namei.c

linux 内核源代码

out_commit:
	ocfs2_commit_trans(osb, handle);
out_unlock_inode:
	ocfs2_meta_unlock(inode, 1);

out:
	ocfs2_meta_unlock(dir, 1);

	if (de_bh)
		brelse(de_bh);
	if (fe_bh)
		brelse(fe_bh);
	if (parent_fe_bh)
		brelse(parent_fe_bh);

	mlog_exit(err);

	return err;
}

/*
 * Takes and drops an exclusive lock on the given dentry. This will
 * force other nodes to drop it.
 */
static int ocfs2_remote_dentry_delete(struct dentry *dentry)
{
	int ret;

	ret = ocfs2_dentry_lock(dentry, 1);
	if (ret)
		mlog_errno(ret);
	else
		ocfs2_dentry_unlock(dentry, 1);

	return ret;
}

static inline int inode_is_unlinkable(struct inode *inode)
{
	if (S_ISDIR(inode->i_mode)) {
		if (inode->i_nlink == 2)
			return 1;
		return 0;
	}

	if (inode->i_nlink == 1)
		return 1;
	return 0;
}

static int ocfs2_unlink(struct inode *dir,
			struct dentry *dentry)
{
	int status;
	int child_locked = 0;
	struct inode *inode = dentry->d_inode;
	struct inode *orphan_dir = NULL;
	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
	u64 blkno;
	struct ocfs2_dinode *fe = NULL;
	struct buffer_head *fe_bh = NULL;
	struct buffer_head *parent_node_bh = NULL;
	handle_t *handle = NULL;
	struct ocfs2_dir_entry *dirent = NULL;
	struct buffer_head *dirent_bh = NULL;
	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
	struct buffer_head *orphan_entry_bh = NULL;

	mlog_entry("(0x%p, 0x%p, '%.*s')\n", dir, dentry,
		   dentry->d_name.len, dentry->d_name.name);

	BUG_ON(dentry->d_parent->d_inode != dir);

	mlog(0, "ino = %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno);

	if (inode == osb->root_inode) {
		mlog(0, "Cannot delete the root directory\n");
		return -EPERM;
	}

	status = ocfs2_meta_lock(dir, &parent_node_bh, 1);
	if (status < 0) {
		if (status != -ENOENT)
			mlog_errno(status);
		return status;
	}

	status = ocfs2_find_files_on_disk(dentry->d_name.name,
					  dentry->d_name.len, &blkno,
					  dir, &dirent_bh, &dirent);
	if (status < 0) {
		if (status != -ENOENT)
			mlog_errno(status);
		goto leave;
	}

	if (OCFS2_I(inode)->ip_blkno != blkno) {
		status = -ENOENT;

		mlog(0, "ip_blkno %llu != dirent blkno %llu ip_flags = %x\n",
		     (unsigned long long)OCFS2_I(inode)->ip_blkno,
		     (unsigned long long)blkno, OCFS2_I(inode)->ip_flags);
		goto leave;
	}

	status = ocfs2_meta_lock(inode, &fe_bh, 1);
	if (status < 0) {
		if (status != -ENOENT)
			mlog_errno(status);
		goto leave;
	}
	child_locked = 1;

	if (S_ISDIR(inode->i_mode)) {
	       	if (!ocfs2_empty_dir(inode)) {
			status = -ENOTEMPTY;
			goto leave;
		} else if (inode->i_nlink != 2) {
			status = -ENOTEMPTY;
			goto leave;
		}
	}

	status = ocfs2_remote_dentry_delete(dentry);
	if (status < 0) {
		/* This vote should succeed under all normal
		 * circumstances. */
		mlog_errno(status);
		goto leave;
	}

	if (inode_is_unlinkable(inode)) {
		status = ocfs2_prepare_orphan_dir(osb, &orphan_dir, inode,
						  orphan_name,
						  &orphan_entry_bh);
		if (status < 0) {
			mlog_errno(status);
			goto leave;
		}
	}

	handle = ocfs2_start_trans(osb, OCFS2_UNLINK_CREDITS);
	if (IS_ERR(handle)) {
		status = PTR_ERR(handle);
		handle = NULL;
		mlog_errno(status);
		goto leave;
	}

	status = ocfs2_journal_access(handle, inode, fe_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (status < 0) {
		mlog_errno(status);
		goto leave;
	}

	fe = (struct ocfs2_dinode *) fe_bh->b_data;

	if (inode_is_unlinkable(inode)) {
		status = ocfs2_orphan_add(osb, handle, inode, fe, orphan_name,
					  orphan_entry_bh, orphan_dir);
		if (status < 0) {
			mlog_errno(status);
			goto leave;
		}
	}

