**** BEGIN LOGGING AT Fri Apr 16 03:00:29 2021 Apr 16 06:33:59 $systemd_system_unitdir points to /usr/lib/systemd/system. Which one points to /etc/systemd/system? Apr 16 07:32:49 yo dudX Apr 16 07:47:04 i added some packages in my yocto build and i want to boost it via sd card. so i have to just add .rootfs.tar.gz in rootfs partition of my sd card? or i have to dd all the zImage and oftree again? thanks Apr 16 07:52:10 skskjaks: it depends. i presume that you are already booting from the sd card? or are we talking about changing the whole boot strategy? Apr 16 08:16:07 yes i am already booting from sd card Apr 16 08:41:47 and what form does the current rootfs have on sdcard? Apr 16 08:42:30 usually one does not use a tar.gz, but it being expanded to a partition. Apr 16 08:53:11 zeddii: did you have any time for kernel bsp ? Apr 16 09:38:59 dl9pf: https://autobuilder.yoctoproject.org/typhoon/#/builders/120/builds/6 Apr 16 09:51:03 scott is fixing already Apr 16 10:45:35 Hello, how can I create an initramfs and add it to a .wic image? Apr 16 10:45:35 I have tried it, although the wic image is larger now. The Kernel is not booting with it. Apr 16 10:45:36 This is my initramfs image https://paste.debian.net/plain/1193951 Apr 16 10:45:36 I have included these variables to machine.conf: Apr 16 10:45:37 INITRAMFS_IMAGE = "image-initramfs" Apr 16 10:45:37 INITRAMFS_IMAGE_BUNDLE = "1" Apr 16 10:45:38 IMAGE_BOOT_FILES_append = " ${KERNEL_IMAGETYPE}-initramfs-${MACHINE}.bin Apr 16 11:00:51 is the initrd missing or is the boot crashing ?? Apr 16 11:01:01 harunn, ^^ Apr 16 11:02:16 It is booting, but theres is no initrd in /boot Apr 16 11:03:07 Prior I had no  initrd, maybe both kernels are in the wic images? Apr 16 11:03:41 With both, I main the bundled and the "default" one Apr 16 11:05:53 never used them wic but its good question if you expsect the initrd to be bundled to the kernel image or as separete file Apr 16 11:08:38 I don't remember if we have examples for npm recipes? Basically my use case is a local build with npm that has dependencies in the registry Apr 16 11:08:42 how to you reference it at your kernel boot cmd ? Apr 16 11:10:49 I have not referenced it in the kernel boot cmd. Apr 16 11:10:50 Shouldn't the kernel "just start it" when it is bundled with the initrd? Apr 16 11:11:54 yes thats what INITRAMFS_IMAGE_BUNDLE so do Apr 16 11:14:13 wasnt there a extra kernel option to make this work in the first place Apr 16 11:17:51 did you rebuild your kernel harunn ? Apr 16 11:18:05 yes Apr 16 11:18:41 so what is in your /boot ? the kernel with initrd should be named differntly Apr 16 11:19:17 and that would be the one you refrence in u-boot either directly or symlinks or such Apr 16 11:22:36 yes, in /boot are only u-boot, the zImage and a few symlinks. Apr 16 11:22:58 well i bet zImag is a symlink .. but to which kernel Apr 16 11:24:02 To the unbundled kernel. Apr 16 11:24:15 well Apr 16 11:24:19 There has to be a "fat kernel" right? Apr 16 11:24:23 that doesnt sound right , does it ? Apr 16 11:24:38 no, but why?^^ Apr 16 11:24:57 is the name telling you which kernel is which .. but yeah the "fatter" one is likely the one with the initrd appended Apr 16 11:25:37 dunno who sets the default zImage symlink in the scripts , sorry Apr 16 11:26:12 Hello all Apr 16 11:26:21 I am trying to secure boot linux. My hardware is a freescale arm64 i.MX8M mini SOC based platform. Apr 16 11:26:50 I am using meta-security/meta-integrity to sign my rootfs. For some reason my keys are not getting Apr 16 11:26:51 loaded by linux integrity subsystem: Apr 16 11:26:51 integrity: Loading X.509 certificate: /etc/keys/x509_ima.der Apr 16 11:26:52 integrity: Problem loading X.509 certificate -126 Apr 16 11:27:01 any pointers/suggestions ? Thank you Apr 16 11:28:30 rob_w: yes, I can distinguish them