[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Pine.LNX.4.64.0706041051090.30836@alien.or.mcafeemobile.com>
Date: Mon, 4 Jun 2007 10:57:31 -0700 (PDT)
From: Davide Libenzi <davidel@...ilserver.org>
To: Andrew Morton <akpm@...ux-foundation.org>
cc: Ingo Molnar <mingo@...e.hu>, Eric Dumazet <dada1@...mosbay.com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Linus Torvalds <torvalds@...ux-foundation.org>,
Ulrich Drepper <drepper@...hat.com>
Subject: Re: [patch 1/2] ufd v1 - unsequential O(1) fdmap core
On Mon, 4 Jun 2007, Andrew Morton wrote:
> > must be RCU friendly (proper barriers) since it's used in
> > lockless code,
>
> Haven't looked at that.
>
> > and must have flags associated to an allocation.
>
> Don't understand that.
It needs to be able to store flags (like close-on-exec, and maybe more)
together with a file*.
> > And I'm
> > leaving out the O(1) part, that for something like this, is just silly not
> > to have it. This is really an array.
>
> Having to walk down a tree in fget_light() would kinda suck.
>
> What about my b)?
Right now, I'm working on a patch that uses fdmap even for legacy
(sequential) fd allocations. With a cleaner interface, w/out having code
all around the kernel that access fdtable members directly. This should
cut some code WRT using two different allocators (code in fs/file.c will
almost zip-away), and maybe it'll look even better and have some new
comments too ;)
It can be, in theory, completely abstracted and moved into lib/. I'll look
into it once I completed the first cleanup, assuming the whole thing won't
look worse than before :)
- Davide
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists