**** BEGIN LOGGING AT Mon Jan 20 02:59:57 2020 Jan 20 08:41:40 Morning Jan 20 08:41:54 Tofe: The vendor.img is on the device now, but mount doesn't seem to work as expected? https://paste.ubuntu.com/p/QFGzVgrQtQ/ Jan 20 08:42:12 Nov 20 21:41:55 mido mount: mounting /dev/loop1 on /android-system/vendor failed: No such file or directory Jan 20 09:59:57 Morning! Jan 20 10:00:14 ah, maybe there isn't any "/vendor" directory in the system image... Jan 20 10:01:05 maybe /system/vendor then ? Jan 20 10:05:57 Tofe: In my / I have: lrwxrwxrwx 1 root root 14 Jul 21 1971 vendor -> /system/vendor Jan 20 10:06:19 In my /system I have: lrw-r--r-- 1 root root 7 Jan 19 2020 vendor -> /vendor Jan 20 10:06:28 So they seem to loop? Jan 20 10:15:01 yes, that's not very useful :) Jan 20 10:15:44 that's on device? Jan 20 10:16:03 or in system.img ? Jan 20 10:17:14 the /system/vendor -> /vendor symlink seems to have been created later, maybe we can change it to a directory Jan 20 10:17:31 On device Jan 20 10:18:39 Though in the system.img file I see a symlink as well Jan 20 10:18:53 So it seems to be created in there already Jan 20 10:19:32 we can probably change /vendor to a directory, https://github.com/Tofee/initramfs-tools-halium/blob/tofe/ab-scheme/scripts/halium#L174 Jan 20 10:20:13 Yes probably Jan 20 10:21:51 ideally, /system/vendor wouldn't be a symlink Jan 20 10:22:26 We'll have to choose :) Jan 20 10:24:58 Tofe: I'm not sure how to sort all this ;) So really need some help here ;) Jan 20 10:25:32 There are different places involved here: 1) for the android-flavored drivers, running in main userspace through libhybris, /vendor and /system/vendor may be used. 2) for LXC, android will only use what's in rootfs, and they are bind mounts of what's in /android. 3) the things in /android are setup by halium at boot time Jan 20 10:26:23 Finally, /system and /vendor should link to a place in /android, normally. Jan 20 10:27:37 in that regard, /vendor pointing to /system/vendor is fine. But /system should point to /android/system iirc, and /android/system/vendor should point to ../vendor or should directly contains vendor.img Jan 20 10:30:53 all this can probably be fixed from the halium script, with a closer look Jan 20 10:31:45 In / I have system symlinking to /android/system and vendor to /system/vendor Jan 20 10:31:51 I'd propose: 1) /vendor points to /android/vendor, a bit like /system 2) we try to mount vendor.img on /android/vendor Jan 20 10:32:44 that way, all the other pointing are correct, I think Jan 20 10:33:11 (if /android/vendor gets automatically mounted in lxc's rootfs/vendor) Jan 20 10:34:51 now I'll need a bit of time to propose you a patch, all these mountpoints are confusing :p Jan 20 11:21:05 Tofe: No worry, happy for all the help :) Jan 20 12:49:33 Tofe: Hmmz this is interesting too: https://github.com/Herrie82/android_build/blob/halium-8.1/core/envsetup.mk#L188 Jan 20 12:50:08 If I understood correctly. If I remove TARGET_COPY_OUT_VENDOR from the boardconfig.mk in my device tree it might include the vendor bits in system.img instead, let me try that Jan 20 13:05:17 It seems so, yes Jan 20 13:05:42 but we probably should get the vendor mounting right instead, it's the right way for later Jan 20 13:06:25 at least it's how google wants to split things up from now on Jan 20 13:06:28 Yeah anyway it doesn't seem to work Jan 20 13:06:36 ah, ok :^ Jan 20 13:49:35 Herrie: can you try with this version? https://bpaste.net/Y2DA Jan 20 13:52:04 ah, wait, I'll insert more traces Jan 20 13:53:09 https://bpaste.net/4K4A here Jan 20 13:59:21 Tofe: building Jan 20 14:17:52 flashing Jan 20 14:20:09 Tofe: https://paste.ubuntu.com/p/HtcgH75xRC/ Jan 20 14:20:45 Seems to work ? Jan 20 14:23:16 I have data in /android/vendor now Jan 20 14:23:40 Symlink in /vendor -> /system/vendor Jan 20 14:24:08 in /system/vendor to /vendor still it seems Jan 20 14:41:51 Lots of SELinux errors again, will look into those Jan 20 14:50:33 do you have the vendor stuff in /var/lib/lxc/android/rootfs/vendor ? Jan 20 14:51:40 Yup Jan 20 14:51:55 https://paste.ubuntu.com/p/VMVQ2GVr4k/ Jan 20 14:54:10 ok great :) Jan 20 14:54:36 I don't see any loop in the logs Jan 20 14:56:20 It looks good in general Jan 20 14:56:44 Issues seem to be more with init.rc and SELinux that doesn't get LD_PRELOADED correctly in some places probably Jan 20 14:56:56 And some .so files are missing leading to crashes Jan 20 14:59:44 ok Jan 20 15:03:25 LXC log: https://paste.ubuntu.com/p/gfqx3ypyWP/ Jan 20 15:12:23 Ok seems I need to LD_PRELOAD vold Jan 20 15:12:30 At least that's what Mer used to do Jan 20 15:12:38 That can explain all the SELinux errors in log Jan 20 15:15:35 it would be nice to see them gone, yes Jan 20 15:16:30 The ueventd ones would be good to go as well Jan 20 15:16:33 Not sure what causes those Jan 20 15:17:50 I'm not sure either Jan 20 15:31:53 Defconfig seems OK there Jan 20 17:20:18 Tofe: I guess I found the cause of the missing touch binary Jan 20 17:23:13 Seems that it was provided by android_external_busybox in Halium 7.1 (https://github.com/Halium/android_external_busybox/commits/halium-7.1) however Mer guys replaced it by their own busybox single binary it seems https://github.com/mer-hybris/android_external_busybox_prebuilt Jan 20 17:24:28 Seems LOS has 14.1 branch as latest: https://github.com/LineageOS/android_external_busybox Jan 20 17:29:04 Found this one for 8.1 though seems to have some changes for MediaTek: https://github.com/SnowCat6/busybox Jan 20 17:30:58 Ah this one seem newer and quite maintained: https://github.com/meefik/busybox Jan 20 17:32:15 Ah and another one from Omnirom. I guess I'll start with that one and add bshah's patches: https://github.com/omnirom/android_external_busybox/commits/android-8.1 Jan 20 17:47:20 Ok this should be it I guess: https://github.com/Herrie82/android_external_busybox/tree/halium-8.1 Jan 20 20:22:12 I just realized we have both ModemManager and oFono in our images... Jan 20 20:33:53 Tofe: That's not helping ;) Jan 20 20:34:28 ModemManager is quite seducing, but we built everything around ofono, so let's just continue with the latter Jan 20 20:36:31 Well everyone uses ConnMan and oFono it seems? Only Ubuntu has ModemManager? Jan 20 20:36:46 Busybox doesn't want to build, but checking Jan 20 20:36:50 Shouldn't be rocket science Jan 20 20:38:22 I'm not even sure UbuntuTouch uses ModemManager... Jan 20 20:40:56 Seems just some missing includes in BusyBox Jan 20 20:41:05 Seems bshah's 7.1 patches don't apply cleanly on 8.1 Jan 20 20:41:09 So reworking them a bit Jan 20 20:42:14 (https://puri.sm/posts/librem5-progress-report-12/ even librem chose oFono last year... though not by far, and ModemManager implemented what was missing since then) Jan 20 20:47:51 Well but we also have ConnMann and all Qt bits connecting with it Jan 20 20:48:00 So might be quite a project to replace it Jan 20 20:52:50 Tofe: Any clues? https://bpaste.net/JG6A Jan 20 20:53:50 https://github.com/Herrie82/android_external_busybox/blob/halium-8.1/networking/udhcp/files.c#L38 Jan 20 22:12:17 Herrie: a missing #include ? Jan 20 22:12:48 nope, it's there Jan 20 22:18:53 Herrie: just to be sure, also include **** ENDING LOGGING AT Mon Jan 20 23:42:21 2020 **** BEGIN LOGGING AT Mon Jan 20 23:45:13 2020 **** ENDING LOGGING AT Mon Jan 20 23:48:17 2020 **** BEGIN LOGGING AT Mon Jan 20 23:51:11 2020 **** ENDING LOGGING AT Tue Jan 21 02:59:56 2020