namei.c

linux 内核源代码

/*
 *   Copyright (C) International Business Machines Corp., 2000-2004
 *   Portions Copyright (C) Christoph Hellwig, 2001-2002
 *
 *   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; either version 2 of the License, or
 *   (at your option) any later version.
 *
 *   This program is distributed in the hope that it will 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 to the Free Software
 *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
 */

#include <linux/fs.h>
#include <linux/ctype.h>
#include <linux/quotaops.h>
#include <linux/exportfs.h>
#include "jfs_incore.h"
#include "jfs_superblock.h"
#include "jfs_inode.h"
#include "jfs_dinode.h"
#include "jfs_dmap.h"
#include "jfs_unicode.h"
#include "jfs_metapage.h"
#include "jfs_xattr.h"
#include "jfs_acl.h"
#include "jfs_debug.h"

/*
 * forward references
 */
struct dentry_operations jfs_ci_dentry_operations;

static s64 commitZeroLink(tid_t, struct inode *);

/*
 * NAME:	free_ea_wmap(inode)
 *
 * FUNCTION:	free uncommitted extended attributes from working map
 *
 */
static inline void free_ea_wmap(struct inode *inode)
{
	dxd_t *ea = &JFS_IP(inode)->ea;

	if (ea->flag & DXD_EXTENT) {
		/* free EA pages from cache */
		invalidate_dxd_metapages(inode, *ea);
		dbFree(inode, addressDXD(ea), lengthDXD(ea));
	}
	ea->flag = 0;
}

/*
 * NAME:	jfs_create(dip, dentry, mode)
 *
 * FUNCTION:	create a regular file in the parent directory <dip>
 *		with name = <from dentry> and mode = <mode>
 *
 * PARAMETER:	dip	- parent directory vnode
 *		dentry	- dentry of new file
 *		mode	- create mode (rwxrwxrwx).
 *		nd- nd struct
 *
 * RETURN:	Errors from subroutines
 *
 */
static int jfs_create(struct inode *dip, struct dentry *dentry, int mode,
		struct nameidata *nd)
{
	int rc = 0;
	tid_t tid;		/* transaction id */
	struct inode *ip = NULL;	/* child directory inode */
	ino_t ino;
	struct component_name dname;	/* child directory name */
	struct btstack btstack;
	struct inode *iplist[2];
	struct tblock *tblk;

	jfs_info("jfs_create: dip:0x%p name:%s", dip, dentry->d_name.name);

	/*
	 * search parent directory for entry/freespace
	 * (dtSearch() returns parent directory page pinned)
	 */
	if ((rc = get_UCSname(&dname, dentry)))
		goto out1;

	/*
	 * Either iAlloc() or txBegin() may block.  Deadlock can occur if we
	 * block there while holding dtree page, so we allocate the inode &
	 * begin the transaction before we search the directory.
	 */
	ip = ialloc(dip, mode);
	if (IS_ERR(ip)) {
		rc = PTR_ERR(ip);
		goto out2;
	}

	tid = txBegin(dip->i_sb, 0);

	mutex_lock_nested(&JFS_IP(dip)->commit_mutex, COMMIT_MUTEX_PARENT);
	mutex_lock_nested(&JFS_IP(ip)->commit_mutex, COMMIT_MUTEX_CHILD);

	rc = jfs_init_acl(tid, ip, dip);
	if (rc)
		goto out3;

	rc = jfs_init_security(tid, ip, dip);
	if (rc) {
		txAbort(tid, 0);
		goto out3;
	}

	if ((rc = dtSearch(dip, &dname, &ino, &btstack, JFS_CREATE))) {
		jfs_err("jfs_create: dtSearch returned %d", rc);
		txAbort(tid, 0);
		goto out3;
	}

	tblk = tid_to_tblock(tid);
	tblk->xflag |= COMMIT_CREATE;
	tblk->ino = ip->i_ino;
	tblk->u.ixpxd = JFS_IP(ip)->ixpxd;

	iplist[0] = dip;
	iplist[1] = ip;

	/*
	 * initialize the child XAD tree root in-line in inode
	 */
	xtInitRoot(tid, ip);

