| T42 | <b100dian> ok, I think I needed to make it real, not just boot /boot, but /dev/block/bootdevice/by-name/boot /boot. Thanks! | 00:04 |
|---|---|---|
| T42 | <b100dian> maybe hybris-boot.img works too now :P | 00:04 |
| T42 | <b100dian> before, I got "Skipping build of hybris-updater-script since HYBRIS_BOOT_PART is not specified" | 00:05 |
| T42 | <b100dian> but now I don't get that and still no output, let me debug it further | 00:05 |
| T42 | <b100dian> hmm make hybris-common ? is that expected to be needed? that generates the file | 00:07 |
| ecrn | Program terminated with signal SIGSEGV, Segmentation fault. | 02:39 |
| ecrn | #0 0x0000007f704a1eb0 in import_exynos_ion_handles(private_handle_t*) () from /android/vendor/lib64/hw/android.hardware.graphics.mapper@4.0-impl.so | 02:39 |
| ecrn | https://pastebin.k4be.pl/view/raw/d60b00fb any ideas? | 04:20 |
| ecrn | the ubuntu touch port I consider a warmup before trying to port SFOS on the device | 04:20 |
| ecrn | because there are no android/lineage sources yet | 04:21 |
| ecrn | but the problem above probably is generic | 04:22 |
| ecrn | https://pastebin.k4be.pl/view/raw/8154a65a | 04:23 |
| ecrn | faulting instruction is => 0x0000007f70121eb0 <+120>: ldr w0, [x20] | 04:24 |
| ecrn | and x20 is 0 | 04:24 |
| ecrn | and well, probably can't be anything else since there is no explicit write to the register between 0x0000007f70121e58 <+32>: mov x20, x0 and the faulting instruction | 04:25 |
| ecrn | oh wait, it is x0 not 0, so it is the function argument that is passed as NULL | 04:26 |
| ecrn | but now it doesnt make any sense | 05:54 |
| ecrn | both x0 and x20 on the faulting instruction are 0/NULL, and the function that calls the import_exynos_ion_handles is almost exactly like https://android.googlesource.com/device/amlogic/yukawa/+/f26afff/gralloc/mali_gralloc_reference.cpp#37 | 05:56 |
| ecrn | with the same validation | 05:57 |
| ecrn | https://pastebin.k4be.pl/view/raw/2f8f7da7 | 06:05 |
| ecrn | i can't see how that would end up passing NULL to import_exynos_ion_handles | 06:11 |
| ecrn | even if called with invalid arguments | 06:12 |
| T42 | <elros34> @b100dian just make hybris-updater-script but that is not a problem.Go bac to basics ff you still dont have init.log and 60s reboot then clearly you are in initramfs with failing /data mount or something missing in defconfig and without corrrect usb initialization. | 06:56 |
| T42 | <elros34> Mind that hybris-recovery is same script, just one variable is different. | 06:58 |
| ecrn | where does android_log_printf output go on libhybris-loaded libraries? | 07:06 |
| mal | logcat | 07:08 |
| mal | I think | 07:08 |
| rinigus | @elros34: something is off and, indeed, we have to go back to basics. although, after adding sleep 1 (instead of usleep), it does boot to sfos in terms of mounting /data and so on. but making telnet and init with debug2 as outlined last night would help. | 07:48 |
| T42 | <b100dian> Yes, it is weird that hybris-boot doesnt have telnet and hybris-recovery does, abd boots. One thing I hoped would differ was the overhaul of mountpoints, which I didnt have locally for some days, I only had boot and userdata when I thought I made this work | 08:29 |
| T42 | <b100dian> Theoretically I can continue edit and boot the build I had from OBS but I thought to also build locally and I am failibg at that:D So - I can boot, with rinigus's build, but cannot yet build. And rinigus builds but cannot boot:) Until yesterday IIUC | 08:31 |
| rinigus | @b100dian: yes, that's correct. now I can boot after switching to ext4 as userdata. but no telnet yet and that would require some troubleshooting tonight | 08:35 |
| rinigus | as for local builds, its mainly to build hal - rest we can push to OBS and be with it doing a job for us | 08:35 |
| ecrn | https://pastebin.k4be.pl/view/raw/0afc2159 | 09:07 |
| ecrn | if 3 is error code then it is ESRCH (No such process) | 09:10 |
| ecrn | is there some service for the "system environment" or what they could mean there? | 09:11 |
| ecrn | apparently that should not be the issue | 10:37 |
| ecrn | funny thing, the host os booted normally and the lxc container booted in charger mode | 12:37 |
| ecrn | displays charging animation | 12:38 |
