not set; it could have been reset but we
		 * got a cancel/die interrupt before getting the signal.
		 */
		if ((buf->flags & BM_PIN_COUNT_WAITER) != 0 &&
			buf->wait_backend_pid == MyProcPid)
			buf->flags &= ~BM_PIN_COUNT_WAITER;

		UnlockBufHdr(buf);

		PinCountWaitBuf = NULL;
	}
}

/*
 * Acquire or release the content_lock for the buffer.
 */
void
LockBuffer(Buffer buffer, int mode)
{
	volatile BufferDesc *buf;

	Assert(BufferIsValid(buffer));
	if (BufferIsLocal(buffer))
		return;					/* local buffers need no lock */

	buf = &(BufferDescriptors[buffer - 1]);

	if (mode == BUFFER_LOCK_UNLOCK)
		LWLockRelease(buf->content_lock);
	else if (mode == BUFFER_LOCK_SHARE)
		LWLockAcquire(buf->content_lock, LW_SHARED);
	else if (mode == BUFFER_LOCK_EXCLUSIVE)
		LWLockAcquire(buf->content_lock, LW_EXCLUSIVE);
	else
		elog(ERROR, "unrecognized buffer lock mode: %d", mode);
}

/*
 * Acquire the content_lock for the buffer, but only if we don't have to wait.
 *
 * This assumes the caller wants BUFFER_LOCK_EXCLUSIVE mode.
 */
bool
ConditionalLockBuffer(Buffer buffer)
{
	volatile BufferDesc *buf;

	Assert(BufferIsValid(buffer));
	if (BufferIsLocal(buffer))
		return true;			/* act as though we got it */

	buf = &(BufferDescriptors[buffer - 1]);

	r