	/*
	 * create entry in parent directory for child directory
	 * (dtInsert() releases parent directory page)
	 */
	ino = ip->i_ino;
	if ((rc = dtInsert(tid, dip, &dname, &ino, &btstack))) {
		if (rc == -EIO) {
			jfs_err("jfs_create: dtInsert returned -EIO");
			txAbort(tid, 1);	/* Marks Filesystem dirty */
		} else
			txAbort(tid, 0);	/* Filesystem full */
		goto out3;
	}

	ip->i_op = &jfs_file_inode_operations;
	ip->i_fop = &jfs_file_operations;
	ip->i_mapping->a_ops = &jfs_aops;

	insert_inode_hash(ip);
	mark_inode_dirty(ip);

	dip->i_ctime = dip->i_mtime = CURRENT_TIME;

	mark_inode_dirty(dip);

	rc = txCommit(tid, 2, &iplist[0], 0);

      out3:
	txEnd(tid);
	mutex_unlock(&JFS_IP(ip)->commit_mutex);
	mutex_unlock(&JFS_IP(dip)->commit_mutex);
	if (rc) {
		free_ea_wmap(ip);
		ip->i_nlink = 0;
		iput(ip);
	} else
		d_instantiate(dentry, ip);

      out2:
	free_UCSname(&dname);

      out1:

	jfs_info("jfs_create: rc:%d", rc);
	return rc;
}


/*
 * NAME:	jfs_mkdir(dip, dentry, mode)
 *
 * FUNCTION:	create a child directory in the parent directory <dip>
 *		with name = <from dentry> and mode = <mode>
 *
 * PARAMETER:	dip	- parent directory vnode
 *		dentry	- dentry of child directory
 *		mode	- create mode (rwxrwxrwx).
 *
 * RETURN:	Errors from subroutines
 *
 * note:
 * EACCESS: user needs search+write permission on the parent directory
 */
static int jfs_mkdir(struct inode *dip, struct dentry *dentry, int mode)
{
	int rc = 0;
	tid_t tid;		/* transaction id */
	struct inode *ip = NULL;	/* child directory inode */
	ino_t ino;
	struct component_name dname;	/* child directory name */
	struct btstack btstack;
	struct inode *iplist[2];
	struct tblock *tblk;

	jfs_info("jfs_mkdir: dip:0x%p name:%s", dip, dentry->d_name.name);

	/* link count overflow on parent directory ? */
	if (dip->i_nlink == JFS_LINK_MAX) {
		rc = -EMLINK;
		goto out1;
	}

	/*
	 * search parent directory for entry/freespace
	 * (dtSearch() returns parent directory page pinned)
	 */
	if ((rc = get_UCSname(&dname, dentry)))
		goto out1;

	/*
	 * Either iAlloc() or txBegin() may block.  Deadlock can occur if we
	 * block there while holding dtree page, so we allocate the inode &
	 * begin the transaction before we search the directory.
	 */
	ip = ialloc(dip, S_IFDIR | mode);
	if (IS_ERR(ip)) {
		rc = PTR_ERR(ip);
		goto out2;
	}

	tid = txBegin(dip->i_sb, 0);

	mutex_lock_nested(&JFS_IP(dip)->commit_mutex, COMMIT_MUTEX_PARENT);
	mutex_lock_nested(&JFS_IP(ip)->commit_mutex, COMMIT_MUTEX_CHILD);

	rc = jfs_init_acl(tid, ip, dip);
	if (rc)
		goto out3;

	rc = jfs_init_security(tid, ip, dip);
	if (rc) {
		txAbort(tid, 0);
		goto out3;
	}

	if ((rc = dtSearch(dip, &dname, &ino, &btstack, JFS_CREATE))) {
		jfs_err("jfs_mkdir: dtSearch returned %d", rc);
		txAbort(tid, 0);
		goto out3;
	}

	tblk = tid_to_tblock(tid);
	tblk->xflag |= COMMIT_CREATE;
	tblk->ino = ip->i_ino;
	tblk->u.ixpxd = JFS_IP(ip)->ixpxd;

	iplist[0] = dip;
	iplist[1] = ip;

	/*
	 * initialize the child directory in-line in inode
	 */
	dtInitRoot(tid, ip, dip->i_ino);