	/* delete the name from the parent dir */
	status = ocfs2_delete_entry(handle, dir, dirent, dirent_bh);
	if (status < 0) {
		mlog_errno(status);
		goto leave;
	}

	if (S_ISDIR(inode->i_mode))
		drop_nlink(inode);
	drop_nlink(inode);
	fe->i_links_count = cpu_to_le16(inode->i_nlink);

	status = ocfs2_journal_dirty(handle, fe_bh);
	if (status < 0) {
		mlog_errno(status);
		goto leave;
	}

	dir->i_ctime = dir->i_mtime = CURRENT_TIME;
	if (S_ISDIR(inode->i_mode))
		drop_nlink(dir);

	status = ocfs2_mark_inode_dirty(handle, dir, parent_node_bh);
	if (status < 0) {
		mlog_errno(status);
		if (S_ISDIR(inode->i_mode))
			inc_nlink(dir);
	}

leave:
	if (handle)
		ocfs2_commit_trans(osb, handle);

	if (child_locked)
		ocfs2_meta_unlock(inode, 1);

	ocfs2_meta_unlock(dir, 1);

	if (orphan_dir) {
		/* This was locked for us in ocfs2_prepare_orphan_dir() */
		ocfs2_meta_unlock(orphan_dir, 1);
		mutex_unlock(&orphan_dir->i_mutex);
		iput(orphan_dir);
	}

	if (fe_bh)
		brelse(fe_bh);

	if (dirent_bh)
		brelse(dirent_bh);

	if (parent_node_bh)
		brelse(parent_node_bh);

	if (orphan_entry_bh)
		brelse(orphan_entry_bh);

	mlog_exit(status);

	return status;
}

/*
 * The only place this should be used is rename!
 * if they have the same id, then the 1st one is the only one locked.
 */
static int ocfs2_double_lock(struct ocfs2_super *osb,
			     struct buffer_head **bh1,
			     struct inode *inode1,
			     struct buffer_head **bh2,
			     struct inode *inode2)
{
	int status;
	struct ocfs2_inode_info *oi1 = OCFS2_I(inode1);
	struct ocfs2_inode_info *oi2 = OCFS2_I(inode2);
	struct buffer_head **tmpbh;
	struct inode *tmpinode;

	mlog_entry("(inode1 = %llu, inode2 = %llu)\n",
		   (unsigned long long)oi1->ip_blkno,
		   (unsigned long long)oi2->ip_blkno);

	if (*bh1)
		*bh1 = NULL;
	if (*bh2)
		*bh2 = NULL;

	/* we always want to lock the one with the lower lockid first. */
	if (oi1->ip_blkno != oi2->ip_blkno) {
		if (oi1->ip_blkno < oi2->ip_blkno) {
			/* switch id1 and id2 around */
			mlog(0, "switching them around...\n");
			tmpbh = bh2;
			bh2 = bh1;
			bh1 = tmpbh;

			tmpinode = inode2;
			inode2 = inode1;
			inode1 = tmpinode;
		}
		/* lock id2 */
		status = ocfs2_meta_lock(inode2, bh2, 1);
		if (status < 0) {
			if (status != -ENOENT)
				mlog_errno(status);
			goto bail;
		}
	}

	/* lock id1 */
	status = ocfs2_meta_lock(inode1, bh1, 1);
	if (status < 0) {
		/*
		 * An error return must mean that no cluster locks
		 * were held on function exit.
		 */
		if (oi1->ip_blkno != oi2->ip_blkno)
			ocfs2_meta_unlock(inode2, 1);

		if (status != -ENOENT)
			mlog_errno(status);
	}

bail:
	mlog_exit(status);
	return status;
}

static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2)
{
	ocfs2_meta_unlock(inode1, 1);

	if (inode1 != inode2)
		ocfs2_meta_unlock(inode2, 1);
}

static int ocfs2_rename(struct inode *old_dir,
			struct dentry *old_dentry,
			struct inode *new_dir,
			struct dentry *new_dentry)
{
	int status = 0, rename_lock = 0, parents_locked = 0;
	int old_child_locked = 0, new_child_locked = 0;
	struct inode *old_inode = old_dentry->d_inode;
	struct inode *new_inode = new_dentry->d_inode;
	struct inode *orphan_dir = NULL;
	struct ocfs2_dinode *newfe = NULL;
	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
	struct buffer_head *orphan_entry_bh = NULL;
	struct buffer_head *newfe_bh = NULL;
	struct buffer_head *old_inode_bh = NULL;
	struct buffer_head *insert_entry_bh = NULL;
	struct ocfs2_super *osb = NULL;
	u64 newfe_blkno, old_de_ino;
	handle_t *handle = NULL;
	struct buffer_head *old_dir_bh = NULL;
	struct buffer_head *new_dir_bh = NULL;
	struct ocfs2_dir_entry *old_inode_dot_dot_de = NULL, *old_de = NULL,
		*new_de = NULL;
	struct buffer_head *new_de_bh = NULL, *old_de_bh = NULL; // bhs for above
	struct buffer_head *old_inode_de_bh = NULL; // if old_dentry is a dir,
						    // this is the 1st dirent bh
	nlink_t old_dir_nlink = old_dir->i_nlink;
	struct ocfs2_dinode *old_di;

