**** BEGIN LOGGING AT Wed Apr 08 03:00:04 2020 Apr 08 03:14:42 denix: Trying to boot bbb I am seeing that uboot is not booting from SD it worked fine few months ago U-Boot 2020.01-gae8ceb7b6e (Apr 08 2020 - 02:53:31 +0000) Apr 08 03:15:03 denix: any magic cmds I might be missing Apr 08 03:15:29 khem: env default -a? Apr 08 03:16:31 did not help Apr 08 03:16:40 Its picking u-boot from SD fine Apr 08 03:17:30 https://hastebin.com/ojehalezul.cs Apr 08 03:20:46 hmm, I can take a look tomorrow Apr 08 03:42:46 denix: ok, Apr 08 04:57:33 denix: I see the problem is https://git.yoctoproject.org/cgit/cgit.cgi/meta-ti/commit/?id=7646efc16a2b0aeafb6790a7d18b8fa2ca4547ad Apr 08 04:58:08 denix: I have a fix which is working http://sprunge.us/OBrLIz Apr 08 05:00:46 denix: if you want to keep the order of partitions then you have to change the root= cmd and bootloader cmd to change rootfs partition to point to /dev/mmcblk0p3 Apr 08 06:33:34 khem: ah, right. needed to add 3rd partition for testing - normally it gets created last one at the end of SD card via scripts. but I wanted wic to have expandable rootfs, so I swapped the order at the last minute... Apr 08 07:32:30 khem, denix: I hit the same problem yesterday Apr 08 07:34:16 Having issues on the Sancloud BBE now we've updated to zeus and wanted to confirm that Arago for the Beaglebone Black works, hit that and some other issues Apr 08 09:21:26 paulbarker: I am thinking of how to merge embed-rootfs and include-path. The first start would be allow 1 or two arguments for include-path. If 1 then old behaviour, if 2 it is ORIGIN DESTINATION. The issue is when I want to add the concept of a rootfs. The simple ways would be: if it is a file/directory, do a simple copy, but if it is a real rootfs take into consideration the pseudo database. Apr 08 09:22:04 something like: --include-path core-image-minimal /export Apr 08 09:23:16 but then I have a clash, what if there is a file/directory named core-image-minimal ? shall I use the image directory tmp/work/machine... or ./core-image-minimal ? Apr 08 09:26:53 ribalda: Email the list and Cc me, I'll get back to you later today or maybe tomorrow Apr 08 09:27:53 sure, I already posted a v3 for the permission fix. There is absolutely no hurry. Apr 08 09:27:55 Even if no pseudo database exists we should do something with pseudo, else the files copied will be owned by the user running wic Apr 08 09:27:55 thanks! Apr 08 09:28:45 Maybe leave files owned by root if no pseudo database exists and clear setuid/setgid bits for safety Apr 08 09:29:54 I think supporting wic running both inside and outside of bitbake is a major complication, if it were up to me I'd only support one or the other Apr 08 09:31:19 if we only support from inside bitbake, then you cannot have disk with multiple images/arches Apr 08 09:32:16 if we allways explicitely declare the pseudo dir that we use, and stop using the one from the environment it will work just fine Apr 08 09:32:20 on both situations Apr 08 09:32:51 You can combine multiple images/arches inside bitbake, you just need to write a recipe to do it Apr 08 09:34:30 Ignoring the pseudo dir given in the env will likely break usage inside bitbake Apr 08 09:34:52 https://git.openembedded.org/openembedded-core/tree/scripts/lib/wic/partition.py#n200 for me is the root of our evil :) Apr 08 09:35:35 on rootfs.py, you know exactly which one is the pseudo directory, so you can pass that to partition.py. Apr 08 09:36:51 I'll take a look later, got to get back to this client work... Apr 08 09:37:32 sure no problem. if you can share a .bb that combines multiple images with different MACHINE I would love to see that, because that will solve my problem until we get this fixed Apr 08 09:37:43 thanks Apr 08 15:54:06 denix: paulbarker I was wondering if wic has an option to specify 'allocate this partition at the end of disk' Apr 08 15:55:30 khem: Not sure, you'd need to check https://git.openembedded.org/openembedded-core/tree/scripts/lib/wic/plugins/imager/direct.py#n359 Apr 08 16:12:37 yeah it seems all fixed partitions should come before the one we want to resize Apr 08 16:37:22 khem: when you create wic image, you don't know how large your SD card will be. and when you want to expand it, you can only resize the last partition w/o damaging others - gparted resizepart + resize2fs. Apr 08 16:40:09 but you can reserve partition slots? I think I did that, i.e. have two 2GB partitions, the actual filesystem for both will only be as big as the rootfs, but you can then resize2fs Apr 08 16:40:40 --fixed-size should reserve the partition space Apr 08 16:41:27 or did I misundertsand the original question? :> Apr 08 17:51:14 denix:right thats what we do in yoe distro for rpi and bbb and some other developer SBC images Apr 08 18:02:43 webOS OSE does the same with systemd service executed on first boot Apr 08 18:17:52 JaMa: we normally do it offline on the host. can I see your systemd service for doing it on boot? Apr 08 18:31:12 denix: use 96boards-tools from meta-96boards layer Apr 08 18:31:25 it works pretty well on systemd and non systemd cases Apr 08 18:31:39 heh, beat me to mentioning it ;) Apr 08 18:31:42 or use yoe distro :) Apr 08 18:31:56 systemd 245 has something built in now, I think Apr 08 18:31:58 