Saturday, 2018-12-01

mortalglitchjust got back into my htc 10 via telnet after a reflash, still have no display. Checked the init log and the last line is ####### Beginning inject loop00:21
malmortalglitch: do you connect to telnet port 2323?00:22
mortalglitchi tried 2323 earlier and got connection refused00:23
mortalglitchI tried the "echo "continue" >/init-ctl/stdin" and it seemed to trigger a reboot00:25
malcheck the init log to see if the mounting what successful00:28
malor if there are some messages about failing kernel config checks, better just pastebin the whole init.log00:29
malthose are the two most likely causes of boot stopping at telnet 2300:29
malmortalglitch: are you sure /dev/mmcblk0p63 is userdata partition?00:35
malwhat device is that?00:36
mortalglitchHTC 1000:36
mortalglitchdouble checking mountpoints and making sure I didn't fat finger something00:37
malmortalglitch: init.log has "mounting /dev/mmcblk0p63 on /data failed: Invalid argument" which to me look like it could be the wrong partition00:37
mortalglitchgood call ty for taking a look. Will go back over the mounting and see what I can find. also I saw the "failed to boot init in real rootfs"00:38
malmortalglitch: yes the failure to boot message is because it cannot find the rootfs because mounting failed00:39
mortalglitchthe mmcblk0p63 is correct for the name of the block but the script looks like it's pointing at /dev/mmcblk0p63 when it should be /dev/block/mmcblk0p63, did I somehow trigger this through the fixup-mountpoints file?00:57
malno, the kernel init doesn't the /dev/block yet01:00
malmortalglitch: are you sure the userdata is not encrypted or something?01:00
mortalglitchI had the device encrypted then wiped thing completely prior to starting this process. I'll see if I can find out if it still is somehow01:02
malmortalglitch: another possiblity, check what filesystem userdata has01:04
malsome android devices have used something else than ext401:04
mortalglitchusing mount | grep "^/dev" - through adb everything there is reporting ext4 but block p63 isn't listed01:11
malso what is p63 then01:14
mortalglitchhaven't been able to figure out how to get it to list it's type01:17
malboot to android and check01:18
malor in recovery01:18
mortalglitchtrying most commands I can find and none seem to show the type of the block.01:34
T4<elros34> blkid, mount non of them? try mounting particular partition manually it should print error01:42
mortalglitchended up dumping twrp recovery log and it shows mcblk0p63 as ext401:50
malare you able to mount it manually in telnet 2301:52
mortalglitchthe recovery log is saying that it's encrypted with a general default password if i'm reading this properly01:54
malso reformat the partition without encryption01:54
*** ChanServ sets mode: +v T402:33
mortalglitchseems to finally be able to mount ty all! now I can telnet into 2323 trying to find the hang here02:42
mortalglitch /sbin/preinit is trigger Trying to run as user instance, but $XDG_RUNTIME_DIR not set, any clue on this one? I thought it was specific to selinux but I thought I disabled all that04:26
UmeaboyHi! Is it to much to ask to get a new build for jfltexx here: ?04:32
UmeaboyIt's quite old. I'm thinking of purchasing the S4 again as I really like it.04:32
*** ChanServ sets mode: +v T408:26
T4<adampigg> @eugenio_g7 i gave up and did the battery disconnect trick and it worked!09:57
piggzmal: one thing to note when reviewing the Qtmm camerabin changes is that i tried to make it so that there was no change in bahavour if the properties dont exist10:12
eugeniopiggz, glad to read that the pad is booting again :)12:38
eugeniopiggz, the pieces required to get UI working via mesa are now in my experimental repository:
piggzeugenio: fab, i'll try them soon ... off out this afternoon13:18
piggzwhat about other port status?13:19
piggzBT, WIFI, sensors?13:19
eugenioWiFi works, provided /system is mounted13:20
eugenioBT and sensors don't13:20
eugenioaudio doesn't work too13:20
eugenioWiFi advertises a random MAC address at every boot though13:23
eugenioanyway, to get UI working these steps should be enough:13:23
eugenio1) remove libhybris13:24
piggzcool ... im keen on BT so will work on that first13:24
eugenio2) install the following packages from my repository: "mesa-i915 mesa-i915-dri-i915-driver mesa-i915-libEGL mesa-i915-libGLESv2 mesa-i915-libgbm mesa-i915-libglapi mesa-i915-libwayland-egl"13:24
eugenio3) upgrade Qt with the recompiled version also available in the repository13:24
eugenio4) change the lipstick env variables as follows:
