md.c

PostgreSQL 8.1.4的源码 适用于Linux下的开源数据库系统

			return false;
	}

	return true;
}

/*
 *	mdnblocks() -- Get the number of blocks stored in a relation.
 *
 *		Important side effect: all segments of the relation are opened
 *		and added to the mdfd_chain list.  If this routine has not been
 *		called, then only segments up to the last one actually touched
 *		are present in the chain...
 *
 *		Returns # of blocks, or InvalidBlockNumber on error.
 */
BlockNumber
mdnblocks(SMgrRelation reln)
{
	MdfdVec    *v = mdopen(reln, false);

#ifndef LET_OS_MANAGE_FILESIZE
	BlockNumber nblocks;
	BlockNumber segno = 0;

	/*
	 * Skip through any segments that aren't the last one, to avoid redundant
	 * seeks on them.  We have previously verified that these segments are
	 * exactly RELSEG_SIZE long, and it's useless to recheck that each time.
	 * (NOTE: this assumption could only be wrong if another backend has
	 * truncated the relation.	We rely on higher code levels to handle that
	 * scenario by closing and re-opening the md fd.)
	 */
	while (v->mdfd_chain != NULL)
	{
		segno++;
		v = v->mdfd_chain;
	}

	for (;;)
	{
		nblocks = _mdnblocks(v->mdfd_vfd, BLCKSZ);
		if (nblocks > ((BlockNumber) RELSEG_SIZE))
			elog(FATAL, "segment too big");
		if (nblocks < ((BlockNumber) RELSEG_SIZE))
			return (segno * ((BlockNumber) RELSEG_SIZE)) + nblocks;

		/*
		 * If segment is exactly RELSEG_SIZE, advance to next one.
		 */
		segno++;

		if (v->mdfd_chain == NULL)
		{
			/*
			 * Because we pass O_CREAT, we will create the next segment (with
			 * zero length) immediately, if the last segment is of length
			 * REL_SEGSIZE.  This is unnecessary but harmless, and testing for
			 * the case would take more cycles than it seems worth.
			 */
			v->mdfd_chain = _mdfd_openseg(reln, segno, O_CREAT);
			if (v->mdfd_chain == NULL)
				return InvalidBlockNumber;		/* failed? */
		}

		v = v->mdfd_chain;
	}
#else
	return _mdnblocks(v->mdfd_vfd, BLCKSZ);
#endif
}

/*
 *	mdtruncate() -- Truncate relation to specified number of blocks.
 *
 *		Returns # of blocks or InvalidBlockNumber on error.
 */
BlockNumber
mdtruncate(SMgrRelation reln, BlockNumber nblocks, bool isTemp)
{
	MdfdVec    *v;
	BlockNumber curnblk;

#ifndef LET_OS_MANAGE_FILESIZE
	BlockNumber priorblocks;
#endif

	/*
	 * NOTE: mdnblocks makes sure we have opened all existing segments, so
	 * that truncate/delete loop will get them all!
	 */
	curnblk = mdnblocks(reln);
	if (curnblk == InvalidBlockNumber)
		return InvalidBlockNumber;		/* mdnblocks failed */
	if (nblocks > curnblk)
		return InvalidBlockNumber;		/* bogus request */
	if (nblocks == curnblk)
		return nblocks;			/* no work */

	v = mdopen(reln, false);