| *** amccarthy is now known as Guest9415 | 17:16 | |
| *** amccarthy_ is now known as amccarthy | 17:16 | |
| ecrn | I have a feeling that something is broken with the libraries or linking | 19:07 |
| rinigus | @b100dian: I have telnet as well. that's with sleep 1 | 19:10 |
| T42 | <b100dian> In hyris-recovery or boot? | 19:10 |
| rinigus | also added init_enter_debug2 and can access it better | 19:10 |
| rinigus | with hybris-boot | 19:10 |
| rinigus | although, it sometimes disconnects and I have to restore connection on pc side (ip addr add...) | 19:11 |
| rinigus | few observations from init: rndis_bam.rndis: No such file or directory | 19:12 |
| T42 | <b100dian> great - then we just have to fix the OBS broken vendor_boot error and get to GUI 😄 | 19:13 |
| rinigus | also, as /etc/udhcpd.conf is symlink to /run, all attempts to write to it lead to nonexistent directory | 19:13 |
| T42 | <b100dian> Yeah, also noticed the bam error but preferred minimal changes to hybris-boot errors so I left it there. I don't know about udhcpd.conf errors' source and I don't use dhcp for some time.. | 19:14 |
| rinigus | well, works as it is for now. | 19:15 |
| T42 | <b100dian> ah, maybe some /run is present in rootfs though | 19:15 |
| rinigus | its link to /run/usb-moded/udhcpd.conf | 19:15 |
| rinigus | so, to make it work, /run/usb-moded has to be created. but let's keep it as it is for now | 19:16 |
| rinigus | what was vendor_boot issue? | 19:16 |
| rinigus | actually, I should check whether hybris-boot via rpm also works. not from out/... | 19:16 |
| rinigus | in addition, found how to get kernel config from boot image. wrote it down at https://github.com/sailfishos-sony-nagara/main/wiki/Extracting-data-from-stock-ROM#extracting-kernel-config | 19:17 |
| T42 | <b100dian> oh, nice | 19:18 |
| T42 | <b100dian> rinigus: nothing provides droid-hal-pdx223-kernel-modules-vendor_boot is one example error in OBS | 19:21 |
| rinigus | @b100dian: forgot to tell you, sorry! those are not needed - those packages are required for lvm root case. we right now have all images in droid-hal | 19:22 |
| rinigus | I have kept those just so it is easy for us to go back to lvm if we want | 19:23 |
| rinigus | so, mic should already work from OBS | 19:23 |
| rinigus | @b100dian: would be nice to replace get_ks with your xmllint approach - I think its better and probably works well from platform sdk too | 19:24 |
| T42 | <b100dian> oh, good think I asked | 19:24 |
| rinigus | :) | 19:25 |
| T42 | <elros34> rinigus rndis_bam is not really an error, it's there for compability with devices which uses different driver | 19:25 |
| T42 | <b100dian> rinigus: xmllint has a small problem - too much standard compliance, meaning that if you have namespaces in your xml some queries fail, that's why there's a sed beforet hat step | 19:26 |
| T42 | <b100dian> Elros: that's what I assumed too | 19:26 |
| rinigus | OK, I will test now with boot/dtbo/vendor_boot images we get from RPMs. hopefully this works too | 19:26 |
| T42 | <b100dian> are images built on OBS? and are they from our hybris-boot? or you've built them locally and updated them with drioid-hal rpms? | 19:28 |
| rinigus | @b100dian: current images are built locally and I push them to obs | 19:30 |
| rinigus | https://build.sailfishos.org/package/show/nemo:devel:hw:sony:nagara/droid-hal-nagara | 19:31 |
| rinigus | and they boot the same way as the ones from out/... | 19:31 |
| rinigus | just their md5sum was different from out/... | 19:31 |
| rinigus | although, I may have regenerated images later with the same kernel/config as a part of make hybris hal | 19:32 |
| rinigus | so, telnet works with them too as long as you have init_enter_debug2 and sleep 1 in init script | 19:33 |
| rinigus | I suspect that these "Skipping ..." lines are a problem: https://paste.mozilla.org/wixNndGa/raw | 19:38 |
| rinigus | probably need to adjust /usr/bin/droid/dmsetup.sh script | 19:38 |
| rinigus | although, it does go to stderr. so, maybe not | 19:40 |
| T42 | <b100dian> not sure if those are a problem | 19:41 |
| T42 | <Mister_Magister> @b100dian zenfone 8 port dead? | 19:43 |
| T42 | <Mister_Magister> killed for good? | 19:43 |
| T42 | <b100dian> @Mister_Magister it has one user 😉 | 19:46 |