	/*
	 * create entry in parent directory for child directory
	 * (dtInsert() releases parent directory page)
	 */
	ino = ip->i_ino;
	if ((rc = dtInsert(tid, dip, &dname, &ino, &btstack))) {
		if (rc == -EIO) {
			jfs_err("jfs_mkdir: dtInsert returned -EIO");
			txAbort(tid, 1);	/* Marks Filesystem dirty */
		} else
			txAbort(tid, 0);	/* Filesystem full */
		goto out3;
	}

	ip->i_nlink = 2;	/* for '.' */
	ip->i_op = &jfs_dir_inode_operations;
	ip->i_fop = &jfs_dir_operations;

	insert_inode_hash(ip);
	mark_inode_dirty(ip);

	/* update parent directory inode */
	inc_nlink(dip);		/* for '..' from child directory */
	dip->i_ctime = dip->i_mtime = CURRENT_TIME;
	mark_inode_dirty(dip);

	rc = txCommit(tid, 2, &iplist[0], 0);

      out3:
	txEnd(tid);
	mutex_unlock(&JFS_IP(ip)->commit_mutex);
	mutex_unlock(&JFS_IP(dip)->commit_mutex);
	if (rc) {
		free_ea_wmap(ip);
		ip->i_nlink = 0;
		iput(ip);
	} else
		d_instantiate(dentry, ip);

      out2:
	free_UCSname(&dname);


      out1:

	jfs_info("jfs_mkdir: rc:%d", rc);
	return rc;
}

/*
 * NAME:	jfs_rmdir(dip, dentry)
 *
 * FUNCTION:	remove a link to child directory
 *
 * PARAMETER:	dip	- parent inode
 *		dentry	- child directory dentry
 *
 * RETURN:	-EINVAL	- if name is . or ..
 *		-EINVAL - if . or .. exist but are invalid.
 *		errors from subroutines
 *
 * note:
 * if other threads have the directory open when the last link
 * is removed, the "." and ".." entries, if present, are removed before
 * rmdir() returns and no new entries may be created in the directory,
 * but the directory is not removed until the last reference to
 * the directory is released (cf.unlink() of regular file).
 */
static int jfs_rmdir(struct inode *dip, struct dentry *dentry)
{
	int rc;
	tid_t tid;		/* transaction id */
	struct inode *ip = dentry->d_inode;
	ino_t ino;
	struct component_name dname;
	struct inode *iplist[2];
	struct tblock *tblk;

	jfs_info("jfs_rmdir: dip:0x%p name:%s", dip, dentry->d_name.name);

	/* Init inode for quota operations. */
	DQUOT_INIT(ip);

	/* directory must be empty to be removed */
	if (!dtEmpty(ip)) {
		rc = -ENOTEMPTY;
		goto out;
	}

	if ((rc = get_UCSname(&dname, dentry))) {
		goto out;
	}

	tid = txBegin(dip->i_sb, 0);

	mutex_lock_nested(&JFS_IP(dip)->commit_mutex, COMMIT_MUTEX_PARENT);
	mutex_lock_nested(&JFS_IP(ip)->commit_mutex, COMMIT_MUTEX_CHILD);

	iplist[0] = dip;
	iplist[1] = ip;

	tblk = tid_to_tblock(tid);
	tblk->xflag |= COMMIT_DELETE;
	tblk->u.ip = ip;

	/*
	 * delete the entry of target directory from parent directory
	 */
	ino = ip->i_ino;
	if ((rc = dtDelete(tid, dip, &dname, &ino, JFS_REMOVE))) {
		jfs_err("jfs_rmdir: dtDelete returned %d", rc);
		if (rc == -EIO)
			txAbort(tid, 1);
		txEnd(tid);
		mutex_unlock(&JFS_IP(ip)->commit_mutex);
		mutex_unlock(&JFS_IP(dip)->commit_mutex);

		goto out2;
	}

	/* update parent directory's link count corresponding
	 * to ".." entry of the target directory deleted
	 */
	dip->i_ctime = dip->i_mtime = CURRENT_TIME;
	inode_dec_link_count(dip);

	/*
	 * OS/2 could have created EA and/or ACL
	 */
	/* free EA from both persistent and working map */
	if (JFS_IP(ip)->ea.flag & DXD_EXTENT) {
		/* free EA pages */
		txEA(tid, ip, &JFS_IP(ip)->ea, NULL);
	}
	JFS_IP(ip)->ea.flag = 0;

