From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from blaine.gmane.org (195-159-176-226.customer.powertech.no [195.159.176.226]) by mail.openembedded.org (Postfix) with ESMTP id 258587F7EF for ; Sun, 27 Oct 2019 22:27:21 +0000 (UTC) Received: from list by blaine.gmane.org with local (Exim 4.89) (envelope-from ) id 1iOr0L-0005rz-Sg for openembedded-core@lists.openembedded.org; Sun, 27 Oct 2019 23:27:21 +0100 X-Injected-Via-Gmane: http://gmane.org/ To: openembedded-core@lists.openembedded.org From: Ferry Toth Date: Sun, 27 Oct 2019 23:27:14 +0100 Message-ID: References: <47af37ed-25aa-65aa-cea7-febfada814f7@gmail.com> Mime-Version: 1.0 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Thunderbird/68.1.2 In-Reply-To: Cc: public-openembedded-core-ZwoEplunGu3dfDuKDZ/zN51Ccm5ICvs9@plane.gmane.org, public-public-openembedded-core-ZwoEplunGu3dfDuKDZ/zN51Ccm5ICvs9-wOFGN7rlS/M9smdsby/KFg@plane.gmane.org Subject: Re: zimage Initramfs booting stuck at Start Kernel X-BeenThere: openembedded-core@lists.openembedded.org X-Mailman-Version: 2.1.12 Precedence: list List-Id: Patches and discussions about the oe-core layer List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 27 Oct 2019 22:27:22 -0000 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Content-Language: en-US Op 27-10-2019 om 22:52 schreef JH: > Hi Ferry, > > On 10/27/19, Ferry Toth wrote: >> Op 27-10-2019 om 02:27 schreef JH: >>> Hi Ferry, >>> >>> On 10/27/19, Ferry Toth wrote: >>>> >>>> As an example you might want to look at my repo, I have 2 images minimal >>>> and initramfs: >>>> https://github.com/edison-fw/meta-intel-edison/tree/master/meta-intel-edison-distro/recipes-core/images >>>> >>>> I don't exactly know how you are doing it. But I'll try to explain what >>>> I do: >>>> I have >>>> 1) a normal rootfs (say edison-image-minimal) and >>>> 2) a initramfs rootfs containing kernel modules needed for booting and a >>>> init script. >>>> >>>> Initramfs is build using core-image-minimal-initramfs.bbappend. Some >>>> additional unneeded packages are pulled, and as size is restricted (10MB >>>> or so?) we remove them here. >>> >>> Thanks for the tips, that confirms my long suspensions that the size >>> of the image could be the cause of the problem, my working zImage size >>> is 8MB, my failing zImager-initramfs size is 35MB. How did you remove >>> the 10 MB size restriction? I checked your >> >> I don't. I reduce the contents as much as possible. > > That is not practical, we cannot limit the kernel + rootfs to 10 MB size. > >>> core-image-minimal-initramfs.bbappend, there is no sign to remove 10 >>> MB size restriction. You have a resize-rootfs.service, but I am not >>> sure that is the one for fixing the 10 MB size restriction. The >> >> No, it isn't. >> >>> edison-image-minimal.bb has the IMAGE_ROOTFS_SIZE = "1048576", sorry >>> for my dumb, is that size for 1048576 KB (1GB) or 1048576 B (1MB)? >> >> The rootfs is on disk (/) and is 1GB. It has nothing to do with initramfs. > > OK. > >>>> The init script is here: >>>> https://github.com/edison-fw/meta-intel-edison/blob/master/meta-intel-edison-distro/recipes-core/initrdscripts/initramfs-framework/init >>>> >>>> The init script loads modules, waits for the disk with the rootfs to >>>> appear and mount it. Then switchroot to it. >>>> If it doesn't appear it drops to a shell so you can investigate what's >>>> wrong. >>> >>> I am confident that the 10 MB size restriction is the problem, I am >>> not clear where that 10 MB size restriction is defined and how to fix >>> the size restriction, appreciate your helps. >> >> I am not sure where it comes from, I think internal in the kernel there >> is a limit, after uncompress. Initramfs is in ram, I think we just need >> to accept there is a size restriction. > > No, there is no limitation in kernel, BTW, the zImage-initramfs is not > just kernel space, the zImage-initramfs = kernel + rootfs which cannot > be limited to 10 MB. I have a zImage-initramfs created by openwrt, the > size is 28 MB. The problem is created by oe-core, most likely in > kernel.bbclass, could anyone in oe-core development provide insights > where is the 10 MB limitation from and how to fix it? I see that Ubuntu keeps the initramfs in a separate cpio, while we are trying to build the cpio into the kernel (afaiu the cpio is unpacked into a kernel directory, and then built-in by the kernels build system). And that cpio is ~64MB, so it must be possible. In my case kernel, cpio and zImage-initramfs are all built. And although U-Boot allows to load kernel and cpio separately I didn't try (or don't remember the result). So, maybe that's the trick. >>>> To build everything edison-image-minimal triggers build of u-boot, >>>> kernel, rootfs, initramfs and kernel-with-built-in-initramfs. >>> >>>> Search for 'init' and 'live' and find the needed lines in >>>> https://github.com/edison-fw/meta-intel-edison/blob/master/meta-intel-edison-distro/recipes-core/images/edison-image-minimal.bb >>> >>> I looked at the init script, but not clear if it is critical for >>> fixing that issue or not. My init is the symblic link to >>> /lib/systemd/systemd in imx6. >> >> Given the size restriction I think we have no choice to use initramfs to >> load minimum required modules and then mount and switchroot to a file >> system on disk. So, yes, you need to modify init for that. After >> switchroot there will be a new init, the symbolic link you mention. > > That is not good, we need to fix it. I am still new to oe-core, > bitbake and Yocto, but I am happy to give a try if someone could point > me where could be the source for the 10 MB limitation of > zImage-initramfs. > > Thank you very much. > > Kind regards, > > - jh >