From: khilman@linaro.org (Kevin Hilman) Date: Fri, 28 Feb 2014 12:24:21 -0800 Subject: [PATCH] ARM: qcom: Enable basic support for Qualcomm platforms in multi_v7_defconfig In-Reply-To: <7heh2nt6uj.fsf@paris.lan> (Kevin Hilman's message of "Fri, 28 Feb 2014 10:44:36 -0800") References: <1393360445-16923-1-git-send-email-galak@codeaurora.org> <7hy50xv3rw.fsf@paris.lan> <7hzjlbtc9y.fsf@paris.lan> <20140228182745.GC2487@codeaurora.org> <7heh2nt6uj.fsf@paris.lan> Message-ID: <7htxbjrnnu.fsf@paris.lan> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org Kevin Hilman writes: > Stephen Boyd writes: > >> On 02/28, Kevin Hilman wrote: >>> Hi Kumar, >>> >>> Kevin Hilman writes: >>> >>> > Kumar Gala writes: >>> > >>> >> Enable support for the MSM8x60, MSM8960, and MSM8974 SoCs, clocks and >>> >> serial console as part of the standard multi_v7_defconfig. >>> >> >>> >> Signed-off-by: Kumar Gala >>> >> --- >>> >> >>> >> I leave this to the arm-soc guys to apply because of possible conflicts >>> >> with other updates to multi_v7_defconfig. >>> > >>> > Seem to be a handful of !qcom changes here, probably resulting from a >>> > run of 'make savedefconfig' without all the new changes in. >>> > >>> > I've thinned this out to only the qcom changes and applied the following >>> > to next/boards. I think it's better to do any 'make savedefconfig' >>> > cleanups as a separate, cleanup-only patch as needed. >>> >>> Bisecting a new boot failure on u8500/snowball came down to this patch. >>> Simply disabling CONFIG_ARCH_QCOM in mutli_v7_defconfig gets the >>> u8500/snowball booting again, so it looks like there may still be some >>> qcom specifics that are being run on non-qcom platforms with a >>> multi-platform config. >>> >>> Could you audit for initcalls/drivers being initialized on non-qcom >>> SoCs? >>> >> >> Can you try disabling CONFIG_ARCH_MSM8X60 and >> CONFIG_ARCH_MSM8960? MY suspicion is the text offset movement is >> causing problems, but I hope I'm wrong. > > Sorry to dash your hopes, but you're right. > > Starting from a fresh multi_v7_defconfig and disabling those two Kconfig > options gets the snowball booting again on multi_v7_defconfig. For the record... After some discussion on IRC, we discovered this is because those options add a 2Mb offset to TEXT_OFFSET, and my boot tool wasn't taking that into account when creating the uImage to boot on snowball. I guess it's time to upgrade u-boot on snowball with bootz support so it can directly boot a zImage. Kevin