From mboxrd@z Thu Jan 1 00:00:00 1970 From: Rob Herring Subject: Re: [PATCH V2 3/3] ARM: tegra: move debug-macro.S to include/debug Date: Thu, 18 Oct 2012 09:15:23 -0500 Message-ID: <50800EFB.9060503@gmail.com> References: <1350328024-30485-1-git-send-email-swarren@wwwdotorg.org> <1350328024-30485-3-git-send-email-swarren@wwwdotorg.org> <507EC303.1080000@gmail.com> <507EDB37.1060102@wwwdotorg.org> <507F1F31.2060503@wwwdotorg.org> <5080088C.9090607@gmail.com> <20121018135444.GT21164@n2100.arm.linux.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20121018135444.GT21164-l+eeeJia6m9vn6HldHNs0ANdhmdF6hFW@public.gmane.org> Sender: linux-tegra-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Russell King - ARM Linux Cc: Stephen Warren , Olof Johansson , linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Stephen Warren , linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org, Arnd Bergmann List-Id: linux-tegra@vger.kernel.org On 10/18/2012 08:54 AM, Russell King - ARM Linux wrote: > On Thu, Oct 18, 2012 at 08:47:56AM -0500, Rob Herring wrote: >> Here is what I mentioned previously. This removes the static mapping from >> the platforms. This is untested and probably breaks on different DEBUG_LL >> options. For now, platforms call debug_ll_io_init, but once all platforms >> are converted, this can be called from devicemaps_init. > > There's a problem with this approach. What if the platform specifically > sets the debug addresses to be within one of it's existing mappings (which > is definitely what you'd want to do with 8250-based UARTs attached to a > PCI bus.) > > This isn't going to work in that case unless we split the debug mapping > from the PCI IO space mapping. It is opt-in, so we may never get to the last step. We're still better off than now. Is there a platform that is doing this? If so, I didn't move any DEBUG_LL users to the fixed i/o space so they would still be using their old i/o space address. Having 2 mappings would work, right? You use the debug mapping initially and when PCI bus and the real console driver are up, you switch to the fixed i/o space. This isn't just a PCI issue. You could also have a mapping that covers a block of SOC peripherals including the uart and you don't want a separate mapping. Rob From mboxrd@z Thu Jan 1 00:00:00 1970 From: robherring2@gmail.com (Rob Herring) Date: Thu, 18 Oct 2012 09:15:23 -0500 Subject: [PATCH V2 3/3] ARM: tegra: move debug-macro.S to include/debug In-Reply-To: <20121018135444.GT21164@n2100.arm.linux.org.uk> References: <1350328024-30485-1-git-send-email-swarren@wwwdotorg.org> <1350328024-30485-3-git-send-email-swarren@wwwdotorg.org> <507EC303.1080000@gmail.com> <507EDB37.1060102@wwwdotorg.org> <507F1F31.2060503@wwwdotorg.org> <5080088C.9090607@gmail.com> <20121018135444.GT21164@n2100.arm.linux.org.uk> Message-ID: <50800EFB.9060503@gmail.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On 10/18/2012 08:54 AM, Russell King - ARM Linux wrote: > On Thu, Oct 18, 2012 at 08:47:56AM -0500, Rob Herring wrote: >> Here is what I mentioned previously. This removes the static mapping from >> the platforms. This is untested and probably breaks on different DEBUG_LL >> options. For now, platforms call debug_ll_io_init, but once all platforms >> are converted, this can be called from devicemaps_init. > > There's a problem with this approach. What if the platform specifically > sets the debug addresses to be within one of it's existing mappings (which > is definitely what you'd want to do with 8250-based UARTs attached to a > PCI bus.) > > This isn't going to work in that case unless we split the debug mapping > from the PCI IO space mapping. It is opt-in, so we may never get to the last step. We're still better off than now. Is there a platform that is doing this? If so, I didn't move any DEBUG_LL users to the fixed i/o space so they would still be using their old i/o space address. Having 2 mappings would work, right? You use the debug mapping initially and when PCI bus and the real console driver are up, you switch to the fixed i/o space. This isn't just a PCI issue. You could also have a mapping that covers a block of SOC peripherals including the uart and you don't want a separate mapping. Rob