#ifndef LET_OS_MANAGE_FILESIZE
	priorblocks = 0;
	while (v != NULL)
	{
		MdfdVec    *ov = v;

		if (priorblocks > nblocks)
		{
			/*
			 * This segment is no longer wanted at all (and has already been
			 * unlinked from the mdfd_chain). We truncate the file before
			 * deleting it because if other backends are holding the file
			 * open, the unlink will fail on some platforms. Better a
			 * zero-size file gets left around than a big file...
			 */
			FileTruncate(v->mdfd_vfd, 0);
			FileUnlink(v->mdfd_vfd);
			v = v->mdfd_chain;
			Assert(ov != reln->md_fd);	/* we never drop the 1st segment */
			pfree(ov);
		}
		else if (priorblocks + ((BlockNumber) RELSEG_SIZE) > nblocks)
		{
			/*
			 * This is the last segment we want to keep. Truncate the file to
			 * the right length, and clear chain link that points to any
			 * remaining segments (which we shall zap). NOTE: if nblocks is
			 * exactly a multiple K of RELSEG_SIZE, we will truncate the K+1st
			 * segment to 0 length but keep it. This is mainly so that the
			 * right thing happens if nblocks==0.
			 */
			BlockNumber lastsegblocks = nblocks - priorblocks;

			if (FileTruncate(v->mdfd_vfd, lastsegblocks * BLCKSZ) < 0)
				return InvalidBlockNumber;
			if (!isTemp)
			{
				if (!register_dirty_segment(reln, v))
					return InvalidBlockNumber;
			}
			v = v->mdfd_chain;
			ov->mdfd_chain = NULL;
		}
		else
		{
			/*
			 * We still need this segment and 0 or more blocks beyond it, so
			 * nothing to do here.
			 */
			v = v->mdfd_chain;
		}
		priorblocks += RELSEG_SIZE;
	}
#else
	if (FileTruncate(v->mdfd_vfd, nblocks * BLCKSZ) < 0)
		return InvalidBlockNumber;
	if (!isTemp)
	{
		if (!register_dirty_segment(reln, v))
			return InvalidBlockNumber;
	}
#endif

	return nblocks;
}

/*
 *	mdimmedsync() -- Immediately sync a relation to stable storage.
 *
 * Note that only writes already issued are synced; this routine knows
 * nothing of dirty buffers that may exist inside the buffer manager.
 */
bool
mdimmedsync(SMgrRelation reln)
{
	MdfdVec    *v;
	BlockNumber curnblk;

	/*
	 * NOTE: mdnblocks makes sure we have opened all existing segments, so
	 * that fsync loop will get them all!
	 */
	curnblk = mdnblocks(reln);
	if (curnblk == InvalidBlockNumber)
		return false;			/* mdnblocks failed */

	v = mdopen(reln, false);

#ifndef LET_OS_MANAGE_FILESIZE
	while (v != NULL)
	{
		if (FileSync(v->mdfd_vfd) < 0)
			return false;
		v = v->mdfd_chain;
	}
#else
	if (FileSync(v->mdfd_vfd) < 0)
		return false;
#endif

	return true;
}

/*
 *	mdsync() -- Sync previous writes to stable storage.
 *
 * This is only called during checkpoints, and checkpoints should only
 * occur in processes that have created a pendingOpsTable.
 */
bool
mdsync(void)
{
	HASH_SEQ_STATUS hstat;
	PendingOperationEntry *entry;

	if (!pendingOpsTable)
		return false;

	/*
	 * If we are in the bgwriter, the sync had better include all fsync
	 * requests that were queued by backends before the checkpoint REDO point
	 * was determined.	We go that a little better by accepting all requests
	 * queued up to the point where we start fsync'ing.
	 */
	AbsorbFsyncRequests();

	hash_seq_init(&hstat, pendingOpsTable);
	while ((entry = (PendingOperationEntry *) hash_seq_search(&hstat)) != NULL)
	{
		/*
		 * If fsync is off then we don't have to bother opening the file at
		 * all.  (We delay checking until this point so that changing fsync on
		 * the fly behaves sensibly.)
		 */
		if (enableFsync)
		{
			SMgrRelation reln;
			MdfdVec    *seg;

