From mboxrd@z Thu Jan  1 00:00:00 1970
From: marc.zyngier@arm.com (Marc Zyngier)
Date: Tue, 29 Jul 2014 14:41:22 +0100
Subject: [PATCH v15 01/12] irq: gic: support hip04 gic
In-Reply-To: <2455679.mpY3ZEfao6@wuerfel>
References: <1406555876-11989-1-git-send-email-haojian.zhuang@linaro.org>
 <1406555876-11989-2-git-send-email-haojian.zhuang@linaro.org>
 <87a97to025.fsf@approximate.cambridge.arm.com> <2455679.mpY3ZEfao6@wuerfel>
Message-ID: <53D7A482.7050705@arm.com>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On 29/07/14 14:05, Arnd Bergmann wrote:
> On Monday 28 July 2014 18:38:58 Marc Zyngier wrote:
>> Overall, this code should be able to sitting within a #ifdef/#endif
>> block, only selected if this platform is enabled, and possibly find a
>> way not to impact all the other platforms when this is not selected.
>>
>> Finally, I'd like to outline how much I dislike the way the GIC
>> architecture has been abused here. Yes, this solves a particular
>> problem, at a given point in time, but this also feels extremely short
>> sighted from whoever has put this thing together. This really feels like
>> a short-term HW hack that is already addressed by GICv3. Merging support
>> for non architecture compliant HW is never the best solution.
> 
> Not merging support for shipping hardware is also not a solution,
> and we already support any number of irqchip drivers and have all
> the logic we need to sort that out at runtime.

I don't dispute this, and maybe I'm just tired of seeing pointlessly
divergent HW being pushed around.

> My feeling is that the original approach taken in the early version
> of the patch set was actually better, given all the problems with
> integrating this nicely into the gic driver.
> 
> Would you be happier with that? The downside of that would be
> that some driver bugs would get fixed in one driver but not the
> other one, which is why we normally try to avoid it. On the other
> hand, it means that the normal GIC driver does not get polluted
> with hacks that are only needed for the hip04 version and there
> is an obvious way to disable the driver at compile time when you
> build for other platforms.

Indeed. I think I'd be much happier with a completely separate driver.
There will even be some benefits to the hip04 driver, as it won't have
to deal with the non-DT stuff and the crazy gic_arch_extn.

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...