md.c

postgresql8.3.4源码,开源数据库

 *
 * 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).
 */
static void
register_dirty_segment(SMgrRelation reln, MdfdVec *seg)
{
	if (pendingOpsTable)
	{
		/* push it into local pending-ops table */
		RememberFsyncRequest(reln->smgr_rnode, seg->mdfd_segno);
	}
	else
	{
		if (ForwardFsyncRequest(reln->smgr_rnode, seg->mdfd_segno))
			return;				/* passed it off successfully */

		if (FileSync(seg->mdfd_vfd) < 0)
			ereport(ERROR,
					(errcode_for_file_access(),
				errmsg("could not fsync segment %u of relation %u/%u/%u: %m",
					   seg->mdfd_segno,
					   reln->smgr_rnode.spcNode,
					   reln->smgr_rnode.dbNode,
					   reln->smgr_rnode.relNode)));
	}
}

/*
 * register_unlink() -- Schedule a file to be deleted after next checkpoint
 *
 * As with register_dirty_segment, this could involve either a local or
 * a remote pending-ops table.
 */
static void
register_unlink(RelFileNode rnode)
{
	if (pendingOpsTable)
	{
		/* push it into local pending-ops table */
		RememberFsyncRequest(rnode, UNLINK_RELATION_REQUEST);
	}
	else
	{
		/*
		 * Notify the bgwriter about it.  If we fail to queue the request
		 * message, we have to sleep and try again, because we can't simply
		 * delete the file now.  Ugly, but hopefully won't happen often.
		 *
		 * XXX should we just leave the file orphaned instead?
		 */
		Assert(IsUnderPostmaster);
		while (!ForwardFsyncRequest(rnode, UNLINK_RELATION_REQUEST))
			pg_usleep(10000L);	/* 10 msec seems a good number */
	}
}

/*
 * RememberFsyncRequest() -- callback from bgwriter side of fsync request
 *
 * We stuff most fsync requests into the local hash table for execution
 * during the bgwriter's next checkpoint.  UNLINK requests go into a
 * separate linked list, however, because they get processed separately.
 *
 * The range of possible segment numbers is way less than the range of
 * BlockNumber, so we can reserve high values of segno for special purposes.
 * We define three:
 * - FORGET_RELATION_FSYNC means to cancel pending fsyncs for a relation
 * - FORGET_DATABASE_FSYNC means to cancel pending fsyncs for a whole database
 * - UNLINK_RELATION_REQUEST is a request to delete the file after the next
 *	 checkpoint.
 *
 * (Handling the FORGET_* requests is a tad slow because the hash table has
 * to be searched linearly, but it doesn't seem worth rethinking the table
 * structure for them.)
 */
void
RememberFsyncRequest(RelFileNode rnode, BlockNumber segno)
{
	Assert(pendingOpsTable);

	if (segno == FORGET_RELATION_FSYNC)
	{
		/* Remove any pending requests for the entire relation */
		HASH_SEQ_STATUS hstat;
		PendingOperationEntry *entry;

		hash_seq_init(&hstat, pendingOpsTable);
		while ((entry = (PendingOperationEntry *) hash_seq_search(&hstat)) != NULL)
		{
			if (RelFileNodeEquals(entry->tag.rnode, rnode))
			{
				/* Okay, cancel this entry */
				entry->canceled = true;
			}
		}
	}
	else if (segno == FORGET_DATABASE_FSYNC)
	{
		/* Remove any pending requests for the entire database */
		HASH_SEQ_STATUS hstat;
		PendingOperationEntry *entry;
		ListCell   *cell, 
				   *prev,
				   *next;

		/* Remove fsync requests */
		hash_seq_init(&hstat, pendingOpsTable);
		while ((entry = (PendingOperationEntry *) hash_seq_search(&hstat)) != NULL)
		{
			if (entry->tag.rnode.dbNode == rnode.dbNode)
			{
				/* Okay, cancel this entry */
				entry->canceled = true;
			}
		}
	
		/* Remove unlink requests */
		prev = NULL;
		for (cell = list_head(pendingUnlinks); cell; cell = next)
		{
			PendingUnlinkEntry *entry = (PendingUnlinkEntry *) lfirst(cell);

			next = lnext(cell);
			if (entry->rnode.dbNode == rnode.dbNode) 
			{
				pendingUnlinks = list_delete_cell(pendingUnlinks, cell, prev);
				pfree(entry);
			}
			else
				prev = cell;
		}
	}
	else if (segno == UNLINK_RELATION_REQUEST)
	{
		/* Unlink request: put it in the linked list */
		MemoryContext oldcxt = MemoryContextSwitchTo(MdCxt);
		PendingUnlinkEntry *entry;

		entry = palloc(sizeof(PendingUnlinkEntry));
		entry->rnode = rnode;
		entry->cycle_ctr = mdckpt_cycle_ctr;

		pendingUnlinks = lappend(pendingUnlinks, entry);

		MemoryContextSwitchTo(oldcxt);
	}
	else
	{
		/* Normal case: enter a request to fsync this segment */
		PendingOperationTag key;
		PendingOperationEntry *entry;
		bool		found;

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

		entry = (PendingOperationEntry *) hash_search(pendingOpsTable,
													  &key,
													  HASH_ENTER,
													  &found);
		/* if new or previously canceled entry, initialize it */
		if (!found || entry->canceled)
		{
			entry->canceled = false;
			entry->cycle_ctr = mdsync_cycle_ctr;
		}

