xfs
[Top] [All Lists]

Re: [PATCH] always set a/c/mtime through ->setattr

To: Christoph Hellwig <hch@xxxxxx>
Subject: Re: [PATCH] always set a/c/mtime through ->setattr
From: Artem Bityutskiy <Artem.Bityutskiy@xxxxxxxxx>
Date: Tue, 20 May 2008 09:19:18 +0300
Cc: viro@xxxxxxxxxxxxxxxxxx, linux-fsdevel@xxxxxxxxxxxxxxx, xfs@xxxxxxxxxxx
In-reply-to: <20080520060838.GA6436@xxxxxx>
Organization: Nokia OYJ
References: <20080520060838.GA6436@xxxxxx>
Reply-to: Artem.Bityutskiy@xxxxxxxxx
Sender: xfs-bounce@xxxxxxxxxxx
User-agent: Thunderbird 2.0.0.14 (X11/20080501)
Christoph,

Christoph Hellwig wrote:
Currently touch_atime and file_update_time directly update a/c/mtime
in the inode and just mark the inode dirty afterwards.  This is pretty
bad for some more complex filesystems that have various different types
of dirtying an inode and/or need to store the data in another place
for example for a buffer to be logged.

This patch changes touch_atime and file_update_time to not update the
inode directly but rather call through ->setattr into the filessystem.

This patch would make our (UBIFS develpers') life easier, thank you!

Could we go a further and allow the file-system returning error if it
for some reasons cannot change the time? For example, the FS could
return -EIO or -ENOSPC up and VFS would have to free resources and
propagate this error to user-space. Is this possible?

--
Best Regards,
Artem Bityutskiy (Артём Битюцкий)


<Prev in Thread] Current Thread [Next in Thread>