From mboxrd@z Thu Jan  1 00:00:00 1970
From: guijianfeng@cn.fujitsu.com (Gui Jianfeng)
Date: Mon, 16 Feb 2009 13:16:19 +0800
Subject: [ltt-dev] [PATCH] LTTng: Make mark-control work in debugfs
In-Reply-To: <20090216043703.GA27302@Krystal>
References: <20090213065458.GA20178@Krystal> <4998D2B6.7030001@cn.fujitsu.com>
	<20090216121549.E80D.A69D9226@jp.fujitsu.com>
	<4998E204.3010405@cn.fujitsu.com> <20090216043703.GA27302@Krystal>
Message-ID: <4998F6A3.4050001@cn.fujitsu.com>

Mathieu Desnoyers wrote:
> * Gui Jianfeng (guijianfeng at cn.fujitsu.com) wrote:
>> KOSAKI Motohiro wrote:
>>>>   Hi Mathieu,
>>>>
>>>>   I guess it's not easy to implement the *mkdir* for marker control.
>>>>   But I have an alternative solution that we can add a control file
>>>>   *new_channel* at the same level of channel directories, and add a 
>>>>   control file *new_marker* in each channel directory.
>>>>   Making use of the following, we can create a marker even if it 
>>>>   doesn't exist by now.
>>>>
>>>>   cd /mnt/ltt/markers
>>>>   echo "channel_name" > new_channel
>>>>   echo "marker_name" > channel_name/new_marker
>>> No.
>>> good user interface is important than easy implementation.
>>>
>>> this UI is really ugly.
>>   Hi Kosaki-san,
>>
>>   Thank you for your kind comments.
>>   I say *not easy* becasue debugfs itself doesn't support *mkdir* completely.
>>   If we are going to implement mkdir in marker control, we have to modify debugfs
>>   itself. IMHO, it's not reasonable and i guess most of the debugfs subsystems don't 
>>   like to see this change.
>>   On the other hand, it's very natual to use a control file to do something in debugfs.
>>   So, I choose this solution. I hope that what I say will clarify the situation.:)
>>   Mathieu, What do you say?
>>
> 
> Hi Gui,
> 
> I would recommend adding mkdir support in our own directories. That
> means we have to override debugfs use of "simple" libfs basic
> operations. We want to use the same operations, but also define a mkdir
> callback. See
> 
> fs/debugfs/inode.c:
> debugfs_get_inode :
> it sets inode->i_op = &simple_dir_inode_operations;
> 
> We will need to override this with our own. See
> fs/libfs.c: simple_dir_inode_operations
> 
  Ah, I see now, thanks for the suggestion, 

> I fully agree with Kosaki : the quality of the user interface is what
> matters and is much more important than implementation difficulty.

  Me too :)

> 
> Thanks,
> 
> Mathieu
> 
>>>
>>>
>>>
>>>
>>>
>> -- 
>> Regards
>> Gui Jianfeng
>>
>>
>> _______________________________________________
>> ltt-dev mailing list
>> ltt-dev at lists.casi.polymtl.ca
>> http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
>>
> 

-- 
Regards
Gui Jianfeng