Apr 16 11:29:01 harunn, so we have to fix the symlink in your case ,, Apr 16 11:30:57 I see Apr 16 12:10:51 Afternoon all Apr 16 12:48:33 o/ Apr 16 12:51:22 qschulz: I looked into SPLASH_IMAGES again, turns out I had it set in my bbappend by: SPLASH_IMAGES_append = " file://psplash-img.png;outsuffix=test" This subsequently oe-pkgdata-util find-path '/usr/bin/psplash-test', but it still does not end up installed on the rootfs. More digging inbound Apr 16 12:52:20 In other news I fixed the annoying white flash of blinding light when loading chromium and also disabled cursor on weston when a cursor plane is available (but not used) Apr 16 12:53:24 I'm starting to get the hang of yocto :) (I migrated from buildroot) Apr 16 13:03:36 escalion: you probably need to add the package returned by oe-pkgdata-util find-path to your image recipe? Apr 16 13:04:55 It's already added, that's the head scratcher Apr 16 13:05:07 escalion: yup, that seems like it, RDEPENDS_${pn}-test will be set to ${PN} automatically but that's about it Apr 16 13:05:59 I compiled with IMAGE_INSTALL_append += " psplash psplash-test" Apr 16 13:06:58 I need to add some debugging to the recipe and see whats happening. Computer tied up with another chromium build now Apr 16 13:07:19 Because I need to get ICECC set up to work with clang somehow :3 Apr 16 13:58:28 hi everyone. Is there any way to somehow specify offset in bytes in wks file (instead of kbytes) ? (https://docs.yoctoproject.org/gatesgarth/ref-manual/ref-kickstart.html) Apr 16 14:01:44 part of the toolchain build includes a gcc-runtime package. where do the files for this package come from? Apr 16 14:02:51 i do not see the typical "git" folder in the /tmp/work//gcc-runtime Apr 16 14:03:03 I've tried already partial values, like 32.5 (I have to write cust meta-data to offset 0x8200), it's neither supported, bitbake just reports ERROR: ../sdimage-imx8-spl-sota.wks:17: argument --offset: Invalid size: '32.5' Apr 16 14:04:59 correction: i do not see the typical "git" folder in the /tmp/work//gcc-runtime// Apr 16 14:05:18 khem: ? Apr 16 14:05:21 any thoughts? Apr 16 14:12:11 * yates listens to some Deep Purple Apr 16 14:17:23 iopaniuk: i haven't looked, but you might try looking at https://docs.yoctoproject.org/dev-manual/common-tasks.html#creating-partitioned-images-using-wic Apr 16 14:19:15 iopaniuk: under the hood the wks is using mkfs.FSTYPE - so any option that you can use with mkfs should be accepted. Source in poky/scripts/lib/wic/partition.py Apr 16 16:13:42 hello guys, I'm running yocto (thud) and I have installed kpartx in the rootfs Apr 16 16:14:10 if I try to list the partition inside the generated wic image on the target: Apr 16 16:49:53 Hello dear Yocto Wozard! What is the best (if possible) to configure a recipe according the image name ? The purpose is to build differently the "apps" recipe based on some variable set inside the image recipe or maybe be I have to use IMAGE_FEATURES but to sure to choose the best option... If someone can point me the best direction. Thank you! Apr 16 16:52:57 NiniC0c0: You can define applications to install for a specific image using IMAGE_INSTALL in the image definition Apr 16 16:53:07 A nice out of tree example: https://github.com/jumpnow/meta-rpi64/blob/dunfell/images/qt5-basic-image.bb Apr 16 16:56:24 NiniC0c0: Yocto chant #1 Apr 16 16:56:30 NiniC0c0: recipe data is local Apr 16 16:56:35 conf data is global Apr 16 16:56:49 so no, you cannot modify a recipe from another recipe (and image recipes are... recipes :) ) Apr 16 16:57:53 so either you go for a different distro, or you create multiple recipes per possible configuration and add the appropriate configuration in your IMAGE_INSTALL Apr 16 16:58:54 I think that would work only if your package recipe is a top-level recipe for you image recipe (meaning, explicitly added to IMAGE_INSTALL for example, not in ANY DEPENDS or RDEPENDS) Apr 16 17:00:22 qschulz always the best, Thx for support ! Apr 16 17:00:30 I personally use a custom python script on top of bitbake that pulls multiple configurations from directories and stitches them together based on an images declaration file Apr 16 17:03:32 My system is to build application sets onto different targets (and cross-targets) - it looks like this: https://imgur.com/a/bpdOevv Apr 16 17:04:19 hello guys, it's advisable to use squasfs and uinonfs nowadays ? Apr 16 17:04:59 squashfs is nice if you only need a read only filesystem because it can be mounted directly and has a high compression rate Apr 16 17:05:02 it seems to be an itneresting solution but at the same not so popular Apr 16 17:05:57 escalion, I want to create an image with a rootfs and a partition that can be rw by the user Apr 16 17:06:31 thekappe: then a squashfs system would work - but only use it if space is a premium Apr 16 17:06:55 Otherwise go with straight ext4 in ro mode and a overlay partition with rw Apr 16 17:07:16 escalion, in my case the buns is that the user won't be able to mess uo with the shipped rfs Apr 16 17:07:27 bonus* Apr 16 17:07:30 just mount the ext4 read-only Apr 16 17:07:45 whoa Apr 16 17:08:10 and overlay on /home Apr 16 17:08:32 sound interesting too Apr 16 17:08:48 It'll be much easier that way. Apr 16 17:09:12 If you want updates etc too, then rauc is pretty nice (and ships bundles with a squashfs) Apr 16 17:09:51 rauc ? Apr 16 17:10:06 rauc, meta-rauc Apr 16 17:10:51 i'll search for it Apr 16 17:12:27 Basically there are advantages and disadvantages of going the squashfs/unionfs route - it all depends really on the use case Apr 16 17:14:09 @escalion thanks Apr 16 17:14:18 im just evaluating all the possibilities Apr 16 17:14:47 Just remember that squashfs saves space at the tradeoff of a small drop in performance Apr 16 17:14:48 basically I'll need something robust and updatable Apr 16 17:15:17 having a ro fs with overlay capabilities sounds interesting Apr 16 17:15:50 It's exactly what I've just done for a project. Using rauc for A/B rootfs updates too Apr 16 17:15:55 I've searched a little bit and I got to squashfs and unionfs Apr 16 17:16:10 that would be probably my case Apr 16 17:16:26 I just call them rootfsA and rootfsB Apr 16 17:18:18 exactly. go take a look at rauc I think it could help with what you want Apr 16 17:19:58 escalion: that ain't built into bitbake/Yocto though, but I guess it is a valid scenario too. To each their own :) Apr 16 17:20:40 qschulz: it isn't but it has been certainly helpful for me to leverage yocto & bitbake to my use case Apr 16 17:23:48 Anyway, I'm off - thanks for the fish! o7 Apr 16 17:30:46 qschulz, what do you think about rauc ? Apr 16 19:32:35 Did you ever package a multiconfig image recipe? Apr 16 19:33:01 I'd like to pre install a container image in my rootfs Apr 16 19:34:38 I can maybe add a task which depends on the multiconfig image to deploy its rootfs into the main image rootfs, but maybe a simple multiconfig image -> package recipe translation would be simpler? this way you could even use SYSTEMD_SERVICE to enable the machine at boot. Apr 16 19:45:53 vdl: I've not tested it recently, but I did a couple of presentations on doing that, sample at: https://github.com/konsulko/meta-container-demo Apr 16 19:47:39 smurray: thank you. "container" image types require packagegroup-core-boot? Apr 16 19:48:25 vdl: no Apr 16 19:49:03 smurray: you added it in build-container.bb Apr 16 19:50:06 vdl: I'd have to dig a bit to see why I did that, it's not a hard requiremnt Apr 16 19:50:28 smurray: btw for the volatile issue (machinectl start foo failed for me as well because /var/log/ was non existent), setting VOLATILE_LOG_DIR = "no" in my (container) distro conf fixes it. Not sure if that can be applied in the container image recipe itself. Apr 16 19:51:53 vdl: as I said, I've not tried it since last Feb, it's possible stuff has changed in gatesgarth and later Apr 16 19:55:09 smurray: ok. I'm currently using a dedicated machine configuration for the containers, because I use linux/dummy for them, but it does not make a lot of sense since a container must be build for the same architecture as the host... Do you know if PREFERRED_PROVIDER_virtual/kernel = "linux/dummy" can be set in the (container) image recipe? Apr 16 19:55:58 otherwise a container distro configuration might be enough I suppose. Apr 16 19:57:50 you can't set it in the image recipe. I've done both setting it in a tweaked machine for the container or in the multiconfig .conf Apr 16 19:58:30 container distro also works Apr 16 20:00:01 maybe the multiconfig is better to centralize these common machine/distro tweaks while still sharing the same machine and distro definitions. Apr 16 20:05:26 smurray: long term for this would be having dynamic packages (similar to kernel module packages) for images of "container" fstype, so one could IMAGE_INSTALL += "container-" and eventually bbappend the package to override the install path or install/enable an .nspawn file. Apr 16 21:33:20 I am bringing a kernel 2.6 into yocto (the exact source tree that a vendor provided) Apr 16 21:35:07 Ostensibly, the code is 2.6.35, and I'd like to diff it with source from kernel.org to see how much mucking around the vendor did. But kernel.org shows 9 revisions for that, from 2.6.35.1 to 2.6.35.9 Apr 16 21:35:50 If I download all nine of them, unpack them, and commit them to a repo in series, is there a quick way to find out which commit has the least edit-distance to the vendor-provided source tree I have? Apr 16 21:37:44 For only nine commits this is something I could do manually, but I'm wondering if there's any tooling that would do this kind of task more automatically before I write my own Apr 16 21:41:11 The script I would write would commit the vendor source tree into another branch, seperate from the successive kernel.org commits. Then I'd call 'git diff vendor_branch kernelorg_commit_hash' for each kernel.org commit hash, then save the diff into a text file with the name of the commit. The smallest text file (by lines?) probably means the fewest differences and should indicate whether the vendor started Apr 16 21:41:17 with 2.6.35.1 or .35.7. Then I could rebase the improvements in the later revisions onto the vendor branch. Apr 16 21:47:07 yates: gcc builds are a bit special, where it shared same source tree with many gcc related recipes, so sources are under work-shared Apr 16 21:47:18 owned by gcc-source recipe usually Apr 17 00:04:06 I'm struggling a bit with my rpi4 and wifi. Does somebody here know which kernel module needs to be started for that? I tried `modprobe b43`, but no wifi interface appears in `ip l` after that... Apr 17 01:12:09 just discovered `cfg80211`, but `modprobe cfg80211` does not bring up a wlan0 either... Apr 17 01:28:51 RP: my musl/mips patch is still forgotten :( Apr 17 01:30:34 jonesv: try. brcmfmac Apr 17 01:31:11 khem: hehe I was just trying that, it seems to do it :) Apr 17 01:31:17 khem: thanks a lot! Apr 17 01:32:58 I always struggle a lot for those kernel modules, I go with trial and error and it takes me a few hours to find which ones to use Apr 17 02:11:30 Hello .. I don't know if here is the correct place to mention ... but it seems (to me) that downloads.yoctoproject.org is timing out :) Apr 17 02:30:22 theodiem, I believe there maybe some Maintenance occurring right now so it may be affecting downloads. Apr 17 02:32:25 thanks. I've managed to get through it and proceeded. **** ENDING LOGGING AT Sat Apr 17 02:59:57 2021