	/* free ACL from both persistent and working map */
	if (JFS_IP(ip)->acl.flag & DXD_EXTENT) {
		/* free ACL pages */
		txEA(tid, ip, &JFS_IP(ip)->acl, NULL);
	}
	JFS_IP(ip)->acl.flag = 0;

	/* mark the target directory as deleted */
	clear_nlink(ip);
	mark_inode_dirty(ip);

	rc = txCommit(tid, 2, &iplist[0], 0);

	txEnd(tid);

	mutex_unlock(&JFS_IP(ip)->commit_mutex);
	mutex_unlock(&JFS_IP(dip)->commit_mutex);

	/*
	 * Truncating the directory index table is not guaranteed.  It
	 * may need to be done iteratively
	 */
	if (test_cflag(COMMIT_Stale, dip)) {
		if (dip->i_size > 1)
			jfs_truncate_nolock(dip, 0);

		clear_cflag(COMMIT_Stale, dip);
	}

      out2:
	free_UCSname(&dname);

      out:
	jfs_info("jfs_rmdir: rc:%d", rc);
	return rc;
}

/*
 * NAME:	jfs_unlink(dip, dentry)
 *
 * FUNCTION:	remove a link to object <vp> named by <name>
 *		from parent directory <dvp>
 *
 * PARAMETER:	dip	- inode of parent directory
 *		dentry	- dentry of object to be removed
 *
 * RETURN:	errors from subroutines
 *
 * note:
 * temporary file: if one or more processes have the file open
 * when the last link is removed, the link will be removed before
 * unlink() returns, but the removal of the file contents will be
 * postponed until all references to the files are closed.
 *
 * JFS does NOT support unlink() on directories.
 *
 */
static int jfs_unlink(struct inode *dip, struct dentry *dentry)
{
	int rc;
	tid_t tid;		/* transaction id */
	struct inode *ip = dentry->d_inode;
	ino_t ino;
	struct component_name dname;	/* object name */
	struct inode *iplist[2];
	struct tblock *tblk;
	s64 new_size = 0;
	int commit_flag;

	jfs_info("jfs_unlink: dip:0x%p name:%s", dip, dentry->d_name.name);

	/* Init inode for quota operations. */
	DQUOT_INIT(ip);

	if ((rc = get_UCSname(&dname, dentry)))
		goto out;

	IWRITE_LOCK(ip, RDWRLOCK_NORMAL);

	tid = txBegin(dip->i_sb, 0);

	mutex_lock_nested(&JFS_IP(dip)->commit_mutex, COMMIT_MUTEX_PARENT);
	mutex_lock_nested(&JFS_IP(ip)->commit_mutex, COMMIT_MUTEX_CHILD);

	iplist[0] = dip;
	iplist[1] = ip;

	/*
	 * delete the entry of target file from parent directory
	 */
	ino = ip->i_ino;
	if ((rc = dtDelete(tid, dip, &dname, &ino, JFS_REMOVE))) {
		jfs_err("jfs_unlink: dtDelete returned %d", rc);
		if (rc == -EIO)
			txAbort(tid, 1);	/* Marks FS Dirty */
		txEnd(tid);
		mutex_unlock(&JFS_IP(ip)->commit_mutex);
		mutex_unlock(&JFS_IP(dip)->commit_mutex);
		IWRITE_UNLOCK(ip);
		goto out1;
	}

	ASSERT(ip->i_nlink);

	ip->i_ctime = dip->i_ctime = dip->i_mtime = CURRENT_TIME;
	mark_inode_dirty(dip);

	/* update target's inode */
	inode_dec_link_count(ip);

	/*
	 *	commit zero link count object
	 */
	if (ip->i_nlink == 0) {
		assert(!test_cflag(COMMIT_Nolink, ip));
		/* free block resources */
		if ((new_size = commitZeroLink(tid, ip)) < 0) {
			txAbort(tid, 1);	/* Marks FS Dirty */
			txEnd(tid);
			mutex_unlock(&JFS_IP(ip)->commit_mutex);
			mutex_unlock(&JFS_IP(dip)->commit_mutex);
			IWRITE_UNLOCK(ip);
			rc = new_size;
			goto out1;
		}
		tblk = tid_to_tblock(tid);
		tblk->xflag |= COMMIT_DELETE;
		tblk->u.ip = ip;