Friday, 2019-09-20

Hacker1245Back with the Moto G5 port, any idea why I can only access during init-debug in real rootfs?11:20
Hacker1245UI does not start11:20
Hacker1245After continuing init connection cuts off11:20
malwhich android base? have you disabled selinux correctly? have you tried masking droid-hal-init systemd service?11:21
Hacker1245LineageOS 14.1, disabled in cmdline, yes (didn't help)11:24
Hacker1245Audit is set to 0 too as the device would reboot if it was unset11:25
malso now it doesn't reboot?11:25
maldo you see anything pstore or last_kmsg after such reboot11:25
malusually audit=0 is not used11:26
Hacker1245Yeah I know that11:26
malhave you tried persistent journal log?11:26
malwhat did that show?11:26
Hacker1245But I don't have any logs after it if I remember correctly11:27
malwhere did you look for the logs?11:27
Hacker1245should config_audit be set to n in defconfig?11:27
malconfig_audit should be =y11:27
Hacker1245Let me recompile the kernel with that set...11:27
malonly some very old devices had it as =n11:28
malmost modern devices need to have audit on11:28
malat least based on experience11:28
Hacker1245Hmm audit is set to y11:29
Hacker1245let me try to remove audit=0 and try again11:29
Hacker1245Btw, where is pstore located?11:29
Hacker1245Got the log, let me paste it11:37
Hacker1245Hmm, maybe removing the console from the boardconfig might help11:39
malyou can try that11:40
Hacker1245Since I remember that I had a similar issue with being unable to open the console with Halium and UT11:40
T42<elros34> you were supposed to fix droid-hal-init. Have you fixed it?11:41
malHacker1245: some watchdog is triggering11:41
Hacker1245Yeah I saw that, but which one?11:41
maldid you do what elros34 said, not sure which exact thing he is talking about but you seem to have talked earlier11:42
Hacker1245I'm not sure what's the problem with droid-hal-init11:43
Hacker1245The defconfig is correct, partitions seem to be mounted correctly11:44
T42<elros34> with audit=0 he just had usual failed to initialize property area11:44
mal@elros34 maybe missing *contexts files from straggler files?11:45
malusually that solved property are issues11:45
Hacker1245removed console, still fails11:45
T42<elros34> I guess we didn't investigate yet11:45
mal@elros34 the issue is that droid-hal build doesn't complain about unpackaged files like it used to11:47
Hacker1245Any ideas on what to try?11:47
Hacker1245put the console back in the cmdline11:47
T42<elros34> @Hacker1245 do you have any *contexts files in /11:47
Hacker1245On the device?11:49
T42<elros34> sure11:49
Hacker1245In the /data partition or in /11:49
Hacker1245In / there are file_contexts, property_contexts, seapp_contexts and service_contexts11:51
T42<elros34> compare them with out/target/product/$DEVICE/root/. Have you added anything to straggler_files?11:54
Hacker1245didn't add anything there11:55
Hacker1245gonna compare11:55
Hacker1245Seems like some things are missing from the property_contexts on the device12:00
Hacker1245file_contexts are missing in out/target/product/cedric/root/12:01
T42<elros34> ee what? Usually it's the opposite.12:02
Hacker1245well that's how it is12:04
Hacker1245There are also some MultiROM lines on the device but that's most likely because I experimented with porting MultiROM on there12:05
Hacker1245There is stuff missing from service_contexts on the device too12:07
T42<elros34> ok I added short instruction: You can skip strace for now, if it helps then make "Notes on LOS14.1 Porting "12:08
Hacker1245I'm back12:10
Hacker1245gonna follow that info12:11
Hacker1245How can I check which files are packaged in the RPM?12:14
Mister_MagisterHacker1245: some programs like Ark can see them12:16
Hacker1245nice, I have ark installed12:16
Hacker1245gonna check that soon12:16
Hacker1245There are no context files in the droid-hal-cedric.rpm12:34
malHacker1245: those are in droid-hal-detritus12:35
Hacker1245well the wiki says Add any symlinks and files located in out/target/product/$DEVICE/root/ which are not already packaged in droid-local-repo/$DEVICE/droid-hal-$DEVICE/droid-hal-$DEVICE.rpm12:37
malHacker1245: that is not completely clear what it mean, it means that you ignore any files you already have in droid-hal-$DEVICE.rpm and add everything else to straggler files, usually the symlinks and *_contexts are the relevant things12:44