| T42 | <b100dian> half from the old days | 19:46 |
| T42 | <Mister_Magister> good, f asoos | 19:46 |
| T42 | <Mister_Magister> i flashed stock onto it and flashed sfos onto fxtec prototype | 19:46 |
| T42 | <Mister_Magister> now its degraded to being dedicated for android crap | 19:46 |
| T42 | <b100dian> you know you have a bootloader unlocked zenfone 8, maybe somebody else wants a small device with sfos | 19:47 |
| T42 | <b100dian> Although me fixing the wifi bug is probably 0 chance | 19:47 |
| T42 | <Mister_Magister> somebody else would have to buy device thats unlockable | 19:47 |
| T42 | <Mister_Magister> and not absolute garbage | 19:47 |
| T42 | <Mister_Magister> that dumps ram on production | 19:48 |
| rinigus | @b100dian: I am pretty sure that dmsetup fails. but not because of Skipped lines. right now, after disabling usb-moded I could check available devices | 19:58 |
| rinigus | and we have only dynpart-odm_a and dynpart-product_a | 19:59 |
| rinigus | if I manually add text to dmsetup for system_a (see my paste) which consists of multiple parts - it fails | 20:00 |
| rinigus | for dynpart-system_ext_a works | 20:01 |
| T42 | <b100dian> Hmm.. I've only analyzed super img from unsin/stock | 20:01 |
| T42 | <b100dian> but I thought the structure and partition names are kept when you flash aosp in fastbootd mode | 20:02 |
| rinigus | @b100dian: they probably are. note that system_a is the only one with multiple linear components | 20:05 |
| rinigus | that's easier to see if I split the string by ; as here: https://paste.mozilla.org/OpaY9ubx/raw | 20:05 |
| T42 | <b100dian> actually I now see that you used sda93 so that is the same structure apparently | 20:05 |
| T42 | <b100dian> god | 20:06 |
| T42 | <b100dian> I see what you mean | 20:07 |
| ecrn | is 93 partitions crazy yet? | 20:07 |
| rinigus | 93 is not the last one :) | 20:08 |
| T42 | <b100dian> it's from '93 | 20:08 |
| T42 | <K31j0> Couldn't we just format something to ext4 and call it / | 20:10 |
| T42 | <Mister_Magister> noo u dumb? | 20:10 |
| T42 | <Mister_Magister> how would that function | 20:10 |
| T42 | <Mister_Magister> you need separate partition for every directory no matter that btrfs subvolumes would achieve the same thing | 20:10 |
| T42 | <Mister_Magister> GPT can have up to 128 partitions so i'm going to use all of them goddamnit! | 20:11 |
| rinigus | re formatting: not our call | 20:13 |
| rinigus | where are the sources of device-mapper spec? | 20:13 |
| T42 | <elros34> soon they will invent something to save space like dynamic partitions v2 -> directories | 20:13 |
| rinigus | https://github.com/sailfishos/device-mapper is archived | 20:14 |
| rinigus | found it - lvm2 | 20:14 |
| T42 | <Mister_Magister> they gonna put everything on one partition and partition it in lvm so that they can have even more partitions (re @elros34: soon they will inven...) | 20:14 |
| T42 | <Mister_Magister> separate lvm disk for each slot so 128 partitions per slot! | 20:15 |
| T42 | <elros34> so you wan to put partition in partition to have more partitions?:) | 20:15 |
| rinigus | lovely. lvm2 is from 6 years ago | 20:16 |
| T42 | <Mister_Magister> there's nothing up-to-date in sfos :) and there wasn't for years | 20:16 |
| T42 | <Mister_Magister> exactly @elros34 | 20:16 |
| T42 | <b100dian> rinigus: but system.. we can package that back into rootfs, that would be compatible even with lineageos build | 20:17 |
| T42 | <b100dian> only system is at problem? | 20:18 |
| T42 | <Mister_Magister> you can any time you want, jolla is doing that | 20:18 |
| rinigus | @b100dian: in principle, yes. so far only system | 20:18 |
| T42 | <Mister_Magister> we did that on fxtec | 20:18 |
| T42 | <K31j0> I will (re @Mister_Magister: noo u dumb?) | 20:18 |
| T42 | <Mister_Magister> tbh if you figured out which partitions does xbl use you could probably remove some | 20:18 |
| T42 | <K31j0> Time to enable zfs | 20:18 |
| T42 | <Mister_Magister> and once in sfos you can just mount images and android would be non the wiser | 20:19 |
| T42 | <K31j0> raid0 deez | 20:19 |
| rinigus | I'll check the sources and see if there was a fix for it | 20:19 |
| T42 | <b100dian> there must be a way to un-super-ize the device, but that would put a high barrier on installing | 20:19 |
