From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?UTF-8?B?VmluY2VudCBTdGVobMOp?= Date: Thu, 19 Mar 2015 16:37:35 +0100 Subject: [Buildroot] [PATCH] configs: add defconfig for Freescale i.MX31 PDK In-Reply-To: <550A0AF9.5050101@mind.be> References: <1426585422-22441-1-git-send-email-vincent.stehle@freescale.com> <550A0AF9.5050101@mind.be> Message-ID: <550AED3F.4070606@freescale.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net On 03/19/2015 12:32 AM, Arnout Vandecappelle wrote: .. > Doesn't the upstream MACH_MX31_3DS work? 2.6.28 is extremely old... Hi Arnout, Thank you for reviewing this patch so quickly and sending feedbacks. This defconfig is indeed based on an old 2.6.28 kernel; the idea was to base the defconfig on the last Freescale "official" release, which works fine for me. You are right that mainline kernel has support for the i.MX31 processor, but there is no dts in there for the i.MX31 PDK right now. The few tries I just did do not boot "as is" but I will continue a bit. Do you insist that this i.MX31 config uses mainline, or would it be acceptable to stay on the "old", Freescale "official" kernel? .. >> +BR2_KERNEL_HEADERS_3_2=y > That doesn't sound like a good idea when the kernel is 2.6.28... Granted, this is not optimal. Those are the kernel headers closest to 2.6.28, "easily" available in buildroot. It is indeed possible to manually specify the 2.6.28 kernel headers version, but this necessitates two patches to fix the linux-headers "build". Do you prefer this solution? .. >> +BR2_TOOLCHAIN_BUILDROOT_EGLIBC=y >> +BR2_TOOLCHAIN_BUILDROOT_CXX=y > Why not uClibc and why C++? Those are "leftovers", sorry. Thanks for spotting those mistakes. I will remove that. .. >> +BR2_ROOTFS_DEVICE_CREATION_STATIC=y > With a more recent kernel we don't need this crappy option... This is correct: this is necessary only with the "official" release kernel. .. >> +# BR2_TARGET_GENERIC_REMOUNT_ROOTFS_RW is not set > Why? Also "leftovers", sorry. Thanks for spotting, I will remove as well. .. > If network boot is really the only option, then perhaps it would be nicer to > use an initramfs linked into the kernel? Or doesn't the board have enough memory > to support that? This is my only boot method for the moment. Do you insist that the defconfig should use MMC? I will continue to try a bit to generate a suitable MMC with buildroot anyway. Maybe that can be changed in the defconfig later on? On the other hand, if we stay with network booting, initramfs is a nice idea as it makes the setup simpler indeed. Also, the i.MX31 PDK has 128 MB of memory, so this is a practical solution. I just tested and an initramfs included in the zImage works fine. Only, Freescale "official" kernel has no support for initrd/initramfs, so we need another patch. Would you prefer this solution? I will send a v2 patch right away, with a few reworks discussed here, and still based on the "old" Freescale "official" release. Please let me know if this is going in the right direction (or not ;) Thanks again for reviewing! Best regards, V.