From mboxrd@z Thu Jan 1 00:00:00 1970 From: Arnd Bergmann Subject: Re: [PATCH] ARM: add get_user() support for 8 byte types Date: Thu, 15 Nov 2012 14:39:41 +0000 Message-ID: <201211151439.41393.arnd@arndb.de> References: <1352495853-9790-1-git-send-email-rob.clark@linaro.org> <201211151339.08016.arnd@arndb.de> Mime-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org To: Rob Clark Cc: Russell King - ARM Linux , linux-arm-kernel@lists.infradead.org, patches@linaro.org, linux-kernel@vger.kernel.org, linux-omap@vger.kernel.org, dri-devel@lists.freedesktop.org List-Id: linux-omap@vger.kernel.org On Thursday 15 November 2012, Rob Clark wrote: > > I still haven't heard a conclusive argument why we need to use get_user() > > rather than copy_from_user() in the DRM code. Is this about a fast path > > where you want to shave off a few cycles for each call, or does this > > simplify the code structure, or something else? > > well, it is mostly because it seemed like a good idea to first try to > solve the root issue, rather than having to fix things up in each > driver when someone from x86-world introduces a 64b get_user().. As pointed out by hpa earlier, x86-32 doesn't have a 64b get_user either. I don't think we have a lot of drivers that are used only on 64-bit x86 and on 32-bit ARM but not on 32-bit x86. > There are some other arch's that don't have 64b get_user(), but I > don't think any that have any DRM drivers. If 64b get_user() is > really not intended to be supported by all archs, it is better to > remove it from x86 and the other arch's that do currently support it, > rather than making it possible to write drivers that are broken on > some archs. The majority of architectures we support have PCI and should be able to build the regular (radeon, nouveau, MGA, VIA, ...) DRM drivers AFAICT. Arnd From mboxrd@z Thu Jan 1 00:00:00 1970 From: arnd@arndb.de (Arnd Bergmann) Date: Thu, 15 Nov 2012 14:39:41 +0000 Subject: [PATCH] ARM: add get_user() support for 8 byte types In-Reply-To: References: <1352495853-9790-1-git-send-email-rob.clark@linaro.org> <201211151339.08016.arnd@arndb.de> Message-ID: <201211151439.41393.arnd@arndb.de> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Thursday 15 November 2012, Rob Clark wrote: > > I still haven't heard a conclusive argument why we need to use get_user() > > rather than copy_from_user() in the DRM code. Is this about a fast path > > where you want to shave off a few cycles for each call, or does this > > simplify the code structure, or something else? > > well, it is mostly because it seemed like a good idea to first try to > solve the root issue, rather than having to fix things up in each > driver when someone from x86-world introduces a 64b get_user().. As pointed out by hpa earlier, x86-32 doesn't have a 64b get_user either. I don't think we have a lot of drivers that are used only on 64-bit x86 and on 32-bit ARM but not on 32-bit x86. > There are some other arch's that don't have 64b get_user(), but I > don't think any that have any DRM drivers. If 64b get_user() is > really not intended to be supported by all archs, it is better to > remove it from x86 and the other arch's that do currently support it, > rather than making it possible to write drivers that are broken on > some archs. The majority of architectures we support have PCI and should be able to build the regular (radeon, nouveau, MGA, VIA, ...) DRM drivers AFAICT. Arnd