Re: [PATCH 4/4] MIPS: bmips: Convert bcm63xx_wdt to use WATCHDOG_CORE

From: Simon Arlott <simon@fire.lp0.eu>
To: Guenter Roeck <linux@roeck-us.net>,
	"devicetree@vger.kernel.org" <devicetree@vger.kernel.org>,
	Ralf Baechle <ralf@linux-mips.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Jason Cooper <jason@lakedaemon.net>,
	Marc Zyngier <marc.zyngier@arm.com>,
	Kevin Cernekee <cernekee@gmail.com>,
	Florian Fainelli <f.fainelli@gmail.com>,
	Wim Van Sebroeck <wim@iguana.be>,
	Miguel Gaio <miguel.gaio@efixo.com>,
	Maxime Bizon <mbizon@freebox.fr>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	linux-mips@linux-mips.org, linux-watchdog@vger.kernel.org
Cc: Rob Herring <robh+dt@kernel.org>, Pawel Moll <pawel.moll@arm.com>,
	Mark Rutland <mark.rutland@arm.com>,
	Ian Campbell <ijc+devicetree@hellion.org.uk>,
	Kumar Gala <galak@codeaurora.org>
Subject: Re: [PATCH 4/4] MIPS: bmips: Convert bcm63xx_wdt to use WATCHDOG_CORE
Date: Sat, 21 Nov 2015 21:44:27 +0000	[thread overview]
Message-ID: <5650E5BB.6020404@simon.arlott.org.uk> (raw)
In-Reply-To: <5650E2FA.6090408@roeck-us.net>

On 21/11/15 21:32, Guenter Roeck wrote:
> On 11/21/2015 11:05 AM, Simon Arlott wrote:
>> Convert bcm63xx_wdt to use WATCHDOG_CORE and add a device tree binding.
>>
>> Adds support for the time left value and provides a more effective
>> interrupt handler based on the watchdog warning interrupt behaviour.
>>
>> This removes the unnecessary software countdown timer and replaces the
>> use of bcm63xx_timer with a normal interrupt when not using mach-bcm63xx.
>>
> 
> Hi Simon,
> 
> this is really doing a bit too much in a single patch.
> Conversion to the watchdog infrastructure should probably be
> the first step, followed by further optimizations and improvements.

I'll split it into two patches, but that won't remove the need for #ifdefs.

> In general, it would be great if we can avoid #ifdef in the code.
> Maybe there is some other means to determine if one code path
> needs to be taken or another. The driver may be part of a
> multi-platform image, and #ifdefs in the code make that all
> but impossible. Besides, it makes the code really hard to read
> and understand.

It's impossible to avoid the #ifdefs because the driver needs to support
mach-bmips while still supporting mach-bcm63xx. I don't think they make
it too difficult to understand. Until there are device tree supporting
drivers for everything mach-bcm63xx needs, it can't be removed.

> We have some infrastructure changes in the works which will move
> the need for soft-timers from individual drivers into the watchdog core.
> Would this possibly be helpful here ? The timer-driven watchdog ping
> seems to accomplish pretty much the same.

There is no need for a software timer. This is not a timer-driven
watchdog ping, there is an unmaskable timer interrupt when the watchdog
timer has less than 50% remaining.

-- 
Simon Arlott