Re: [PATCH 1/6] staging: android: binder: Remove some funny && usage

From: Daniel Walker <dwalker@fifo99.com>
To: Brian Swetland <swetland@google.com>
Cc: "Arve Hjønnevåg" <arve@android.com>,
	"Jeremy Fitzhardinge" <jeremy@goop.org>,
	"Greg Kroah-Hartman" <greg@kroah.com>,
	linux-kernel@vger.kernel.org, hackbod@android.com
Subject: Re: [PATCH 1/6] staging: android: binder: Remove some funny && usage
Date: Wed, 24 Jun 2009 15:49:38 -0700	[thread overview]
Message-ID: <1245883778.32124.214.camel@desktop> (raw)
In-Reply-To: <a55d774e0906241514w761f86fva294784583cdd8a4@mail.gmail.com>

On Wed, 2009-06-24 at 15:14 -0700, Brian Swetland wrote:
> 
> I guess it depends on the questions -- for good or ill, it's what is
> in use and replacing it with a different mechanism would be a pretty
> huge effort that there's unlikely to be time for in the near future.
> Quite a bit of the userspace framework depends on the binder handling
> the remote reference counting, object lifespan, etc, stuff and it's
> subtle and hairy to debug.  Moving to a different transport is
> possible (it's just software, as they say), but there's a lot of code
> that depends on the behavior that exists today.

I'm really interested in how binder was picked. It seems like a dead
technology (hasn't been touched in 3 years according to open-binder.org)

I think the best option is to write a layer over D-Bus that implements
the binder API. I don't think that would be too difficult since D-Bus
does IPC, binder does IPC, and the differences should be worked out with
a layer between the two. D-Bus seems like the best solution since it has
the best community backing..

> If the binder is absolutely unacceptable as something to be included
> in the linux kernel (that's a message that I seem to be getting here),
> I think the best thing for us to do is maintain it in our tree for now
> and focus on stuff that is far less controversial.

Yes... I'm refraining from doing more cleanups because it seems like a
lost cause. 

> For example the msm7k SoC code may need some various cleanup, but I
> think at the end of the day adding support for a new ARM based SoC and
> peripherals is not going to be that contentious.  The
> wakelock/suspendblock stuff is a bit further out there, but there
> seemed to be some good progress on getting it reviewed and revised on
> the linux-pm list, last I saw.

Do you have a msm branch someplace that is strictly msm support with
absolutely no generic changes mixed in?

The only thing in staging that is less controversial from my perspective
is ram_console.c that could actually be cleaned up and possibly merged.

Daniel