Return to xfs_export.c CVS log

Up to [Development] / linux-2.6-xfs / fs / xfs / linux-2.6

Radix tree based inode caching

One of the perpetual scaling problems XFS has is indexing
it's incore inodes. We currently uses hashes and the default
hash sizes chosen can only ever be a tradeoff between memory
consumption and the maximum realistic size of the cache.

As a result, anyone who has millions of inodes cached on a
filesystem needs to tunes the size of the cache via the ihashsize
mount option to allow decent scalability with inode cache
operations.

A further problem is the separate inode cluster hash, whose size is
based on the ihashsize but is smaller, and so under certain
conditions (sparse cluster cache population) this can become
a limitation long before the inode hash is causing issues.

The following patchset removes the inode hash and cluster hash
and replaces them with radix trees to avoid the scalability
limitations of the hashes. It also reduces the size of the
inodes by 3 pointers....
Merge of xfs-linux-melb:xfs-kern:29481a by kenmcd.

  Convert xfs inode caches from hashes to radix trees.

/*
 * Copyright (c) 2004-2005 Silicon Graphics, Inc.
 * All Rights Reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
#include "xfs.h"
#include "xfs_types.h"
#include "xfs_inum.h"
#include "xfs_log.h"
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
#include "xfs_export.h"

static struct dentry dotdot = { .d_name.name = "..", .d_name.len = 2, };

/*
 * XFS encodes and decodes the fileid portion of NFS filehandles
 * itself instead of letting the generic NFS code do it.  This
 * allows filesystems with 64 bit inode numbers to be exported.
 *
 * Note that a side effect is that xfs_vget() won't be passed a
 * zero inode/generation pair under normal circumstances.  As
 * however a malicious client could send us such data, the check
 * remains in that code.
 */

STATIC struct dentry *
xfs_fs_decode_fh(
	struct super_block	*sb,
	__u32			*fh,
	int			fh_len,
	int			fileid_type,
	int (*acceptable)(
		void		*context,
		struct dentry	*de),
	void			*context)
{
	xfs_fid2_t		ifid;
	xfs_fid2_t		pfid;
	void			*parent = NULL;
	int			is64 = 0;
	__u32			*p = fh;

#if XFS_BIG_INUMS
	is64 = (fileid_type & XFS_FILEID_TYPE_64FLAG);
	fileid_type &= ~XFS_FILEID_TYPE_64FLAG;
#endif

	/*
	 * Note that we only accept fileids which are long enough
	 * rather than allow the parent generation number to default
	 * to zero.  XFS considers zero a valid generation number not
	 * an invalid/wildcard value.  There's little point printk'ing
	 * a warning here as we don't have the client information
	 * which would make such a warning useful.
	 */
	if (fileid_type > 2 ||
	    fh_len < xfs_fileid_length((fileid_type == 2), is64))
		return NULL;

	p = xfs_fileid_decode_fid2(p, &ifid, is64);

	if (fileid_type == 2) {
		p = xfs_fileid_decode_fid2(p, &pfid, is64);
		parent = &pfid;
	}

	fh = (__u32 *)&ifid;
	return sb->s_export_op->find_exported_dentry(sb, fh, parent, acceptable, context);
}


STATIC int
xfs_fs_encode_fh(
	struct dentry		*dentry,
	__u32			*fh,
	int			*max_len,
	int			connectable)
{
	struct inode		*inode = dentry->d_inode;
	int			type = 1;
	__u32			*p = fh;
	int			len;
	int			is64 = 0;
#if XFS_BIG_INUMS
	bhv_vfs_t		*vfs = vfs_from_sb(inode->i_sb);

	if (!(vfs->vfs_flag & VFS_32BITINODES)) {
		/* filesystem may contain 64bit inode numbers */
		is64 = XFS_FILEID_TYPE_64FLAG;
	}
#endif

	/* Directories don't need their parent encoded, they have ".." */
	if (S_ISDIR(inode->i_mode))
	    connectable = 0;

	/*
	 * Only encode if there is enough space given.  In practice
	 * this means we can't export a filesystem with 64bit inodes
	 * over NFSv2 with the subtree_check export option; the other
	 * seven combinations work.  The real answer is "don't use v2".
	 */
	len = xfs_fileid_length(connectable, is64);
	if (*max_len < len)
		return 255;
	*max_len = len;

	p = xfs_fileid_encode_inode(p, inode, is64);
	if (connectable) {
		spin_lock(&dentry->d_lock);
		p = xfs_fileid_encode_inode(p, dentry->d_parent->d_inode, is64);
		spin_unlock(&dentry->d_lock);
		type = 2;
	}
	BUG_ON((p - fh) != len);
	return type | is64;
}

STATIC struct dentry *
xfs_fs_get_dentry(
	struct super_block	*sb,
	void			*data)
{
	bhv_vnode_t		*vp;
	struct inode		*inode;
	struct dentry		*result;
	bhv_vfs_t		*vfsp = vfs_from_sb(sb);
	int			error;

	error = bhv_vfs_vget(vfsp, &vp, (fid_t *)data);
	if (error || vp == NULL)
		return ERR_PTR(-ESTALE) ;

	inode = vn_to_inode(vp);
	result = d_alloc_anon(inode);
        if (!result) {
		iput(inode);
		return ERR_PTR(-ENOMEM);
	}
	return result;
}

STATIC struct dentry *
xfs_fs_get_parent(
	struct dentry		*child)
{
	int			error;
	bhv_vnode_t		*vp, *cvp;
	struct dentry		*parent;

	cvp = NULL;
	vp = vn_from_inode(child->d_inode);
	error = bhv_vop_lookup(vp, &dotdot, &cvp, 0, NULL, NULL);
	if (unlikely(error))
		return ERR_PTR(-error);

	parent = d_alloc_anon(vn_to_inode(cvp));
	if (unlikely(!parent)) {
		VN_RELE(cvp);
		return ERR_PTR(-ENOMEM);
	}
	return parent;
}

struct export_operations xfs_export_operations = {
	.decode_fh		= xfs_fs_decode_fh,
	.encode_fh		= xfs_fs_encode_fh,
	.get_parent		= xfs_fs_get_parent,
	.get_dentry		= xfs_fs_get_dentry,
};