From mboxrd@z Thu Jan 1 00:00:00 1970 From: Zoltan Gyarmati Date: Fri, 15 Jan 2016 02:53:11 +0100 Subject: [Buildroot] [PATCH 1/1] configs: beaglebone: update kernel to 4.1.13 and u-boot to 2015.10 In-Reply-To: <20160113095153.6b4f95cb@free-electrons.com> References: <1452653542-5562-1-git-send-email-mr.zoltan.gyarmati@gmail.com> <20160113095153.6b4f95cb@free-electrons.com> Message-ID: <56985107.2090006@gmail.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Deat Thomas & All, thanks for looking into this, please see the answers inline... On 01/13/2016 09:51 AM, Thomas Petazzoni wrote: > Zoltan, > > On Wed, 13 Jan 2016 03:52:22 +0100, Zoltan Gyarmati wrote: > >> This patch is basically the cleaned-up version of Chris >> LaRocque's patch from his mail > > Thanks for respining this patch. > >> "[PATCH 1/2] Updated the Beaglebone configuration, 01/03/2016 >> 10:01 PM" and the cleanup followed Arnout's corresponding >> answers. Changes to Chris' patches: * dtc patches fetched from GH >> and placed in 'board/beaglebone/patches/dtc' * >> BR2_GLOBAL_PATCH_DIR set accordingly * added seperated kernel >> defconfig instead of patching the upstream defconfig * cleanups >> in uEnv.txt * fix u-boot version to 2015.10 >> >> The heavy-lifting was done by Chris here, the rest is mostly >> cosmetics. If anything els seems to be tweaked, please let us >> know. >> >> >> board/beaglebone/linux-3.12.config | 251 --- >> board/beaglebone/linux_4.1.13_bb.org_defconfig | 2043 >> ++++++++++++++++++++ > > Looking at its size, this file looks like a full config file and > not a defconfig. Please use a defconfig. It's indeed a defconfig (generated with `make linux-savedefconfig`), but it enables tons of modules (it's just a slightly modified version of the upstream one, see at https://github.com/beagleboard/linux/blob/4.1/arch/arm/configs/bb.org_de fconfig) I'm going to cook a shrinker defconfig version of this for Buildroot, with just a basic set of modules enabled, especially as this version also takes a lot of time to build... > > Or better: please use the defconfig that is inside the kernel tree, > and if you need additional tweaks, use config fragments. > >> ...comment-block-of-fdt_get_property_namelen.patch | 32 + >> .../0002-fdt-Add-a-function-to-count-strings.patch | 246 +++ >> ...d-a-function-to-get-the-index-of-a-string.patch | 159 ++ >> ...004-fdt-Add-functions-to-retrieve-strings.patch | 173 ++ > > I am not supper happy with the solution of having those DTC > patches specifically for the BeagleBone platform. If I understand > correctly, those patches are needed to get the DT overlay to work, > right ? > > We had the same problem for the C.H.I.P platform, so it would also > be useful for us to have a overlay-capable DTC. You can see what > my colleague Antoine T?nart did: > > https://github.com/NextThingCo/CHIP-buildroot/commit/1a73b514d204d0c6e 9100d53011a11d99fd8b778 > > https://github.com/NextThingCo/CHIP-buildroot/commit/f3334df58611ff3cae5 fd62906423fdce3d80d24 > https://github.com/NextThingCo/CHIP-buildroot/commit/c4457eba3099f5ca5 14c06bb35b42772178a151a > > It is not ideal because it relies on a separate dtc-overlay > package, which packages a different version of dtc, which Antoine > has cooked on his own github repository. > > We really need to find a solution that is acceptable in the > upstream Buildroot, and applies to multiple platforms. > >> ...cache-flush-management-symbols-when-MULTI.patch | 34 - >> .../0001-am335x_evm-uEnv.txt-bootz-n-fixes.patch | 754 >> ++++++++ > > This huge U-Boot patch is really ugly and mixes lots of stuff. Why > do you need it? Agreed, it's an ugly patch, coming from https://github.com/eewiki/u-boot-patches/blob/master/v2015.10/0001-am335 x_evm-uEnv.txt-bootz-n-fixes.patch , and i couldn't find a nicer patchset. Looking ahead to 2016.01, there is a smaller patch for it (https://github.com/eewiki/u-boot-patches/blob/master/v2016.01/0001-am33 5x_evm-uEnv.txt-bootz-n-fixes.patch), where most of the changes are config-related and just a small actual code is patched, i'm going to test it and and report. (maybe we can go with a small patch extracted from the linked one, and a custom defconfig file?) > > >> -# bootloader BR2_TARGET_UBOOT=y >> BR2_TARGET_UBOOT_BOARDNAME="am335x_evm" >> BR2_TARGET_UBOOT_CUSTOM_VERSION=y >> -BR2_TARGET_UBOOT_CUSTOM_VERSION_VALUE="2013.10" >> +BR2_TARGET_UBOOT_CUSTOM_VERSION_VALUE="2015.10" >> +BR2_TARGET_UBOOT_NEEDS_DTC=y > > So this is going to use the host-dtc package for U-Boot, but isn't > the kernel still going to use its internal copy of dtc ? If so, how > are your dtc patches useful for the kernel ? > > Thanks! > > Thomas > Thanks, regards zgyarmati -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iQEcBAEBAgAGBQJWmFEGAAoJEAGmEeeR8iVhCVgIAKQpqvofRUQXyy10wKcAGCvg 8SPtWt09Cl69Y7sNx32kSkrfAWDHI0O3ez+Arz3lodIORAyV0Dq//BwYkof4l/ed uHkI5v+bH9g1JnkxbJ4dssVO2hTFE95rtxmsIwXSTmmoXG9hsHMaXkY34Ka615AJ ZoLQTnY5KL16MINA1N0vfi9KypRHQ69UUJugO9NoNhOpiwwXoTQMCCWmCIdnkcH5 Xn8ahqymgMuXPlA/KqIg54hsEj3/qHx2zvSsNIOSz+efRgCBtkVcH51wqHv04vYi 8/ME1pneG7OA1v5Ki/A9PG1xszlpXxaSA2B/Pxxm0plj30eB5SziYLFVrXzLc6s= =WJ8j -----END PGP SIGNATURE-----