**** BEGIN LOGGING AT Sat Jul 30 02:59:57 2022 Jul 30 08:25:05 Morning! Jul 30 08:25:30 Herrie: moboot also boots fine my webOS on TP, so I'll put in place the new binary in /boot Jul 30 08:30:06 done Jul 30 08:54:53 Tofe: I put it on my Go too Jul 30 09:10:42 Just battery was empty Jul 30 09:10:51 So put it to charge for a bit Jul 30 09:13:18 Tofe: Make sure to have this commit in your moboot build: 9fed1761855dbb97a6e8f7fd85e4275c19b65b75 Jul 30 09:13:40 I have 2 branches (a regular Go and Go-Dev branch. This commit is only in my Go-Dev branch) Jul 30 09:13:55 This commit will enable verbose + serial console by default Jul 30 09:14:05 So you don't need to push the files moboot.sercon.LuneOS etc Jul 30 09:14:37 ah, yes, good idea Jul 30 09:16:07 mmh doesn't help that much in my case Jul 30 09:16:23 stuck in "Booting...", not sure if something is running Jul 30 09:17:25 cmdline isn't so great, I just see: concerning the console, there is just "console=tty1", no serial or anything Jul 30 09:17:39 could be the defconfig here Jul 30 09:20:21 yes, the 5.7 defconfig from mbd doesn't have any CMDLINE in the defconfig, so I restored the previous value and I'm rebuilding Jul 30 09:21:40 i also removed the KERNEL_EXTRA_ARG: looks like it buiilds fine without it? Jul 30 09:23:50 Tofe: OK well I specify the following in my kernel recipe: CMDLINE = "LUNEOS_NO_OUTPUT_REDIRECT console=ttyMSM0,115200,n8" Jul 30 09:23:58 Which should give console, no? Jul 30 09:24:46 well looks like this is overriden by https://gist.github.com/mbd98/dc973ade2746b5589ab19869d2b31ca8#file-config-L473 somehow Jul 30 09:25:06 we'll see after the kernel rebuild Jul 30 09:25:12 Ah yes it could be Jul 30 09:25:37 we also had CONFIG_CMDLINE_FROM_BOOTLOADER=y which I put back Jul 30 09:29:26 cmdline is better now, but still nothing in minicom :( Jul 30 09:30:26 ah no wait, it says console=ttyS0, not ttyMSM0 Jul 30 09:30:42 damn, where does that come from Jul 30 09:35:40 Tofe: That's what moboot outputs Jul 30 09:35:47 But it's not enforced Jul 30 09:35:59 oh, ok Jul 30 09:38:00 Comes from here: https://github.com/Herrie82/moboot/blob/herrie/go-dev/lib/bootlinux/bootlinux.c#L40 Jul 30 09:38:44 But I had serial output even when moboot was displaying this, so it's misleading really Jul 30 09:38:47 It's a hardcoded value Jul 30 09:39:40 https://github.com/Herrie82/moboot/blob/herrie/go-dev/lib/bootlinux/bootlinux.c#L247 Jul 30 09:40:02 but we have also CONFIG_ARM_ATAG_DTB_COMPAT_CMDLINE_FROM_BOOTLOADER=y Jul 30 09:40:57 ok I've disabled the CMDLINE from bootloader Jul 30 09:41:18 which is incorrect for moboot anyway Jul 30 09:41:30 rebuild in progress... Jul 30 09:41:49 and I've changed inherit kernel_android into inherit kernel, long overdue Jul 30 09:42:11 Tofe: Yeah there needs to be some tweaking there Jul 30 09:42:22 But seems we're getting close to something that should start to boot and give logs now :D Jul 30 09:45:48 I hope so! Jul 30 09:50:34 nope :/ Jul 30 10:01:13 I got logs! I hacked a bit more moboot's cmdline Jul 30 10:01:50 Herrie: https://paste2.org/Ph44yHBm Jul 30 10:02:58 Tofe: init_functions.sh: line 56: /sys/kernel/uevent_helper: Permission denied Jul 30 10:03:02 There's a defconfig flag for that Jul 30 10:03:12 yes, but it's not that fatal Jul 30 10:03:24 the real reason is the md init Jul 30 10:03:31 dm* Jul 30 10:05:43 VIRTUAL-RUNTIME_android-initramfs-scripts = "initramfs-scripts-android" Jul 30 10:05:43 is incorrect; this is dedicated to android devices with a userdata and such Jul 30 10:06:00 we're more related to the pinephone approach here Jul 30 10:08:14 we need initramfs-scripts-simple from meta-pine64-luneos Jul 30 10:08:29 time to move that one somewhere else... Jul 30 10:09:30 Tofe: But you also use it for HAmmerhead? Jul 30 10:09:45 The naming and the comments in the recipe looked funny to me already LOL Jul 30 10:10:02 no, hammerhead is an an android device, in the sense that the partitions are dedicated to android from the start Jul 30 10:10:10 Ah OK Jul 30 10:10:44 also it's booted by fastboot, though it should have an influence on initrd here Jul 30 10:10:49 +not Jul 30 10:11:53 OK Jul 30 10:11:59 not sure where to put this initramfs-scripts-simple recipe... meta-smartphone/meta-android just looks confusing