		/*
		 * NB: it's intentional that we don't change cycle_ctr if the entry
		 * already exists.	The fsync request must be treated as old, even
		 * though the new request will be satisfied too by any subsequent
		 * fsync.
		 *
		 * However, if the entry is present but is marked canceled, we should
		 * act just as though it wasn't there.  The only case where this could
		 * happen would be if a file had been deleted, we received but did not
		 * yet act on the cancel request, and the same relfilenode was then
		 * assigned to a new file.	We mustn't lose the new request, but it
		 * should be considered new not old.
		 */
	}
}

/*
 * ForgetRelationFsyncRequests -- forget any fsyncs for a rel
 */
void
ForgetRelationFsyncRequests(RelFileNode rnode)
{
	if (pendingOpsTable)
	{
		/* standalone backend or startup process: fsync state is local */
		RememberFsyncRequest(rnode, FORGET_RELATION_FSYNC);
	}
	else if (IsUnderPostmaster)
	{
		/*
		 * Notify the bgwriter about it.  If we fail to queue the revoke
		 * message, we have to sleep and try again ... ugly, but hopefully
		 * won't happen often.
		 *
		 * XXX should we CHECK_FOR_INTERRUPTS in this loop?  Escaping with an
		 * error would leave the no-longer-used file still present on disk,
		 * which would be bad, so I'm inclined to assume that the bgwriter
		 * will always empty the queue soon.
		 */
		while (!ForwardFsyncRequest(rnode, FORGET_RELATION_FSYNC))
			pg_usleep(10000L);	/* 10 msec seems a good number */

		/*
		 * Note we don't wait for the bgwriter to actually absorb the revoke
		 * message; see mdsync() for the implications.
		 */
	}
}

/*
 * ForgetDatabaseFsyncRequests -- forget any fsyncs and unlinks for a DB
 */
void
ForgetDatabaseFsyncRequests(Oid dbid)
{
	RelFileNode rnode;

	rnode.dbNode = dbid;
	rnode.spcNode = 0;
	rnode.relNode = 0;

	if (pendingOpsTable)
	{
		/* standalone backend or startup process: fsync state is local */
		RememberFsyncRequest(rnode, FORGET_DATABASE_FSYNC);
	}
	else if (IsUnderPostmaster)
	{
		/* see notes in ForgetRelationFsyncRequests */
		while (!ForwardFsyncRequest(rnode, FORGET_DATABASE_FSYNC))
			pg_usleep(10000L);	/* 10 msec seems a good number */
	}
}


/*
 *	_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(reln, v) <= ((BlockNumber) RELSEG_SIZE));

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

/*
 *	_mdfd_getseg() -- Find the segment of the relation holding the
 *		specified block.
 *
 * If the segment doesn't exist, we ereport, return NULL, or create the
 * segment, according to "behavior".  Note: isTemp need only be correct
 * in the EXTENSION_CREATE case.
 */
static MdfdVec *
_mdfd_getseg(SMgrRelation reln, BlockNumber blkno, bool isTemp,
			 ExtensionBehavior behavior)
{
	MdfdVec    *v = mdopen(reln, behavior);

#ifndef LET_OS_MANAGE_FILESIZE
	BlockNumber targetseg;
	BlockNumber nextsegno;

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

	targetseg = blkno / ((BlockNumber) RELSEG_SIZE);
	for (nextsegno = 1; nextsegno <= targetseg; nextsegno++)
	{
		Assert(nextsegno == v->mdfd_segno + 1);

		if (v->mdfd_chain == NULL)
		{
			/*
			 * Normally we will create new segments only if authorized by the
			 * caller (i.e., we are doing mdextend()).	But when doing WAL
			 * recovery, create segments anyway; this allows cases such as
			 * 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.
			 *
			 * We have to maintain the invariant that segments before the last
			 * active segment are of size RELSEG_SIZE; therefore, pad them out
			 * with zeroes if needed.  (This only matters if caller is
			 * extending the relation discontiguously, but that can happen in
			 * hash indexes.)
			 */
			if (behavior == EXTENSION_CREATE || InRecovery)
			{
				if (_mdnblocks(reln, v) < RELSEG_SIZE)
				{
					char	   *zerobuf = palloc0(BLCKSZ);

					mdextend(reln, nextsegno * ((BlockNumber) RELSEG_SIZE) - 1,
							 zerobuf, isTemp);
					pfree(zerobuf);
				}
				v->mdfd_chain = _mdfd_openseg(reln, nextsegno, O_CREAT);
			}
			else
			{
				/* We won't create segment if not existent */
				v->mdfd_chain = _mdfd_openseg(reln, nextsegno, 0);
			}
			if (v->mdfd_chain == NULL)
			{
				if (behavior == EXTENSION_RETURN_NULL &&
					FILE_POSSIBLY_DELETED(errno))
					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(SMgrRelation reln, MdfdVec *seg)
{
	long		len;

	len = FileSeek(seg->mdfd_vfd, 0L, SEEK_END);
	if (len < 0)
		ereport(ERROR,
				(errcode_for_file_access(),
		errmsg("could not seek to end of segment %u of relation %u/%u/%u: %m",
			   seg->mdfd_segno,
			   reln->smgr_rnode.spcNode,
			   reln->smgr_rnode.dbNode,
			   reln->smgr_rnode.relNode)));
	/* note that this calculation will ignore any partial block at EOF */
	return (BlockNumber) (len / BLCKSZ);
}