From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Thompson Subject: Re: [PATCH v4 11/13] serial: asc: Adopt readl_/writel_relaxed() Date: Thu, 19 Jun 2014 14:12:01 +0100 Message-ID: <53A2E1A1.3010705@linaro.org> References: <1401961994-18033-1-git-send-email-daniel.thompson@linaro.org> <1403174303-25456-1-git-send-email-daniel.thompson@linaro.org> <1403174303-25456-12-git-send-email-daniel.thompson@linaro.org> <53A2C9A3.9090507@linaro.org> <53A2CD8D.1050106@linaro.org> <53A2D135.4020700@linaro.org> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <53A2D135.4020700@linaro.org> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: "linux-arm-kernel" Errors-To: linux-arm-kernel-bounces+linux-arm-kernel=m.gmane.org@lists.infradead.org To: Srinivas Kandagatla , Jason Wessel Cc: Mark Rutland , kernel@stlinux.com, kgdb-bugreport@lists.sourceforge.net, Linus Walleij , linux-kernel@vger.kernel.org, Jiri Slaby , Dirk Behme , Russell King , Nicolas Pitre , patches@linaro.org, Anton Vorontsov , "David A. Long" , linux-serial@vger.kernel.org, Catalin Marinas , kernel-team@android.com, devicetree@vger.kernel.org, linaro-kernel@lists.linaro.org, Pawel Moll , Ian Campbell , Colin Cross , Rob Herring , John Stultz , Thomas Gleixner , linux-arm-kernel@lists.infradead.org, Maxime Coquelin , Greg List-Id: devicetree@vger.kernel.org On 19/06/14 13:01, Srinivas Kandagatla wrote: >>> Why are we concern about x86 for this driver? >>> As per my understanding this IP is only seen on ARM and SH based CPUs so >>> why cant we just use relaxed versions, why ifdefs? >>> I think, this would involve fixing the kconfig and make it depend on SH >>> and ARM based platforms only. >> >> You mean just drop the COMPILE_TEST? >> >> In generally I like as much code as possible to compile on x86. Its >> worthwhile protection against the excessive/accidental ARMisms which >> could easily impact less common architectures (such as SH). > > That's fair. Does this mean that we are going do similar changes to > other ST drivers too? I didn't give any thought at all to other ST drivers. I don't see why a *general* preference (of mine or anyone else) would override what is right for any particular driver. I don't think "both manage ST peripherals" means drivers have much in common. >>> On the other hand, This patch looks more generic and applicable to most >>> of the drivers. Am not sure which way is the right one. >> >> I'm particularly keen on doing the right thing where readl_relaxed() is >> concerned because this function has a compiler barrier on ARM but not on >> x86. >> > My only concern is code duplication all across ST drivers. I really struggle to understand this. Why would anyone copy code out of the asc driver into the network driver (or any other ST driver)? >> Since having asc_in/asc_out made it easy to portably make these changes >> I decided is was better to be redundantly exemplary than conceal secret >> portability issues. > Your change would fit in nicely with as asc_in/out are wrappers and fix > st-asc but this would be just for asc driver. > What about other drivers which fall in same category? > > So I think we should just drop COMPILE_TEST and possibly make it > specific to ARM and SH or ARM only. I'm slightly uneasy about this primarily because all the rationale above describes a concern about drivers other than the one I seek to change. They ought to be outside the scope of this change. Nevertheless, since I said I don't feel that strongly about it, as you wish... I'll change this in v5.