Jul 30 10:12:00 I took hammerhead mainline as rexample Jul 30 10:12:15 meta-smartphone/meta-mainline ? Jul 30 10:12:25 exactly what I was going to propose Jul 30 10:12:27 Doesn't exist yet, but we can create that of course Jul 30 10:12:36 JaMa: ^ no problem with that ? Jul 30 10:15:28 Tofe: I suggest we do that for now for testing, if JaMa has objections we can always move it somewhere else Jul 30 10:15:35 But to me the most logical place Jul 30 10:15:44 already on it Jul 30 10:19:09 At least it doesn't crash the kernel right away, so that's already something positive :P Jul 30 10:20:36 yes, and it also starts initrd, which gives us a bit of space to tinker Jul 30 10:29:37 There will be challenges ahead with the kernel itself for sure, but it's not a bad start Jul 30 10:38:56 Tofe: For sometime later came across this branch from z3ntu for msm8974 (Hammerhead): https://github.com/z3ntu/linux/commits/qcom-msm8974-5.19.y-fixes Jul 30 11:28:01 Herrie: yep Jul 30 11:28:33 for the moment, on TP, I'm stuck after mounting boot and before (I think?) starting lvm.static Jul 30 12:06:36 Tofe: OK. Didn't we do something for lvm.static specifically for Tenderloin in the initramfs? Jul 30 12:07:12 no strong objection, it's pity to create whole layer just for 1 recipe (maybe a few later), but probably better than duplicating it in meta-hp and meta-pine Jul 30 12:08:11 my TP, didn't catch fire yet, but also still doesn't start, switched back to original charger and different cable Jul 30 12:08:44 JaMa: It's picky when it comes to cable Jul 30 12:08:58 The original Palm/HP cables were actually very decent Jul 30 12:10:01 is there any marking on cable? Jul 30 12:15:00 No the HP one has the round shiny silver on both ends on the connector Jul 30 12:15:26 https://image.coolblue.nl/600x315/products/69471 Jul 30 12:18:39 Tofe: You remember we have init-tenderloin.sh in initramfs-scripts-halium right? Jul 30 12:20:41 We also might not need this on mainline? https://github.com/shr-distribution/meta-smartphone/blob/master/meta-hp/recipes-core/initrdscripts/initramfs-scripts-halium/init-tenderloin.sh#L82-L84 Jul 30 12:53:40 Herrie: thanks, that's the one which I don't have anymore.. will let it charge for another day, then disassemle to jump-start Jul 30 12:59:41 Herrie: this .sh was not in use with halium; however, we had a startup script and a bottom script Jul 30 13:00:22 in the bottom script we had these lines Jul 30 13:00:58 but here, I'm stuck during the S01-mount-boot.sh script just after mounting /mnt/boot Jul 30 13:04:22 Tofe: Ah OK Jul 30 13:04:44 I guess it doesn't hurt to clean up these scripts then Jul 30 13:04:49 If not used Jul 30 13:08:16 well, I'll first try to have it running Jul 30 13:57:33 Tofe: It might be a good idea now to compare 3.4 defconfig with current one now Jul 30 13:57:55 Could be there are some fs or uevent flags in there that are needed but that we don't have now Jul 30 13:58:39 At least that would be my first guess Jul 30 13:58:57 Because lvm technically shouldn't fail normally Jul 30 13:59:09 Unless you got a locked partition somehow maybe? Jul 30 13:59:33 But we saw other messages further down then if I recall correctly Jul 30 13:59:39 well, it looks like I have issues with the logs being redirected in the wrong way Jul 30 13:59:43 so I'm deactivating that Jul 30 14:01:09 What do you mean? They don't get to serial console but somewhere else? Jul 30 14:02:06 ah, finally: https://paste2.org/9CDP6j2Y Jul 30 14:04:21 Well that gives leads :) Jul 30 14:04:37 yes, it's mount_root_partition which fails Jul 30 14:05:44 ah, I expect luneos-rootfs, but it's named luneos-root...damn Jul 30 14:09:39 "Welcome to LuneOS 1.0!" Jul 30 14:10:44 :D Jul 30 14:11:21 https://paste2.org/2bpHjK1H Jul 30 14:12:26 I'm logged in, through the cable Jul 30 14:14:33 Lots of rcu logspam Jul 30 14:14:38 yes Jul 30 14:23:34 ah battery is dead :/ Jul 30 14:26:11 Time for a break or to continue on Go :P Jul 30 14:27:34 well I don't even have a luneos partition on the Go Jul 30 14:28:28 Tofe: Same here actually Jul 30 14:40:25 https://github.com/shr-distribution/meta-smartphone/compare/kirkstone...Tofee:meta-smartphone-1:kirkstone/mainline-tenderloin?expand=1 and https://github.com/webOS-ports/meta-pine64-luneos/compare/tofe/kirkstone?expand=1 