	/* At some point it might be nice to break this function up a
	 * bit. */

	mlog_entry("(0x%p, 0x%p, 0x%p, 0x%p, from='%.*s' to='%.*s')\n",
		   old_dir, old_dentry, new_dir, new_dentry,
		   old_dentry->d_name.len, old_dentry->d_name.name,
		   new_dentry->d_name.len, new_dentry->d_name.name);

	osb = OCFS2_SB(old_dir->i_sb);

	if (new_inode) {
		if (!igrab(new_inode))
			BUG();
	}

	/* Assume a directory hierarchy thusly:
	 * a/b/c
	 * a/d
	 * a,b,c, and d are all directories.
	 *
	 * from cwd of 'a' on both nodes:
	 * node1: mv b/c d
	 * node2: mv d   b/c
	 *
	 * And that's why, just like the VFS, we need a file system
	 * rename lock. */
	if (old_dentry != new_dentry) {
		status = ocfs2_rename_lock(osb);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
		rename_lock = 1;
	}

	/* if old and new are the same, this'll just do one lock. */
	status = ocfs2_double_lock(osb, &old_dir_bh, old_dir,
				   &new_dir_bh, new_dir);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	parents_locked = 1;

	/* make sure both dirs have bhs
	 * get an extra ref on old_dir_bh if old==new */
	if (!new_dir_bh) {
		if (old_dir_bh) {
			new_dir_bh = old_dir_bh;
			get_bh(new_dir_bh);
		} else {
			mlog(ML_ERROR, "no old_dir_bh!\n");
			status = -EIO;
			goto bail;
		}
	}

	/*
	 * Aside from allowing a meta data update, the locking here
	 * also ensures that the vote thread on other nodes won't have
	 * to concurrently downconvert the inode and the dentry locks.
	 */
	status = ocfs2_meta_lock(old_inode, &old_inode_bh, 1);
	if (status < 0) {
		if (status != -ENOENT)
			mlog_errno(status);
		goto bail;
	}
	old_child_locked = 1;

	status = ocfs2_remote_dentry_delete(old_dentry);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	if (S_ISDIR(old_inode->i_mode)) {
		u64 old_inode_parent;

		status = ocfs2_find_files_on_disk("..", 2, &old_inode_parent,
						  old_inode, &old_inode_de_bh,
						  &old_inode_dot_dot_de);
		if (status) {
			status = -EIO;
			goto bail;
		}

		if (old_inode_parent != OCFS2_I(old_dir)->ip_blkno) {
			status = -EIO;
			goto bail;
		}

		if (!new_inode && new_dir != old_dir &&
		    new_dir->i_nlink >= OCFS2_LINK_MAX) {
			status = -EMLINK;
			goto bail;
		}
	}

	status = ocfs2_lookup_ino_from_name(old_dir, old_dentry->d_name.name,
					    old_dentry->d_name.len,
					    &old_de_ino);
	if (status) {
		status = -ENOENT;
		goto bail;
	}

	/*
	 *  Check for inode number is _not_ due to possible IO errors.
	 *  We might rmdir the source, keep it as pwd of some process
	 *  and merrily kill the link to whatever was created under the
	 *  same name. Goodbye sticky bit ;-<
	 */
	if (old_de_ino != OCFS2_I(old_inode)->ip_blkno) {
		status = -ENOENT;
		goto bail;
	}

	/* check if the target already exists (in which case we need
	 * to delete it */
	status = ocfs2_find_files_on_disk(new_dentry->d_name.name,
					  new_dentry->d_name.len,
					  &newfe_blkno, new_dir, &new_de_bh,
					  &new_de);
	/* The only error we allow here is -ENOENT because the new
	 * file not existing is perfectly valid. */
	if ((status < 0) && (status != -ENOENT)) {
		/* If we cannot find the file specified we should just */
		/* return the error... */
		mlog_errno(status);
		goto bail;
	}

	if (!new_de && new_inode) {
		/*
		 * Target was unlinked by another node while we were
		 * waiting to get to ocfs2_rename(). There isn't
		 * anything we can do here to help the situation, so
		 * bubble up the appropriate error.
		 */
		status = -ENOENT;
		goto bail;
	}