| T42 | <Mister_Magister> naah if you made nice installation script there should be no issue | 20:20 |
| T42 | <Mister_Magister> i remember resizing partitions on my old htc explorer | 20:20 |
| T42 | <Mister_Magister> just remove partition by partition until it no longer boots fastboot | 20:21 |
| T42 | <b100dian> yeah, but there's always that guy that pastes from word^H^H and everything blows up | 20:21 |
| T42 | <Mister_Magister> ah classic, i was helping guy once who couldn't copy paste | 20:22 |
| rinigus | @b100dian: as we probably will want to base the port on stock after this aosp exercise . so, would be good to fix it on dmsetup / parse-android-dynparts level | 20:22 |
| T42 | <b100dian> what's team Verdandi working on now? OP6/T? | 20:22 |
| T42 | <Mister_Magister> op6(t) is semi-done | 20:22 |
| T42 | <b100dian> yes, but that system also has sources | 20:22 |
| T42 | <Mister_Magister> i'm planning jolla tablet v2 | 20:22 |
| T42 | <Mister_Magister> nobody ported tablet in a while | 20:22 |
| T42 | <Mister_Magister> op6 needs rebasing to android 11 but first apps need work | 20:23 |
| T42 | <Mister_Magister> i had guy complain that he can't open 50 apps because oom kicks in but tbh he also complained he has to follow instructions to get it to work | 20:25 |
| T42 | <Mister_Magister> but other than that more and more people are buying it and i hear little complaints | 20:25 |
| rinigus | @b100dian: looking at dmsetup source it seems all to be OK. should support multiple elements as well | 20:26 |
| rinigus | the error message is actually "device-mapper: reload ioctl on dynpart-system_a failed: Invalid argument" | 20:29 |
| rinigus | also, we have wait_for_keymaster[1630]: [gbinder] ERROR: Can't open /dev/hwbinder: No such file or directory | 20:30 |
| rinigus | but that /dev/hwbinder is listed in ls -l | 20:31 |
| T42 | <b100dian> I just booted with fresh mic and old hybris-recovery. I have /dev/binderfs/hwbinder though, not /dev/ one | 20:36 |
| T42 | <b100dian> I see that only two partitions are mounted on dmsetup as you mentioned | 20:36 |
| rinigus | @b100dian: my mistake, /dev/hwbinder is missing indeed | 20:38 |
| T42 | <b100dian> Maybe we can link it from droid hal pre init. But lets focus on dyn parts for now | 20:39 |
| rinigus | agreed | 20:42 |
| T42 | <b100dian> dmsetup sh in dmesg | 20:43 |
| T42 | <b100dian> device-mapper: table: 253:2: linear: Invalid argument count | 20:43 |
| T42 | <b100dian> device-mapper: ioctl: error adding target to table | 20:43 |
| rinigus | can't find these messages in lvm2 sources. in kernel? | 20:50 |
| rinigus | looks to be from kernel/drivers/md/dm-linear.c | 20:53 |
| T42 | <b100dian> yes, they're in kernel/drivers/md/dm-ioctl.c 2192 and 1396 | 20:53 |
| T42 | <b100dian> and dm-linear.c yes | 20:53 |
| T42 | <b100dian> hopefully there's some config that would unlock the support | 20:55 |
| rinigus | Let's hope so. Would have to read the sources and maybe see how android is mounting it. | 21:02 |
| T42 | <b100dian> looks as if parse-android-dynparts adapts android C code | 21:04 |
| rinigus | @b100dian: I'll continue tomorrow, let's see what you will find | 21:11 |
| T42 | <b100dian> ringus: still an updated dmsetup / device-mapper package sounds like a fun project. | 21:17 |
| T42 | <b100dian> https://gitlab.com/lvmteam/lvm2/-/blob/main/WHATS_NEW_DM didn't find anything that jumps out | 21:17 |
| ecrn | is android libraries abi stable within given android version? | 21:30 |
| ecrn | because if not, that could explain the sigsegv I'm getting | 21:37 |
| ecrn | something is overwriting class members | 21:37 |
| ecrn | in the counstructor, mind you | 21:38 |
| T42 | <b100dian> ecrn: you mean like an inline initializer that steps on virtual methods? or, what type of members, data? | 21:57 |
| T42 | <b100dian> mal: do we have aio / as in https://gitlab.com/lvmteam/lvm2/-/blob/main/lib/device/bcache.c?blame=1#L29 ? | 22:12 |
| T42 | <b100dian> at least my host dmsetup works with the wonky super.img https://paste.mozilla.org/RafuTi9i/raw | 22:15 |
| ecrn | data members | 22:19 |
| ecrn | the method tables possibly too, but unproven | 22:20 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!