xfs
[Top] [All Lists]

Re: extended attributes support on powerpc

To: linux-xfs@xxxxxxxxxxx
Subject: Re: extended attributes support on powerpc
From: Ethan Benson <erbenson@xxxxxxxxxx>
Date: Sun, 27 Jan 2002 17:28:36 -0900
In-reply-to: <2576.1012184666@xxxxxxxxxxxxxxxxxxxxx>; from kaos@xxxxxxx on Mon, Jan 28, 2002 at 01:24:26PM +1100
Mail-copies-to: nobody
Mail-followup-to: linux-xfs@xxxxxxxxxxx
References: <20020127170434.K14742@xxxxxxxxxxxxxxx> <2576.1012184666@xxxxxxxxxxxxxxxxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
User-agent: Mutt/1.2.5i
On Mon, Jan 28, 2002 at 01:24:26PM +1100, Keith Owens wrote:
> On Sun, 27 Jan 2002 17:04:35 -0900, 
> Ethan Benson <erbenson@xxxxxxxxxx> wrote:
> >extended attributes (and thus acls) do not appear to be supported on
> >powerpc, as far as i can tell this is because the acl-xattr patches do
> >not add the syscall and unistd.h defines for the powerpc arch (only
> >i386 and ia64).  is there any more to it then that? and if not how
> >difficult is it to do so?  (it looks trivial, but a kernel hacker i am
> >not...)
> 
> It should be that simple.  Find 3 unused syscall numbers in
> sys_call_table in arch/ppc/kernel/misc.S, in 2.4.17 the next 3 are
> 208-210.  Add these lines after sys_gettid, adjusting to fit current
> ppc syscall usage.

in the i386 version you seem to be adding about 25 dummy syscalls, am
i correct in assuming this is just `padding' to avoid conflicts if
Linus allocates some new syscalls?  is the same necessary/wise to do
on powerpc as well?

> #ifdef CONFIG_HAVE_ATTRCTL
>       .long SYMBOL_NAME(sys_attrctl)
>       .long SYMBOL_NAME(sys_acl_get)
>       .long SYMBOL_NAME(sys_acl_set)          /* 210 */
> #else
>       .long SYMBOL_NAME(sys_ni_syscall)
>       .long SYMBOL_NAME(sys_ni_syscall)
>       .long SYMBOL_NAME(sys_ni_syscall)       /* 210 */
> #endif        /* CONFIG_HAVE_ATTRCTL */
> 
> Add these to include/asm-ppc/unistd.h, adjusting as required.
> 
> #define __NR__attrctl         208
> #define __NR__acl_get         209
> #define __NR__acl_set         210
> 
> >btw, somewhat related but i noticed that the -misc patch adds the
> >syscall defines to ia64's unistd.h, but the -acl-extattr patch adds it
> >to i386's any reason for this inconsistency? or a bug in the patch
> >generation scripts?
> 
> Human error, I will fix it in the next patch set.
> 

-- 
Ethan Benson
http://www.alaska.net/~erbenson/

Attachment: pgpMQAQCjytW7.pgp
Description: PGP signature

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