| rinigus | @b100dian: morning! Looks like you forgot to make SDK target with sdk-assistant | 05:01 |
|---|---|---|
| rinigus | Looked at config changes and I think we need to adjust dynpart setup. It's currently using sda10 which is tz partition. I guess it's a typo | 05:03 |
| rinigus | Also not sure if makes sense to use before in systemd dmsetup service for mounts. Instead, I would suggest to use after in mounts. | 05:08 |
| rinigus | @b100dian: we also need to add parse-android-dynparts to patterns. I have pushed "requires" to nolvm branch" | 06:36 |
| rinigus | @b100dian: I have also fixed super partition setup in droid-config. checked in fp4 repos - they had super on sda10; we have on sda93 | 07:58 |
| T42 | <b100dian> Oops yup, forgot to change that. | 08:06 |
| *** Ischwitch is now known as Ingvix | 08:26 | |
| T42 | <b100dian> rinigus: I did add Requires to patterns/patterns-sailfish-device-adaptation.inc, it was on a lower line (90-ish) | 08:31 |
| T42 | <b100dian> rinigus: you're right, I had the target with nagara instead of xqcq54 | 08:34 |
| rinigus | @b100dian: sorry, missed your requires. will remove mine | 08:51 |
| rinigus | actually, I would keep the one closer to hybris as it is more related to overall android-sfos glue | 08:52 |
| T42 | <b100dian> Sure, but no rush, we can edit with the next fix anyway | 08:58 |
| rinigus | done already - otherwise it will slip | 09:08 |
| rinigus | obs rebuilds triggered as well | 09:09 |
| rinigus | @b100dian: do we need parse-android-dynparts in hal? or should it be just one more package at OBS? | 09:12 |
| T42 | <b100dian> just the one package at OBS should work, I was thinking to use hal as a checkout-everything-needed don't feel strong about it | 09:13 |
| rinigus | @b100dian: here our styles are different. I would prefer to keep there only those packages that need android bits. rest can be done at obs and we don't have to worry about keeping submodules in sync with updates | 09:16 |
| rinigus | I'll add it to OBS as well. otherwise mic image building will be problematic | 09:16 |
| *** phlixi is now known as Guest9288 | 09:58 | |
| *** phlixi_ is now known as phlixi | 09:58 | |
| *** phlixi is now known as Guest9300 | 12:27 | |
| *** phlixi_ is now known as phlixi | 12:27 | |
| rinigus | @b100dian: fyi mic create fs fails to make zip file if we ask it to pack root into tar.gz . it works for tar.bz2 :) | 14:45 |
| T42 | <b100dian> Ah, that may require a hyvbis-boot change. But maybe bzip2 unpacking works on our devices? | 14:46 |
| T42 | <b100dian> *hybris-boot | 14:47 |
| T42 | <b100dian> This change if bzip2 doesn't work: https://github.com/sailfishos-on-sake/hybris-boot/commit/85b3bdb68ae2029400ffc184c0ad4a4ad3caded3 | 14:48 |
| rinigus | @b100dian: let's keep generating tgz and failing on the last step. maybe we will go back to lvm anyway - so no need to start fixing it :) | 14:54 |
| rinigus | also, your sake "alternative" flashing instructions are handy in this case | 14:54 |
| T42 | <b100dian> Yeah, these days recovery is not what it used to be:D | 14:56 |
| rinigus | @b100dian: turns out I don't know how to get /data mounted in adb. please instruct | 17:52 |
| rinigus | (in recovery) | 17:53 |
| rinigus | reading back the logs, looks like I just need to format userdata (sda94) | 17:57 |
| rinigus | @b100dian: flashed the latest image. it gets stuck on sony logo (no sfos logo appears anymore) and reboots after some time. will see if there are some logs/traces on device | 18:07 |
| rinigus | have init.log with sleep 60 and reboot after that. | 18:08 |
| rinigus | and have no rndis btw. same in that log | 18:09 |
| T42 | <b100dian> Which boit inage are you using? | 18:34 |
| T42 | <b100dian> *boot | 18:34 |
| T42 | <b100dian> Just mkfs.ext4 /dev/block/sda94 | 18:35 |
| mal | why not use fastboot to format userdata? | 18:45 |
| rinigus | @b100dian: hybris-boot. but now which in particular, would have to trace down. just a sec | 18:49 |
| rinigus | mal: as I was not sure :) . will use it next time | 18:49 |
| rinigus | @b100dian: added this flashing procedure as far as I understood into our docs (end of hadk) | 18:50 |
| T42 | <b100dian> Try a fresh build of hybris-recovery | 18:50 |
| rinigus | btw, for me, fastbootd disappears when I flash hybris boot (and hybris dtbo + vendor_boot). to get back to it, have to flash all 3 from aosp. | 18:51 |
| rinigus | is that expected? | 18:51 |
| mal | known issue | 18:52 |
| mal | fastbootd lives split in vendor_boot and boot, hybris-boot.img doesn't support it, yet | 18:52 |
| rinigus | mal: thanks! | 18:54 |
| rinigus | @b100dian: do you tahe your hybris-boot from RPM or from hadk/out/target/product/pdx224/hybris-boot.img | 18:55 |
