xfs
[Top] [All Lists]

Re: XFS ACL Patch

To: dcox@xxxxxxxxxx
Subject: Re: XFS ACL Patch
From: "Nathan Scott" <nathans@xxxxxxxxxxxxxxxxxxxxxxxx>
Date: Fri, 17 Nov 2000 10:15:13 -0400
Cc: linux-xfs@xxxxxxxxxxx, Scott Smyth <ssmyth@xxxxxxxxxx>, John Trostel <jtrostel@xxxxxxxxxx>, Robert Stickel <rstickel@xxxxxxxxxx>
In-reply-to: Danny <danscox@xxxxxxxxxxxxxx> "Re: XFS ACL Patch" (Nov 16, 5:50pm)
References: <200011161859.MAA12613@xxxxxxxxxxxxxxxxxxxx> <3A1464C9.F3027609@xxxxxxxxxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
hi Danny,

On Nov 16,  5:50pm, Danny wrote:
> Subject: Re: XFS ACL Patch
> ...
>       Okay, here's a patch which should be much happier!  It's against
> test10, and seems to work on my system.
> ...
>       I noticed that the attr_get, set, remove, et. al. have been changed to
> be just one system call instead of five (or so).  Should I also emulate
> that and create only one system call for ACLs?
> 

They were cut down from eight system calls to one, primarily to
solve the system call collision problem which cropped up between
XFS and RedHat 7.0, but also to get some SGI folk looking into
the issues related to EAs & to start thinking about ways to
properly "fix" the XFS source.

WRT system calls for ACLs, I think you need to ask that on the
linux-acl, linux-fsdevel and linux-kernel lists.  I'd suggest
having a look at the approach that Andreas has taken for ext2
ACLs and make use of that if at all possible - would be a good
idea to contact Andreas (Gruenbacher <ag@xxxxxxxxxxxxxxx>) and
work together to ensure a consistent interface is used.

As with extended attributes, once an interface is agreed upon,
that is what XFS will have to use (i.e. the existing one-syscall
"attrctl" interface is guaranteed to change, so don't base any
decisions on that!).

cheers.

-- 
Nathan

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