From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ben Dooks Subject: Re: [Ct173-kernel-internal] [PATCH 0/4] ARM: tegra: Support for big-endian (version 3) Date: Tue, 20 Jan 2015 11:45:05 +0000 Message-ID: <54BE3FC1.5010107@codethink.co.uk> References: <1420817231-8346-1-git-send-email-bob.mottram@codethink.co.uk> <54BE2E77.9050909@codethink.co.uk> <54BE3DD2.1060204@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: In-Reply-To: <54BE3DD2.1060204-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> Sender: linux-tegra-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Dmitry Osipenko , Bob Mottram , linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, swarren-3lzwWm7+Weoh9ZMKESR00Q@public.gmane.org, thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org, gnurou-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org, ct173-kernel-internal-81qHHgoATdFT9dQujB1mzip2UmYkHbXO@public.gmane.org List-Id: linux-tegra@vger.kernel.org On 20/01/15 11:36, Dmitry Osipenko wrote: > 20.01.2015 13:31, Ben Dooks =D0=BF=D0=B8=D1=88=D0=B5=D1=82: >> On 09/01/15 15:27, Bob Mottram wrote: >>> During testing of big-endian on a Jetson TK1 with baserock, the >>> following >>> issues were found and fixed. Note that to get the system running wi= th >>> big-endian you will need to disable ARCH_SUSPEND_POSSIBLE within >>> arch/arm/Kconfig. >>> >>> Bob Mottram (4): >>> ARM: tegra: Add a separate big-endian configuration >>> ARM: gpio: tegra: Maintain endianness >>> ARM: tegra: enable big-endian support >>> ARM: tegra: fix big-endian issue with irq code >> >> Is it possible to get the tegra core patches applied? These seem to = work >> enough to get a system running and once the PM fixes are in there sh= ould >> be no problems running big-endian tegra? >> > I don't think it's enough. Upstream kernel doesn't accept half workin= g > solutions. It's not only PM that needs fixing, but i2c, timer, reset > handler and possibly more. You are enabling big-endian across all teg= ra > generations, but tegra2 won't boot at all for example. Patches must b= e > tested on all hw before applying. And nit: big-endian should be enabl= ed > after all dependencies resolved. Ok, we hadn't found any issue with the reset. I thought we had the timer fix in this series. Most of the other drivers didn't use any __raw code so we're reasonably confident they will work. I didn't see any issues with the i2c from visual inspection, what's wrong with it in big endian? --=20 Ben Dooks http://www.codethink.co.uk/ Senior Engineer Codethink - Providing Genius