| T42 | <b100dian> The latter | 18:55 |
| rinigus | @b100dian: hybris-recovery does show up rndis in dmesg. and shows sfos logo after which all goes dark and rndis disappears. will try once more with -recovery after removing init.log (not sure whether its old or new) | 19:45 |
| T42 | <b100dian> Chroot from recovery into sailfishos and mask usb-moded | 19:46 |
| T42 | <b100dian> Or mask it very fast:) | 19:47 |
| rinigus | @b100dian: doesn't look like possible. lots of permission denied in init.log: https://paste.mozilla.org/O4CUzRPp/raw | 19:57 |
| rinigus | that's with hybris-recovery, dtbo, and vendor_boot from sfos hadk | 19:57 |
| rinigus | from out/target/product/pdx223 | 19:57 |
| rinigus | and there is also journal, will try to pull that | 19:58 |
| mal | why is that trying to use android_usb and not configfs? | 20:00 |
| rinigus | no idea. would have to find the corresponding init | 20:03 |
| mal | is seems to find /sys/class/android_usb/android0 folder and then tries to use it | 20:05 |
| mal | some mistake in kernel config changes? | 20:05 |
| T42 | <elros34> change if/else order so script will use configfs | 20:06 |
| rinigus | mal: re config mistakes, don't know. would have to look probably | 20:09 |
| rinigus | @elros34: can try, thanks | 20:09 |
| T42 | <elros34> my wild guess is something initialize usb (but what?) that is why android0 is available but why b100dian didn't have this issue | 20:12 |
| rinigus | @elros34: I hope we have the same builds, but cannot be 100% sure. also I don't know if b100dian flashed vendor_boot and dtbo. also, its all while booting using hybris-recovery. no idea if it has any impact | 20:20 |
| rinigus | while we do have different devices (1iv vs 5iv), platform is the same and it shouldn't matter | 20:20 |
| rinigus | going through the journal to see if there is something obvious | 20:21 |
| T42 | <b100dian> rinigus: I have only flashed boot not anything else, over aosp15 r67. | 20:22 |
| T42 | <b100dian> Do you have the defconfig changes in kernel repo and did you manually build it? Because it picks some prebuilt otherwise | 20:23 |
| rinigus | @b100dian: OK, so flashing is different. re kernel - did build it manually using ./build-kernels-clang.sh | 20:26 |
| rinigus | I wonder if I can check config from hybris-recovery image | 20:26 |
| T42 | <b100dian> Should be possible, some proc/zConfig if I remember correctly | 20:27 |
| T42 | <elros34> sure zcat /proc/config.gz | 20:27 |
| rinigus | yes, but that's while booted. probably can extract that from image as well | 20:29 |
| rinigus | one of the first things in journal: dmsetup.service: Main process exited, code=exited, status=1/FAILURE | 20:30 |
| rinigus | so, dynamic partitions are bust | 20:30 |
| rinigus | not configured during boot | 20:30 |
| T42 | <b100dian> Hmm.. Ok, so you are using a build I have not yet tried | 20:31 |
| T42 | <b100dian> But still you did have *some* telnet access? | 20:31 |
| rinigus | re telnet: not really, didn't had a chance to configure it on pc side. and if you look into init.log above then its not very promising either | 20:34 |
| rinigus | re dyn partitions: they somehow are still on as I have "Found device /dev/mapper/dynpart-odm_a." and later its mounted | 20:34 |
| T42 | <b100dian> Ok, so that is log from android recovery. Ok, then Elros suggestion of switching the if/else could do it. I have a static IP on my laptops usb interface since.. 2021 probably, first port:) | 20:35 |
| T42 | <elros34> wait do you flash regular hybris-recovery.img? Why not hybris-boot.img? | 20:40 |
| rinigus | towards the end of a journal I do get lots of "Dependency failed for Droid mount for /vendor" due to failed dmsetup I think. | 20:40 |
| rinigus | re hybris-boot: can try that again as well. that ended up just rebooting. | 20:41 |
| rinigus | will test again | 20:41 |
| T42 | <elros34> create /init_enter_debug2 | 20:41 |
| T42 | <elros34> it's all little strange, you shouldn't be able to boot to real rootfs when using hybris-recovery | 20:43 |
| rinigus | it still gets stuck on sony and reboots on 60s | 20:47 |
| rinigus | log coming up | 20:47 |
| T42 | <elros34> maybe I am wrong and if recovery init usb and doesn't findy any issue it just continue booting to second stage and then it fails with usb initialization | 20:49 |
| rinigus | https://paste.mozilla.org/x3b0r2Ah/raw | 20:50 |
| rinigus | not sure it matters, but there is no usleep (usleep: not found) in these init.logs | 20:51 |
| T42 | <b100dian> usleep: not found | 20:51 |
| T42 | <b100dian> Wow | 20:51 |
| T42 | <b100dian> Change it to sleep 1, but I am really curious how it worked for me now | 20:52 |
| rinigus | I have no idea. rather strange | 20:54 |