eugenio5) chmod -R 777 /dev, otherwise UI doesn't start (haven't actually see why, will take a look)13:24
eugeniopiggz, great13:24
piggzeugenio: r0kk3rz i think posted a neat udev change that makes the touchscreen device predictable13:28
piggzeugenio: also, did you have any issues mounting android partitions? i havnt looked into yet, but im sure /system wasnt mounted for me before I lost access to the tab13:31
maleugenio: could you grab output of evdev_trace -i (from mce-tools package) to see if the sensors are available as event devices13:37
r0kk3rznot predictable, but it seems like theres some auto sensing in the evdevtouch plugin or something13:37
piggzmal: my system.mount has the correct paths ... and when I mount manually it works fine...13:58
malpiggz: could you show the mount file14:00
piggz but when mounted with systemd, it times out waiting for14:00
piggzDec 01 13:53:17 Sailfish systemd[1]: dev-mmcblk0p11.device: Job dev-mmcblk0p11.device/start failed with result 'timeout'.14:00
malpiggz: increasing timeout doesn't help?14:01
piggzmal: the device exists, and i can mount it14:02
piggzi dont know where that depend job is from14:02
piggzeven me manually mounting now, it doesnt14:03
piggzusing systemctl start system.mount14:03
piggzbut using mount works fine14:03
piggzno, increasing timeout didnt help14:07
piggzhave a think, im not in a rush out now for the afternnon, i'll check telegram for suggestions :D14:07
piggzmal: a google sussgested config_fhandle, but i have that enabled14:14
piggzmal: another suggestion is systemd waiting on udev to show the device exists...14:17
piggzudevadm info --export-db14:17
piggzbut this doesnt print anything14:17
piggzso maybe udev isnt working14:22
piggzDec 01 14:15:45 Sailfish systemd-udevd[2436]: unknown key 'DEVLINKS' in /lib/udev/rules.d/999-android-system.rules:26814:23
piggzDec 01 14:15:45 Sailfish systemd-udevd[2436]: invalid rule '/lib/udev/rules.d/999-android-system.rules:268'14:23
malpiggz: which rule is that line?14:27
piggzSUBSYSTEM=="block", DEVLINKS=="* platform/pci0000:00/80860F14:00/by-name/android_persistent *|platform/pci0000:00/80860F14:00/by-name/android_persistent *|* platform/pci0000:00/80860F14:00/by-name/android_persistent|platform/pci0000:00/80860F14:00/by-name/android_persistent", MODE="0660", GROUP="system", OWNER="system"14:27
*** ChanServ sets mode: +v T414:41
*** nic is now known as Guest7447314:57
eugeniopiggz, yup /system doesn't mount for me automatically but that is a moot point as I plan to get rid of it eventually :)14:58
eugeniomal, this is the output of evdev_trace:
eugenioI don't think they're exposed as event devices :/ but that might be some firmware issue15:03
mortalglitch /sbin/preinit is triggering "Trying to run as user instance, but $XDG_RUNTIME_DIR not set", any clue on this one? I thought it was specific to selinux but I thought I disabled all that15:10
mortalglitchgetting that through telnet 2323 trying to force it to finish initializing15:11
malmortalglitch: get output of dmesg and journalctl -a, also systemctl15:16
maleugenio: yep, no sensors in there15:17
eugeniocherry trail uses this thing:
T4<adampigg> We should be able to get sensors via hybris tho? But can we mix parts of hybris and native mesa?15:58
eugenioI guess as long as we don't replace libegl/libgles/etc with the ones provided by libhybris it should work, but I think that they should be exposed as event devices16:12
eugenioand there is something else wrong16:13
eugeniothe android init mixins load the modules manually (in fact, in /lib/modules there are a bunch of kernel modules for the various sensors)16:15
eugenioafter loading the hid-heci-ish.ko module I get this stack trace:
eugenioplus there seems that firmware loading is broken somehow16:18
eugenios/there seems/it seems/16:18
eugeniothe kernel cmdline contains firmware_class.path=/system/etc/firmware but I've copied firmwares there so that they are available even though /system isn't mounted16:19
eugeniothe only device that correctly picks up the firmware is the wifi adapter, but I think that it doesn't search for it (the firmware is specified manually via the CONFIG_BCMDHD_FW_PATH kernel option)16:21
maleugenio: so it uses iio sensors?16:22
maleugenio: sensorfwd has support for some iio sensors16:23
eugeniogreat, so we could avoid hybris entirely16:24
maleugenio: I'll have a look how those should work16:25
maleugenio: pastebin output of ls -l /dev/16:26
eugeniogreat, thanks!16:27
malwondering what sensors the device actually has?16:27
maleugenio: can you have a look at /sys/class and see if there is anything that looks like sensors or iio16:28