Hacker1245Add init too?12:45
malhow about you just show what you have in root/12:48
Hacker1245contents of root/
Hacker1245What I have in straggler_files so far12:49
malremove init.zygote32.rc and fstab.qcom, add vendor12:56
Hacker1245anything else?12:56
malnothing critical12:57
Hacker1245building packages13:10
Hacker1245next rebuilding the rootfs13:10
Hacker1245btw, should I have audit set to 0 or not?13:13
malno audit=0 to kernel command line13:35
Hacker1245mal: After reinstalling the rootfs mask droid-hal-init?14:03
malof course try to see if that helps14:08
Hacker1245If I can even boot it with audit=014:09
Hacker1245The detritus package does not build14:25
Hacker1245mal: the package does not build for me14:32
malHacker1245: show you droid-hal spec14:33
Hacker1245mal: there14:37
malHacker1245: do you don't have anything related to detritus in droid-local-repo?14:40
Hacker1245mal: I only have this in droid-hal-cedric
Hacker1245and yes I did rebuild droid-hal14:42
Hacker1245contents of jolla-hw-adaptation-cedric.yaml14:44
malthat makes no sense at all14:46
maldetritus in generated automatically if straggler_files define is present14:47
mal*is generated14:47
Hacker1245That's why I was weirded out too14:47
Hacker1245Anything I could try?14:47
malcheck droid-hal build log14:48
malit should be in ANDROID_ROOT14:49
maldid you run without any parameters?14:49
malor with some parameter14:49
Hacker1245I think with no parameters14:49
Hacker1245Well detritus is in logs14:52
Hacker1245mal: See this
malare you sure you don't have some mistake in environmental variables (ANDROID_ROOT) and it would be writing it to some wrong place?14:58
malHacker1245: if you look at the dates in the rpms those seem to be from several days ago14:59
Hacker1245I think I found the issue14:59
malHacker1245: so none of the changes you have made earlier today actually ended up on the device14:59
Hacker1245let me rebuild the packages15:00
Hacker1245now I have detritus15:03
malwhat was the issue?15:03
Hacker1245environment variable15:04
Hacker1245I don't know how I had a hadk1 folder15:04
piggzmal: updated piggz-o-vision work nicely15:52
Hacker1245mal: well device still crashes15:57
Hacker1245didn't mask droid-hal or enable debug yet15:57
Hacker1245Mask droid-hal-init?16:00
Hacker1245something still seems to kick the watchdog16:04
Hacker1245mal:pstore output
malHacker1245: are you 100% sure you have correctly disabled selinux?16:53
Hacker1245hmm, disable CONFIG_SECURITY_SELINUX in the kernel too?16:55
Hacker1245Then everything should be properly disabled16:57
Hacker1245the selinux bootparam is enabled in the kernel16:57
Hacker1245selinux is disabled in cmdline16:57
malverify the kernel config one more time, show the .config you have in out/16:59
Hacker1245hmm kernel even says [    0.002467,0] SELinux:  Disabled at boot.16:59
malkernel config17:01
Hacker1245SELinux is enabled in the out kernel config17:05
malcan't you just show the file17:05
malnot sure if you understand how it's disabled17:06
piggzprint /proc/cmdline ?17:06
Hacker1245That config is in out/17:07
malthat looks ok17:13
Hacker1245What about this? CONFIG_SECURITY_SELINUX=y17:14
Hacker1245still selinux is set to 0 on boot17:15
piggzHacker1245: and what is /proc/cmdline? if u canget that far17:21
T42<srLuxint> I have some progress on my issue so far! Actually it seems like my clue on audioflinger callbacks not triggering was correct. I've managed to trigger 'em manually with help of`/system/bin/service` and right now I have the right vsid to set HW params to turn on voice, where vsid is voice session ID I guess. Now it's time to find out w17:21
T42hy the callback are broken.17:21
Hacker1245piggz: Can't check /proc/cmdline as it crashes before I get access to telnet17:21
Hacker1245You can check the cmdline from the pstore output17:21
malHacker1245: you don't understand how it's disabled, it's meant that those configs are like that, it's kernel commandline that then disables it with selinux=017:24
piggzHacker1245: what is in your where the command line is specified17:24
Hacker1245I know that cmdline disables it mal17:24
malHacker1245: but you had connection before17:24
Hacker1245Yes, with audit-017:25
