From: Andreas Dilger <adilger@sun.com>
Subject: Re: [PATCH 1/5] ext4 online defrag header file changes
Date: Tue, 18 Mar 2008 01:08:46 +0800
Message-ID: <20080317170846.GE3542@webber.adilger.int>
References: <200803141202.AA00330@TNESG9526.rs.jp.nec.com>
 <20080314125848.GE3542@webber.adilger.int> <47DE0F8D.2060106@rs.jp.nec.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7BIT
Cc: cmm@us.ibm.com, tytso@mit.edu, linux-ext4@vger.kernel.org,
	linux-fsdevel@vger.kernel.org
To: Akira Fujita <a-fujita@rs.jp.nec.com>
Return-path: <linux-fsdevel-owner@vger.kernel.org>
In-reply-to: <47DE0F8D.2060106@rs.jp.nec.com>
Content-disposition: inline
Sender: linux-fsdevel-owner@vger.kernel.org
List-Id: linux-ext4.vger.kernel.org

On Mar 17, 2008  15:28 +0900, Akira Fujita wrote:
> Hi Andreas,
>>> How about renaming this structure from ext4_extent_data to
>>> ext4_extent_info which used the extent in-core structure
>>> and put it into ext4_fs_extents.h? Because in-core super_block is 
>>> ext4_sb_info.
>>>
>>> /*
>>>  * This is the extent in-core structure.
>>>  */
>>> struct ext4_extent_info {
>>> 	ext4_lblk_t	ee_block;	/* first logical block extent covers */
>>> 	int		ee_len;		/* number of blocks covered by extent */
>>> 	ext4_fsblk_t	ee_start;	/* first physical block extent covers */
>>> };          
>>
>> Would it be possible to use the FIEMAP interface to get the file mapping
>> data to userspace?  That would avoid the need to have a second ioctl
>> with nearly the same data.
>   
> Reducing ioctl makes sense.
> But EXT4_EXTENTS_INFO needs to get the file mapping data
> with inode number not file name.
> So we should keep ext4_extents_info structure
> or put ino entry in fiemap structure.

We introduced a special ioctl for this which would allow root to perform
ioctl() on any inode by inode number.  We call this EXT4_IOC_WRAPPER, and
it is called on the root inode and specifies the inode number and real
ioctl number before the rest of the ioctl data.

struct ext4_ioctl_wrapper {
	__u32 ioctl_cmd;
	__u32 padding;
	__u64 ioctl_ino;
	char ioctl_data[0];
};

More patch is available, but it isn't in a self-contained patch.

Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.