lkml.org 
[lkml]   [2010]   [Aug]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH 2/5] vfs: make i_op->permission take a dentry instead of an inode
From
Date
On Fri, 2010-08-27 at 14:11 +1000, Neil Brown wrote:
> On Thu, 26 Aug 2010 16:24:02 -0400
> "David P. Quigley" <dpquigl@tycho.nsa.gov> wrote:
>
> > I may be missing something but I looked at your patch series and I see
> > no good reason for this patch at all. You just churned a lot of code for
> > something that you don't even have a need for in the patch set. Your
> > only two new callers of this function could just as easily have used the
> > inode since it isn't doing anything special with the dentry. It actually
> > pulls the inode out of it and uses it in generic_permission and
> > security_inode_permission. If you are going to change this you should
> > also change generic_permission as well. Honestly I'd rather see the
> > dentry requirement removed from inode operations instead but
> > unfortunately this isn't possible as I found out with my attempts to
> > remove the dentry requirement for get/setxattr
>
>
> union_permission needs the dentry to get access to d_fsdata, which caches the
> upperpath and lowerpath which were found at lookup time.
>
> Is that what you missed?
>

You're correct I missed the line where that was being pulled out of the
dentry. The better question for me would be why do it this way as
opposed to what the union file systems are doing. If neither UnionFS or
AUFS are having to make this change so I'd like a much better
explination for this change. I'm not seeing enough information in the
form of why he designed the prototype this way to justify a change that
the other implementations don't seem to need.

Dave



\
 
 \ /
  Last update: 2010-08-27 20:25    [W:0.101 / U:14.024 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site