Hacker1245piggz:BOARD_KERNEL_CMDLINE := console=ttyHSL0,115200,n8 androidboot.console=ttyHSL0 androidboot.hardware=qcom user_debug=30 msm_rtb.filter=0x23717:25
Hacker1245BOARD_KERNEL_CMDLINE += ehci-hcd.park=3 androidboot.bootdevice=7824900.sdhci lpm_levels.sleep_disabled=1 vmalloc=350M selinux=017:25
deathmist@srLuxint I actually still have the exact same issue :(  no call audio with afglue or hidl module. still not sure if it's a config issue on my side, very well could be as Danct12 got it working with the hidl module on the same 8.1 base17:36
maldeathmist: did you try hidl module on your device?17:37
malah you did, strange that it didn't work17:38
maldeathmist: did you update submodule before trying hidl module?17:38
mal*config submodule17:38
piggzHacker1245: mal: so, should selinux be 0 or not?17:40
deathmistmal: I remember doing so, /usr/libexec/pulse/hidl-helper is running for sure after booting the new builds (and completely dropping afglue packages)17:40
piggzits set to 0 in BOARD_KERNEL_CMDLINE17:41
malpiggz: it should be 0 afaik17:41
piggzok, its forced off, and should be, but reboots?17:43
malpiggz: the issue must be something else17:45
piggzyes, looking at the panic, it looks like something lese17:46
malpiggz: although the panic suggests selinux based on web searching17:47
maldeathmist: which pulse droid module version do you have on the device?17:49
deathmistmal: pulseaudio-modules-droid is @ 12.2.79-1 and hidl module is v1.2.0-1 on the build I run currently17:51
malok, those seem to be the latest versions17:53
T42<srLuxint> @deathmist [@srLuxint I actually still have the exact same …], The only thing I can say for certain is that somehow on my system PA doesn't get any call info. As I say I've managed to fix it manually and now I'm totally sure it's somewhat an issue with higher level component then PA itself. Still have no idea what HIDL is, will inves17:54
T42tigate on that on Monday.17:54
T42<srLuxint> Guess I'll keep you informed on any info I get.17:56
mal@srLuxin hidl means the android binder communication thing17:59
T42<srLuxint> Well but the AF is binder-baser too, isn't it?18:00
malwell not the same way, it goes via the android api18:01
malhidl module talks directly to binder18:01
T42<srLuxint> As far as I get it we have AF service which serves callbacks from other system components via some interface name and each component gets an access to this interface from servicemanager via interface name (at least af glue related modules are written in such a manner). And for now we have AF on systems at least lower than 8.0 to only18:06
T42 set call info. Which is fine, yet I don't know who exactly talks to AF service. What's the story with this HIDL thing?18:06
malhidl module does pretty much the same thing as afglue but using different api18:09
malsome devices don't support the afglue way anymore18:09
T42<srLuxint> I mean basically I have AF service name and worst case scenario I just grep all the binaries to find the one which would had to talk to AF.18:10
T42<srLuxint> Well I guess then I'll dig into hidl module sources too. :D18:12
Hacker1245mal: Any idea what the kernel panic is caused by?18:13
deathmist@srLuxint I used this before as it allowed me to have call audio with "pactl set-card-profile droid_card.primary voicecall-voicemmode1" during a call. I even made a script for it but it just ended up horribly bad and didn't work properly18:29
deathmistit was purely for debugging tho and concluded with this
T42<srLuxint> deathmist: ln which way horribly? Did something break or just things were not quite smooth enough?18:31
deathmistbasically it didn't work reliably at all, I'd have to run it manually from a computer each time and the voicemmode number changed if I answered vs if I placed a call, it just was a mess18:32
T42<srLuxint> Thanks for your links. Seems to be quite useful.18:35
T42<srLuxint> Ok, I see. Guess it the very VSID I was talking about previously.18:35
deathmistI'd love to sort this out once and for all since basically all other functions of my W.I.P port for the OnePlus 5 work perfectly fine18:35
Hacker1245mal: Strange19:30
Hacker1245Now it crashes even with audit=019:31
Hacker1245Leaving the pstore log here19:35
Hacker1245Bye everyone19:48

Generated by 2.17.1 by Marius Gedminas - find it at!