| T42 | <b100dian> I mean, I didnt have telnet without one pause there but I'm confuses how did I have one with usleep if it is not defined | 20:54 |
| T42 | <b100dian> *confused | 20:55 |
| T42 | <b100dian> Still away from my keyboard, maybe in 10 mins I'll check.. | 20:55 |
| rinigus | that helped, now shows sfos logo that blanks out later. so, similar to recovery | 20:59 |
| rinigus | and rndis disappears | 21:00 |
| T42 | <elros34> if you would have init_enter_debug2 then you shouldn't see sfos log and have persistent telnet | 21:01 |
| rinigus | @elros34: I see. that would have to wait for tomorrow - rather late over here. sorry | 21:01 |
| T42 | <elros34> sure thing, of course you can create it in recovery | 21:02 |
| rinigus | yes, that's a plan. unless @b100dian will fix everything meanwhile :) | 21:05 |
| T42 | <b100dian> well, I also have the usleep: not found, so I was just lucky in a race with a kernel. gonna set back to 1s, and yes, I'm going to continue creating mic again myself, hopefully with more brainpower left than yesterday | 21:06 |
| T42 | <b100dian> There's another usleep 500000 usb_setup_android_usb and I haven't introduced that one | 21:08 |
| T42 | <b100dian> I _think_ that this init script is ran twice, once in ramfs and once in real rootfs, and in ramfs there's usleep | 21:09 |
| mal | yes, both are run | 21:10 |
| T42 | <elros34> it is but usb init part is not | 21:10 |
| rinigus | OK, that's good. some progress is there even without telnet for me - can pull journal using adb from recovery. better than empty boots | 21:10 |
| T42 | <elros34> I wonder if this strange android_usb behavior is because of using hybris-recovery. I saw something similar in past and now I think it was because porters tried recovery (usually nobody uses it) | 21:13 |
| T42 | <b100dian> I did manage to get telnet on hybris-boot with empty ext4 /data partition, but not with unpacked rootfs | 21:14 |
| mal | why are you using hybris-recovery? | 21:14 |
| T42 | <b100dian> So I assumed you were correct in saying that hybris-boot does not have kernel access on aosp, but really didn't check | 21:14 |
| T42 | <b100dian> And that's why I switched to hybris-recovery | 21:14 |
| T42 | <elros34> but init_enter_debug2 should work in your case | 21:15 |
| T42 | <b100dian> It may just be that I added that usleep instead of sleep, it didn't work, and I lost the kernel race one time randomly and won it another time, still randomly | 21:15 |
| T42 | <b100dian> Yeah, gonna try again now, first without init_enter_debug2 | 21:15 |
| T42 | <b100dian> then with | 21:16 |
| T42 | <b100dian> *kernel access = telnet access | 21:16 |
| T42 | <elros34> people seems to try to avoid it like it would break something but it only save time, you can easily continue booting with it | 21:16 |
| mal | yeah, I add init_enter_debug2 when do testing on new device | 21:18 |
| mal | and also always at first disable droid-hal-init and usb-moded | 21:18 |
| mal | to prevent losing telnet and possible boot loops | 21:18 |
| T42 | <b100dian> so with hybris boot no init.log and no telnet, reboot in 60 seconds, with init_enter_debug2 in root nothing either. Looks like /data is not mounted. Aaand of course | 21:28 |
| T42 | <b100dian> /boot appears to live on | 21:28 |
| T42 | <b100dian> sounds like fixup-mounpoints again, where we have boot_a :D | 21:28 |
| T42 | <elros34> most likely you don't have /boot defined in fstab | 21:29 |
| mal | yeah, many new devices don't have it in fstab | 21:30 |
| mal | but build should complain about it | 21:30 |
| T42 | <b100dian> but that doesn't explain why recovery has telnet and boot doesnt , right? | 21:33 |
| mal | assuming those are otherwise the same | 21:56 |
| T42 | <b100dian> cp: cannot stat './out/target/product/pdx224/hybris-updater-script' | 23:38 |
| T42 | <b100dian> I see that this needs the /boot entry in fstab, but what to rebuild? hybris-boot would be my first guess, but doesn't make it | 23:39 |
| T42 | <b100dian> the error is on `rpm/dhd/helpers/build_packages.sh -d` | 23:39 |
| T42 | <elros34> build hybris-hal, it t should build everything needed | 23:43 |
| T42 | <b100dian> hmm I did | 23:44 |
| T42 | <elros34> so you still have errors about /boot? | 23:45 |
| T42 | <b100dian> not sure, just found this missing hybris-updater-script and searched the logs here that it may be related to boot | 23:52 |
| T42 | <b100dian> to be fair, I have not built this nagara repos locally, rinigus did all that on OBS :) so I'm doing this now | 23:53 |
| T42 | <elros34> see Android.mk from hybris-boot. If you don't have b | 23:53 |
| T42 | <elros34> boot entry in fstab then there will be missing hybris-updater-script | 23:54 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!