xfs
[Top] [All Lists]

Re: [PATCH] xfs: remove unused locking flags

To: Alexey Dobriyan <adobriyan@xxxxxxxxx>
Subject: Re: [PATCH] xfs: remove unused locking flags
From: Nathan Scott <nathans@xxxxxxx>
Date: Mon, 10 Jul 2006 08:47:24 +1000
Cc: Andrew Morton <akpm@xxxxxxxx>, xfs@xxxxxxxxxxx, linux-fsdevel@xxxxxxxxxxxxxxx
In-reply-to: <20060709170529.GA7539@martell.zuzino.mipt.ru>; from adobriyan@gmail.com on Sun, Jul 09, 2006 at 09:11:03PM +0400
References: <20060708215324.GA7522@martell.zuzino.mipt.ru> <20060709105454.D1640104@wobbly.melbourne.sgi.com> <20060709170529.GA7539@martell.zuzino.mipt.ru>
Sender: xfs-bounce@xxxxxxxxxxx
User-agent: Mutt/1.2.5i
Hi Alexey,

On Sun, Jul 09, 2006 at 09:11:03PM +0400, Alexey Dobriyan wrote:
> On Sun, Jul 09, 2006 at 10:54:54AM +1000, Nathan Scott wrote:
> > On Sun, Jul 09, 2006 at 01:53:24AM +0400, Alexey Dobriyan wrote:
> > > Signed-off-by: Alexey Dobriyan <adobriyan@xxxxxxxxx>
> >
> > NACK.  These macros get used by other SGI code (not merged in mainline).
> > Their presence here has zero runtime cost, and keeps merges simpler for
> > me, so they need to stay.
> 
> In this case, yes, runtime overhead in nil.
> 
> What about passing dummy credentials?

Yes, what about it?  Is there any measurable cost there?  Show me, I'm
interested, really.  If there is, I'm sure we can do things differently
to remove that.  I'm starting to get a little hesitant about your
intentions here though to be honest, you seem to have a bit of an axe
to grind (do you?  why?  was it something I said?).

> What about DMAPI stubbed to errors
> since XFS hit mainline (at least 900 lines which can be removed)? They
> have runtime overhead.

DMAPI is useful, and used - if its unfit for mainline (which a number
of people seem to agree to) then why not be constructive and massage
it into a form more fit for mainline, if thats the issue here?

I did have a patch a little while back which made the mainline DMAPI
"if (event-enabled)" stuff be compile-time conditional (which should
be done independently of whether full-blown DMAPI is in mainline) --
I got side-tracked before thoroughly testing that change though.  If
you'd like that WIP patch, lemme know, and I'll happily pass it along.

> I can add "behavoir chains" here but patch for dealing with them doesn't
> exists yet, so I won't.

You're a bit confused, there I think - behaviours are used today, in
mainline, and there is active XFS feature development work relying on
them even further.  As such, that patch would also get NACKed.  There
is no harm in asking though.

If you're unhappy with the way XFS is maintained, then it is GPL code
and you're free to make of it as you will, of course, as far as the
license permits.  I'd prefer to work with you though to make useful
changes to the mainline version, and would caution you that it is a
very complex body of code that a few weeks of cleanup efforts really
wont have prepared you for.  Its also my hope that you'll find us XFS
folks all round nice guys though :) and I'd vastly prefer to work with
you to cleanup the version we maintain, in mutually satisfactory ways,
but the choice is always yours.

cheers.

ps: you've not answered my question as to whether you tested your last
change at all - did you?  thanks.

-- 
Nathan


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