if you dont want to bake but eat the cake Apr 08 18:32:14 so maybe an option in 3.2 Apr 08 18:32:44 yeah systemd-growfs Apr 08 18:33:10 I think they should implement systemd-do-what-i-mean and be done with it Apr 08 18:33:24 LOL Apr 08 18:34:45 ah its now called systemd-repart Apr 08 18:34:49 denix: you mean after writing the image to uSD? Apr 08 18:34:59 khem: rofl Apr 08 18:35:17 JaMa: yeah Apr 08 18:35:26 https://github.com/webosose/meta-webosose/commit/fa114668e650837e1094eb5938f88bd8eec161e2 Apr 08 18:36:38 https://github.com/poettering/systemd/commit/533ac98255947310299db3381e259d24fb3ac2e9#diff-91be33be198dcb660ec3e06b561540db Apr 08 18:36:48 heh, seems like a candidate for oe-core if everyone is reinventing it ;) Apr 08 18:37:40 smurray: indeed Apr 08 18:38:02 yeah, mentor's using 96boards-tools too Apr 08 18:40:20 khem: heh, seems like you're constrained to filesystems that support online resizing with the systemd stuff Apr 08 18:40:24 I have been fixing it for non-systemd cases too Apr 08 18:41:06 smurray: yeah you can only fix what you got with repart kind of approach Apr 08 18:41:37 need to reboot into shiny 5.6 kernel brb Apr 08 19:16:33 denix: so what fix are you going for fixing the wic issue ? Apr 08 19:16:51 perhaps a revert might be ok too unless you have plans for different layout Apr 08 19:17:45 khem: no reverts - layout has to stay :) Apr 08 19:17:59 khem: would need to pass root= Apr 08 19:20:54 you are forcing people to buy larger SD cards Apr 08 19:21:06 or perhaps override default wic Apr 08 19:21:11 khem: btw, it sets WKS_FILE conditionally with ?= so you can easily override it back to sdimage-bootpart.wks for now Apr 08 19:21:51 denix: yeah I have distro override for all SBCs so have same wic format Apr 08 19:22:06 so they look almost same to user Apr 08 19:26:40 khem: yeah, I was debating whether to add it to distro or to bsp. the thing is our test team wants to use it regardless of distro, hence had to add to meta-ti Apr 08 19:27:57 its fine as long as it does not come in distro's way however, I wonder if you change boot params to use different partition number how it will reflect Apr 08 19:28:17 my suggestion would be to not force end users do what your test team wants Apr 08 19:43:11 khem: well, it's not that clean. I see more and more BSPs define their own layouts Apr 08 19:43:33 http://git.yoctoproject.org/cgit/cgit.cgi/meta-rockchip/tree/wic/rk3399-boot.wks and http://git.yoctoproject.org/cgit/cgit.cgi/meta-rockchip/tree/wic/rock-pi-4.wks Apr 08 19:45:06 khem: with all the bootloaders, FIT images, DTBs, some reserved partitions, etc. Apr 08 19:47:11 khem: BBB may seem simple and only require boot+root. but there are other platforms more complicated - more changes will be coming Apr 08 19:47:37 khem: but the good news is that there's always the way to override it by distros Apr 08 19:49:26 denix: people having to now do extra work to get back to what's been the default partitioning for BBB for so long seems a bit odd, tbh Apr 08 19:54:39 smurray: actually 3-partition SD cards have been the default for BBB since the beginning :) Apr 08 19:55:34 smurray: as a product, 3rd partition's been used for documentation and initial setup Apr 08 19:55:48 denix: in rcn's stuff, you mean? Apr 08 19:58:30 denix: perhaps I'm misremembering what I've gotten using bitbake for my own images in the past, been a while Apr 08 19:58:51 smurray: you mean debian? no, it's not specific to debian. same was the case for production SDKs from TI Apr 08 19:59:51 denix: so this change to meta-ti is a no op, or pushing that down from some non-meta-ti product? Apr 08 20:00:08 smurray: but production SDKs were not using wic until now, so wic was staying barebone for a while Apr 08 20:00:48 denix: where I'm coming from is whether picking this up will now need fixing stuff for the bbb and bbe images in AGL Apr 08 20:00:55 denix: it's sounding like it will Apr 08 20:03:00 looks like AGL overrides the wic file for one of them but not the other, I'll make a note in case there's an issue when we pick this up Apr 08 20:07:35 since when do we get something on SOC SDKs that is product ready ? Apr 08 20:07:50 I was just hoping for one Apr 08 20:09:09 perhaps an installer like raspian might be a good onboarding experience Apr 08 20:13:39 khem: product as in reference :) Apr 08 20:13:50 khem: like RDK :) Apr 08 20:14:00 not a final product Apr 08 21:09:18 denix: http://errors.yoctoproject.org/Errors/Build/100947/ this is bbb world build Apr 08 21:09:34 some issues are perhaps meta-ti alone you might be interested in those Apr 08 21:14:23 product meant product, reference for sure, R in RDK says that too Apr 08 21:15:59 denix: can we enable CONFIG_NET_UDP_TUNNEL in bbb kernel ? this will make wireguard work out of box Apr 08 21:20:00 khem: from our perspective, SOC is a product and SDK that enables it is also a product. obviously, it's not a retail product - you can't buy it in a supermarket... Apr 08 21:21:05 khem: config fragments are supported, so all it takes is to add a file with CONFIG_NET_UDP_TUNNEL=y and SRC_URI += it Apr 08 21:58:47 denix: yeah already have it here, was just mentioning if you wanted some out of box experience with wireguard on bbb **** ENDING LOGGING AT Thu Apr 09 02:59:57 2020