eugenionothing, but that could be because the kernel modules don't load properly (
eugeniolooking at /lib/modules I do have a bunch of sensor modules indeed:
malcan you find the firmware somewhere, maybe there is some symlink missing and it cannot find the firmware files?16:32
T4<adampigg> So much fun, kinda wish i wasnt out today!!16:34
eugeniohm, no symlinks:
eugeniodfw_sst.bin and fw_sst_22a8.bin should be related to the audio card16:35
eugeniobut the kernel doesn't manage to load them either16:35
eugenio@adampigg: I won't steal all the fun, don't worry :)16:36
malthere can be other places for firmware files like /firmware or something16:36
maleugenio: add symlink /etc/firmware pointing to /system/etc/firmware16:37
eugenioyup done that, also /lib/firmware16:37
eugeniohm there is /system/vendor/firmware though16:38
malwhat does it contain?16:39
maldo you have /vendor?16:39
malcan you point me to your kernel16:40
eugenioyup /vendor exist too, /vendor/firmware has the same contents of /system/vendor/firmware (but actually /vendor is not a symlink, maybe I have copied it myself during my tests months ago and I don't remember)16:41
eugeniokernel source:
mortalglitchdidn't find much in systemctl, dmesg ended up showing quite a bit looks like at several points it's unable to read qcom,display-id or issue commands16:54
malolesalscheider: don't guess what is relevant, just show all16:55
malwrong highligt16:56
malmortalglitch: have you masked droid-hal-init?16:56
mortalglitchyeah, good call16:57
mortalglitchdid an unmask, rebooted : Systemctl and jounralctl going back over them now17:09
malmortalglitch: which android base?17:11
mortalglitchbase off LineageOS 14.1 (android 7.1)17:13
malmortalglitch: so you are probably missing the 14.1 porting stuff mentioned in faq (linked in channel topic)17:14
malmortalglitch: search for "14.1 porting" in that17:15
malmortalglitch: so you need to run one script to create some needed symlinks17:15
mortalglitchI had briefly saw that on r0kk3rz github earlier didn't see the exec it had mentioned and though the info may have been dated. Will check it out, thanks again mal17:16
malmortalglitch: there used to be more 14.1 specific things but now it's only running one script17:17
malin case you don't want to rebuild the image now then you can edit the script to create the links on the device but I recommend rebuilding the image so you won't be missing those next time17:20
eugenioby loading the firmwares through userspace at least the audio card is now detected18:48
eugeniotoo bad there aren't any ALSA UCM available for it (audio driver is rt5659)18:49
mortalglitchkeep making it a few steps forward at least, ran the script change rebuilt, reflashed los, flashed sailfish, was encrypted out of userdata again, had to fix that one more now i'm getting a ton of /dev/cpuset/xxxx : No space left on device20:07
mortalglitchI tried some forums suggestions of push 0 to the .mem files in certain cpuset directories but I still can't start the services and if I reboot it seems to revert all the changes20:08
*** ChanServ sets mode: +v T420:10
malmortalglitch: cpuset warnings or errors are irrelevant20:15
malmortalglitch: just get new versions of the logs I asked before20:16
mortalglitchah kk, will do20:17
mortalglitchSystemctl : Really Long DMESG: journalctl (fresh boot) :
malmortalglitch: have you masked any systemd services?20:53
mortalglitchdidn't mask anything this go round20:54
malI didn't see any sign of it trying to start the UI in journalctl maybe it did reach that part yet, also output of /usr/libexec/droid-hybris/system/bin/logcat could be useful21:05
mortalglitchlogcat dump
malmortalglitch: could you get a new journalctl output but wait a bit longer after the boot before you get it so it will contain all21:28
malalso check systemctl-user or systemctl --user whichever works21:30
mortalglitch long running journal dump21:32
mortalglitchI did see systemd messages as I scrolled through that time21:32
mortalglitchsystemctl-user results in sysd "user session is not running"21:33
mortalglitchsince reboot autologin@100000 user@0 and user@100000 all failed21:34
malhmm, what could cause user session to fail22:02
mortalglitchit's weird because after awhile I can systemctl start autologin@100000 and then start the other services22:17
malthere shouldn't be need to run that manually, need to think more tomorrow22:46
mortalglitchthank you for taking a look into it. I'm gonna to finish cleaning the house and then hack on it some more this afternoon.23:15
mortalglitchpasted wrong address lol23:31

Generated by 2.17.1 by Marius Gedminas - find it at!