			/*
			 * Find or create an smgr hash entry for this relation. This may
			 * seem a bit unclean -- md calling smgr?  But it's really the
			 * best solution.  It ensures that the open file reference isn't
			 * permanently leaked if we get an error here. (You may say "but
			 * an unreferenced SMgrRelation is still a leak!" Not really,
			 * because the only case in which a checkpoint is done by a
			 * process that isn't about to shut down is in the bgwriter, and
			 * it will periodically do smgrcloseall().	This fact justifies
			 * our not closing the reln in the success path either, which is a
			 * good thing since in non-bgwriter cases we couldn't safely do
			 * that.)  Furthermore, in many cases the relation will have been
			 * dirtied through this same smgr relation, and so we can save a
			 * file open/close cycle.
			 */
			reln = smgropen(entry->rnode);

			/*
			 * It is possible that the relation has been dropped or truncated
			 * since the fsync request was entered.  Therefore, we have to
			 * allow file-not-found errors.  This applies both during
			 * _mdfd_getseg() and during FileSync, since fd.c might have
			 * closed the file behind our back.
			 */
			seg = _mdfd_getseg(reln,
							   entry->segno * ((BlockNumber) RELSEG_SIZE),
							   true);
			if (seg)
			{
				if (FileSync(seg->mdfd_vfd) < 0 &&
					errno != ENOENT)
				{
					ereport(LOG,
							(errcode_for_file_access(),
							 errmsg("could not fsync segment %u of relation %u/%u/%u: %m",
									entry->segno,
									entry->rnode.spcNode,
									entry->rnode.dbNode,
									entry->rnode.relNode)));
					return false;
				}
			}
		}

		/* Okay, delete this entry */
		if (hash_search(pendingOpsTable, entry,
						HASH_REMOVE, NULL) == NULL)
			elog(ERROR, "pendingOpsTable corrupted");
	}

	return true;
}

/*
 * register_dirty_segment() -- Mark a relation segment as needing fsync
 *
 * If there is a local pending-ops table, just make an entry in it for
 * mdsync to process later.  Otherwise, try to pass off the fsync request
 * to the background writer process.  If that fails, just do the fsync
 * locally before returning (we expect this will not happen often enough
 * to be a performance problem).
 *
 * A false result implies I/O failure during local fsync.  errno will be
 * valid for error reporting.
 */
static bool
register_dirty_segment(SMgrRelation reln, MdfdVec *seg)
{
	if (pendingOpsTable)
	{
		PendingOperationEntry entry;

		/* ensure any pad bytes in the struct are zeroed */
		MemSet(&entry, 0, sizeof(entry));
		entry.rnode = reln->smgr_rnode;
		entry.segno = seg->mdfd_segno;

		(void) hash_search(pendingOpsTable, &entry, HASH_ENTER, NULL);
		return true;
	}
	else
	{
		if (ForwardFsyncRequest(reln->smgr_rnode, seg->mdfd_segno))
			return true;
	}

	if (FileSync(seg->mdfd_vfd) < 0)
		return false;
	return true;
}

/*
 * RememberFsyncRequest() -- callback from bgwriter side of fsync request
 *
 * We stuff the fsync request into the local hash table for execution
 * during the bgwriter's next checkpoint.
 */
void
RememberFsyncRequest(RelFileNode rnode, BlockNumber segno)
{
	PendingOperationEntry entry;

	Assert(pendingOpsTable);

	/* ensure any pad bytes in the struct are zeroed */
	MemSet(&entry, 0, sizeof(entry));
	entry.rnode = rnode;
	entry.segno = segno;

	(void) hash_search(pendingOpsTable, &entry, HASH_ENTER, NULL);
}

/*
 *	_fdvec_alloc() -- Make a MdfdVec object.
 */
static MdfdVec *
_fdvec_alloc(void)
{
	return (MdfdVec *) MemoryContextAlloc(MdCxt, sizeof(MdfdVec));
}