Jul 30 14:43:18 (and bblayer too, of course) Jul 30 14:50:10 thnx Jul 30 14:56:14 let me merge it :) Jul 30 14:56:40 to do something useful while qtwebengine is rebuilding Jul 30 15:00:16 merged with additional fix for layer.conf files, please try to use savedefconfig when refresing defconfig (much easier to review) Jul 30 15:08:25 JaMa: Thnx Jul 30 16:17:35 JaMa: In this case we completely switched defconfig Jul 30 16:18:06 From Dragonboard one to one known to work with Tenderloin Jul 30 16:28:55 Tofe: Both defconfigs for 5.7 and 5.15 have quite some stuff that's not needed really Jul 30 16:31:02 The 5.15 defconfig has 368 values that don't exist in the 5.7 defconfig. The 5.7 defconfig has 1123 values that don't exist in 5.15 :S Jul 30 16:32:48 So I'm not sure what they used for those, but seems they could be cleaned up quite a bit Jul 30 16:40:26 These 719 values are in the 3.4 defconfig, but not in the 5.7 one: https://bpa.st/EQHA Jul 30 17:02:41 Herrie: well, I'm all ok for cleaning it up, as long as it works as well as now Jul 30 17:05:45 Tofe: I'm not sure where to start :P Jul 30 17:05:58 Just the 5.15 has quite some wrong values for sure Jul 30 17:06:48 I.e. it has these: CONFIG_ARCH_OMAP3, CONFIG_ARCH_OMAP4, CONFIG_SOC_OMAP5, CONFIG_SOC_AM33XX, CONFIG_SOC_AM43XX, CONFIG_SOC_DRA7XX to name a few Jul 30 17:07:10 So not sure what's the best one to start with to be honest Jul 30 17:07:35 Maybe take the qcom_defconfig and add what's needed there instead? Jul 30 17:12:57 My new image is building here (quite a rebuild like JaMa predicted some days ago) Jul 30 17:21:13 I don't see anything that shocking in 3.4 defconfig that seems to be lacking in 5.7 though. At least nothing quickly I can tie back to the issues you're seeing Jul 30 17:22:13 Tofe: I see this in your log: Failed to start Enable SWAP from LVM partition Jul 30 17:25:08 Seems that's this one: https://github.com/shr-distribution/meta-smartphone/blob/master/meta-hp/recipes-core/systemd/systemd-machine-units/tenderloin/tenderloin-swap.service Jul 30 17:35:52 ah, well, there's plenty of things to fix in these logs yes Jul 30 17:36:19 also /dev was a bit too empty to my taste Jul 30 17:37:26 Tofe: Well the defconfig was really basic so not surprised Jul 30 18:31:35 Herrie: that's the point of savedefconfig (to remove quite most of the irelevant churn between 5.7 and 5.15 and to save only the non-default options Jul 30 18:32:03 FWIW my tenderloin build failed with ERROR: linux-hp-tenderloin-5.19-rc7+gitrAUTOINC+f55df88ad1_6d76455e1d-r0 do_deploy: Required initramfs uImage /OE/build/luneos-langdale/webos-ports/tmp-glibc/deploy/images/tenderloin/initramfs-uboot-image-tenderloin.cpio.gz.u-boot is not available! Jul 30 18:32:52 JaMa: Well we "inherited" some defconfigs from kernel devs who attempted to mainline tenderloin Jul 30 18:33:00 Though config options seem to vary a lot Jul 30 18:33:21 JaMa: Error looks like a mistake in Tofe's recent changes somewhere Jul 30 18:37:53 it's probably missing '.rootfs' due to my oe-core artifacts changes Jul 30 18:42:48 jup that was it Jul 30 19:40:12 JaMa: ah, yes, I didn't rebase my build latetly Jul 30 20:27:36 Herrie: "Booting Linux using a console connection that is too slow to keep up with the boot-time console-message rate. For example, a 115Kbaud serial console can be -way- too slow to keep up with boot-time message rates, and will frequently result in RCU CPU stall warning messages." https://www.kernel.org/doc/Documentation/RCU/stallwarn.txt Jul 30 20:28:03 so we shouldn't worry too much about those I think Jul 30 20:28:28 Tofe: Ah OK Jul 30 20:28:31 Makes sense Jul 30 20:28:39 Then why they put a 115200 baud rate :D Jul 30 20:29:17 PPP has 1500000 ;) Jul 30 20:29:24 Which is almost 10 times faster Jul 30 20:29:27 But it's an exception Jul 30 20:29:53 I don't know :p hardware limitation maybe ? Jul 30 20:30:40 Yeah well.. Why can it boot faster compared to what console can outptu... Silly Jul 30 20:31:07 Anyway journalctl should be fine I guess Jul 30 20:31:16 Since you can login you can generate that? Jul 30 20:35:18 sure, then it's a bit of a hassle to retrieve the file, but it's possible :p Jul 30 21:03:47 Tofe: Yeah true **** ENDING LOGGING AT Sun Jul 31 02:59:56 2022