	/* In case we need to overwrite an existing file, we blow it
	 * away first */
	if (new_de) {
		/* VFS didn't think there existed an inode here, but
		 * someone else in the cluster must have raced our
		 * rename to create one. Today we error cleanly, in
		 * the future we should consider calling iget to build
		 * a new struct inode for this entry. */
		if (!new_inode) {
			status = -EACCES;

			mlog(0, "We found an inode for name %.*s but VFS "
			     "didn't give us one.\n", new_dentry->d_name.len,
			     new_dentry->d_name.name);
			goto bail;
		}

		if (OCFS2_I(new_inode)->ip_blkno != newfe_blkno) {
			status = -EACCES;

			mlog(0, "Inode %llu and dir %llu disagree. flags = %x\n",
			     (unsigned long long)OCFS2_I(new_inode)->ip_blkno,
			     (unsigned long long)newfe_blkno,
			     OCFS2_I(new_inode)->ip_flags);
			goto bail;
		}

		status = ocfs2_meta_lock(new_inode, &newfe_bh, 1);
		if (status < 0) {
			if (status != -ENOENT)
				mlog_errno(status);
			goto bail;
		}
		new_child_locked = 1;

		status = ocfs2_remote_dentry_delete(new_dentry);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}

		newfe = (struct ocfs2_dinode *) newfe_bh->b_data;

		mlog(0, "aha rename over existing... new_de=%p new_blkno=%llu "
		     "newfebh=%p bhblocknr=%llu\n", new_de,
		     (unsigned long long)newfe_blkno, newfe_bh, newfe_bh ?
		     (unsigned long long)newfe_bh->b_blocknr : 0ULL);

		if (S_ISDIR(new_inode->i_mode) || (new_inode->i_nlink == 1)) {
			status = ocfs2_prepare_orphan_dir(osb, &orphan_dir,
							  new_inode,
							  orphan_name,
							  &orphan_entry_bh);
			if (status < 0) {
				mlog_errno(status);
				goto bail;
			}
		}
	} else {
		BUG_ON(new_dentry->d_parent->d_inode != new_dir);

		status = ocfs2_check_dir_for_entry(new_dir,
						   new_dentry->d_name.name,
						   new_dentry->d_name.len);
		if (status)
			goto bail;

		status = ocfs2_prepare_dir_for_insert(osb, new_dir, new_dir_bh,
						      new_dentry->d_name.name,
						      new_dentry->d_name.len,
						      &insert_entry_bh);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
	}

	handle = ocfs2_start_trans(osb, OCFS2_RENAME_CREDITS);
	if (IS_ERR(handle)) {
		status = PTR_ERR(handle);
		handle = NULL;
		mlog_errno(status);
		goto bail;
	}

	if (new_de) {
		if (S_ISDIR(new_inode->i_mode)) {
			if (!ocfs2_empty_dir(new_inode) ||
			    new_inode->i_nlink != 2) {
				status = -ENOTEMPTY;
				goto bail;
			}
		}
		status = ocfs2_journal_access(handle, new_inode, newfe_bh,
					      OCFS2_JOURNAL_ACCESS_WRITE);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}

		if (S_ISDIR(new_inode->i_mode) ||
		    (newfe->i_links_count == cpu_to_le16(1))){
			status = ocfs2_orphan_add(osb, handle, new_inode,
						  newfe, orphan_name,
						  orphan_entry_bh, orphan_dir);
			if (status < 0) {
				mlog_errno(status);
				goto bail;
			}
		}

		/* change the dirent to point to the correct inode */
		status = ocfs2_update_entry(new_dir, handle, new_de_bh,
					    new_de, old_inode);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
		new_dir->i_version++;

		if (S_ISDIR(new_inode->i_mode))
			newfe->i_links_count = 0;
		else
			le16_add_cpu(&newfe->i_links_count, -1);

		status = ocfs2_journal_dirty(handle, newfe_bh);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
	} else {
		/* if the name was not found in new_dir, add it now */
		status = ocfs2_add_entry(handle, new_dentry, old_inode,
					 OCFS2_I(old_inode)->ip_blkno,
					 new_dir_bh, insert_entry_bh);
	}

	old_inode->i_ctime = CURRENT_TIME;
	mark_inode_dirty(old_inode);

	status = ocfs2_journal_access(handle, old_inode, old_inode_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (status >= 0) {
		old_di = (struct ocfs2_dinode *) old_inode_bh->b_data;

		old_di->i_ctime = cpu_to_le64(old_inode->i_ctime.tv_sec);
		old_di->i_ctime_nsec = cpu_to_le32(old_inode->i_ctime.tv_nsec);

		status = ocfs2_journal_dirty(handle, old_inode_bh);
		if (status < 0)
			mlog_errno(status);
	} else
		mlog_errno(status);

	/*
	 * Now that the name has been added to new_dir, remove the old name.
	 *
	 * We don't keep any directory entry context around until now
	 * because the insert might have changed the type of directory
	 * we're dealing with.
	 */
	old_de_bh = ocfs2_find_entry(old_dentry->d_name.name,
				     old_dentry->d_name.len,
				     old_dir, &old_de);