#ifndef LET_OS_MANAGE_FILESIZE

/*
 * Open the specified segment of the relation,
 * and make a MdfdVec object for it.  Returns NULL on failure.
 */
static MdfdVec *
_mdfd_openseg(SMgrRelation reln, BlockNumber segno, int oflags)
{
	MdfdVec    *v;
	int			fd;
	char	   *path,
			   *fullpath;

	path = relpath(reln->smgr_rnode);

	if (segno > 0)
	{
		/* be sure we have enough space for the '.segno' */
		fullpath = (char *) palloc(strlen(path) + 12);
		sprintf(fullpath, "%s.%u", path, segno);
		pfree(path);
	}
	else
		fullpath = path;

	/* open the file */
	fd = PathNameOpenFile(fullpath, O_RDWR | PG_BINARY | oflags, 0600);

	pfree(fullpath);

	if (fd < 0)
		return NULL;

	/* allocate an mdfdvec entry for it */
	v = _fdvec_alloc();

	/* fill the entry */
	v->mdfd_vfd = fd;
	v->mdfd_segno = segno;
	v->mdfd_chain = NULL;
	Assert(_mdnblocks(fd, BLCKSZ) <= ((BlockNumber) RELSEG_SIZE));

	/* all done */
	return v;
}
#endif   /* LET_OS_MANAGE_FILESIZE */

/*
 *	_mdfd_getseg() -- Find the segment of the relation holding the
 *		specified block.  ereport's on failure.
 *		(Optionally, can return NULL instead of ereport for ENOENT.)
 */
static MdfdVec *
_mdfd_getseg(SMgrRelation reln, BlockNumber blkno, bool allowNotFound)
{
	MdfdVec    *v = mdopen(reln, allowNotFound);

#ifndef LET_OS_MANAGE_FILESIZE
	BlockNumber segstogo;
	BlockNumber nextsegno;

	if (!v)
		return NULL;			/* only possible if allowNotFound */

	for (segstogo = blkno / ((BlockNumber) RELSEG_SIZE), nextsegno = 1;
		 segstogo > 0;
		 nextsegno++, segstogo--)
	{
		if (v->mdfd_chain == NULL)
		{
			/*
			 * We will create the next segment only if the target block is
			 * within it.  This prevents Sorcerer's Apprentice syndrome if a
			 * bug at higher levels causes us to be handed a ridiculously
			 * large blkno --- otherwise we could create many thousands of
			 * empty segment files before reaching the "target" block.	We
			 * should never need to create more than one new segment per call,
			 * so this restriction seems reasonable.
			 *
			 * BUT: when doing WAL recovery, disable this logic and create
			 * segments unconditionally.  In this case it seems better to
			 * assume the given blkno is good (it presumably came from a
			 * CRC-checked WAL record); furthermore this lets us cope in the
			 * case where we are replaying WAL data that has a write into a
			 * high-numbered segment of a relation that was later deleted.	We
			 * want to go ahead and create the segments so we can finish out
			 * the replay.
			 */
			v->mdfd_chain = _mdfd_openseg(reln,
										  nextsegno,
								(segstogo == 1 || InRecovery) ? O_CREAT : 0);
			if (v->mdfd_chain == NULL)
			{
				if (allowNotFound && errno == ENOENT)
					return NULL;
				ereport(ERROR,
						(errcode_for_file_access(),
						 errmsg("could not open segment %u of relation %u/%u/%u (target block %u): %m",
								nextsegno,
								reln->smgr_rnode.spcNode,
								reln->smgr_rnode.dbNode,
								reln->smgr_rnode.relNode,
								blkno)));
			}
		}
		v = v->mdfd_chain;
	}
#endif

	return v;
}

/*
 * Get number of blocks present in a single disk file
 */
static BlockNumber
_mdnblocks(File file, Size blcksz)
{
	long		len;

	len = FileSeek(file, 0L, SEEK_END);
	if (len < 0)
		return 0;				/* on failure, assume file is empty */
	return (BlockNumber) (len / blcksz);
}