Wednesday, 2018-12-05

*** ChanServ sets mode: +v T402:41
*** Gizmokid2005 is now known as Mogwai04:05
T4<adampigg> Messaging from depecher on latte today!07:57
T4<adampigg> @eugenio_g7 do you get reboots? Mine seems to turn off when not in use...when im using it it seems fine08:17
*** ChanServ sets mode: +v T408:25
T4<eugenio_g7> Yup! I guess I disabled the iTCO watchdog but I haven't tested it yet08:26
T4<eugenio_g7> s/I guess//08:26
T4<adampigg> Atleast im not alone!08:28
T4<adampigg> mal, spiiron, what needs configured in kernel/android side for watchdogs. tablet it rebooting when not in use. on boot, dsme says it cant open watchdog0, i guess it can open /dev/watchdog12:06
T4<adampigg> yes, lsof says /dev/watchdo is open by dsme, but nothing has watchdog012:07
T4<adampigg> kernel is doing some pinging12:20
T4<adampigg> 1.626955] iTCO_wdt: Intel TCO WatchDog Timer Driver v1.11 … [    1.627043] iTCO_wdt: Found a Cherry Trail SoC TCO device (Version=3, TCOBASE=0x0460) … [    1.627287] iTCO_wdt: initialized. heartbeat=30 sec (nowayout=1) … [    8.435644] iTCO_wdt: iTCO_wdt_ping … [    8.435777] iTCO_wdt: iTCO_wdt_ping … [   10.308781] iTCO_wdt: iTCO_wdt_pi12:20
T4<adampigg> krnlyng: which rule needs changed for....12:40
T4<adampigg> i have a /dev/block/platform/by-name/* with invalid links12:41
T4<adampigg> i  … need a /dev/block/by-name/* with valid ones :)12:41
r0kk3rz@adampigg: get your bt to do anything?12:42
T4<adampigg> r0kk3rz: sort of, it scans and finds stuff, but doesnt pair12:43
r0kk3rzwell thats something12:43
spiiroinT4: IIRC so far the "usual situation" has been that (without heavy patching) android kernels tend to have no usable-by-dsme wd device nodes12:44
T4<adampigg> spiiron, so, the usual fix is? :)12:45
T4<adampigg> it gets a small mention in the hadk, but not a lot12:45
spiiroinT4: porting the grumpy old driver written for jolla1 kernel / not using wdd12:47
spiiroinT4: the situation has been more like: we'd like to get wd reboots when user side system gets stuck, but the default setup is happy until kernel is whacked too12:48
spiiroinT4: so if you get hit by wd reboots without trying really hard, it would be something novel12:48
T4<adampigg> spiiroin: what does android do? afaict it has a watchdog service ... will that do? what is the point of dsme-wdd if it doesnt work?12:48
T4<adampigg> it seems to happen shortly after suspend, if not connected12:49
T4<adampigg> while in use its fine12:49
T4<adampigg> if usb is connected it stays up12:49
T4<adampigg> but, if i unplug and waiti a minute, it will reboot12:50
T4<adampigg> intel cherry-trail iTCO driver in all intel motherboards i think12:50
spiiroinT4: even without wdd, the dsme-wdd process still participates in system heartbeat and watches over dsme-server process - which in turn handles things like process-watchdog, etc12:51
T4<adampigg> what would happen if i didnt compile in itco-wdt to the kernel?12:52
spiiroinT4: hmmm.. "intel", "shortly after suspend" ... again IIRC, but I think what we have used have suspend/resume handlers that disable the hw wd on suspend and enable + kick at resume -> user space kicking matter only when device stays out of suspend12:56
T4<adampigg> Got links to them???13:06
T4<adampigg> Spiirion^^ would something similar be on tbj?13:46
*** ChanServ sets mode: +v T414:44
T4<adampigg> May have found a kernel patch to help15:14
T4<abhishek_0> Mi pad 2 runs on intel atom right?15:16
T4<adampigg> Yup15:17
T4<abhishek_0> Can it run mainline ?15:18
T4<adampigg> Possibly15:20
T4<adampigg> We have gfx through mesa, not hybris, and sensors should work15:20
mal@adampigg do you see the needed files in sysfs?15:21
malfor sensors15:21
T4<adampigg> Mal, i havnt messed with sensors yet,  touch, bt and suspend current focus15:22
T4<adampigg> I have the sensor kernel modules15:22
T4<adampigg> And fojnd some comments in the android init aboit load order15:23
r0kk3rzpoor piggz needs his keyboard to work15:25
T4<adampigg> I do115:26
r0kk3rz@adampigg: so why you using old kernel?15:27
T4<adampigg> Its the los kernel, and i havnt thought about the build sydtem changes to use a new one...but, i can maybe just use upstream android, doesnt have to be kernel.org15:29
T4<eugenio_g7> I had some trouble with modesetting on newer kernels (not on sailfishos, tried with generic linux distros)15:29
r0kk3rzi should hope cherrytrail support isnt so bad these days15:32
T4<adampigg> I read 4.11 had imptovements15:33
T4<adampigg> Mal, you know which rules file needs changed for fixing links? I probably should read about udev rule writing and syntax16:08
malthat can be overriden in device sparse16:13
T4<adampigg> So, telling dsme to not send wd ticks is a.great.way to scew up the user experience, and general joy of using a device17:34
r0kk3rz@adampigg: sounds amazing17:53
piggz_r0kk3rz: actually, it may have worked.....confirming18:18
piggz_r0kk3rz: it seems to have18:24
piggz_@eugenio_g7 maybe fixed reboots18:24
piggzmal: should all /sys mounts be removed on the grounds systemd refuses to mount them anyway?18:58
piggzkrnlyng_: mal: this works for me:19:13
piggzENV{ID_PART_ENTRY_SCHEME}=="gpt", ENV{ID_PART_ENTRY_NAME}=="?*", SYMLINK+="block/by-name/$env{ID_PART_ENTRY_NAME}"19:14
malpiggz: what difference does it have to the existing ones?19:36
piggzmal: it removes the platform folder and platform name folders from the path, which are not expected...19:37
piggzalso, one of those variables toesnt expand to anything anyway, and results in broken links19:37
piggzi think it is platform_name19:37
piggzmal: so, should all /sys mounts be removed?19:38
malpiggz: maybe ignore those in droid-hal spec19:47
piggzmal: i have.....19:47
malso what is the problem?19:47
piggzthe reason i say, is that systemd appears to refult them anyway19:47
piggzDec 05 19:20:33 Sailfish systemd[1]: sys-firmware-efi-efivars.mount: Cannot create mount unit for API file system /sys/firmware/efi/efivars. Refusing.19:48
piggzDec 05 19:20:33 Sailfish systemd[1]: sys-fs-pstore.mount: Cannot create mount unit for API file system /sys/fs/pstore. Refusing.19:48
piggzDec 05 19:20:33 Sailfish systemd[1]: sys-fs-cgroup-bfqio.mount: Cannot create mount unit for API file system /sys/fs/cgroup/bfqio. Refusing.19:48
piggzDec 05 19:20:33 Sailfish systemd[1]: sys-fs-cgroup-cpuset.mount: Cannot create mount unit for API file system /sys/fs/cgroup/cpuset. Refusing.19:48
piggzseems to be a systemd feature ?19:48
malpiggz: show the spec file19:49
piggzmal: no, this is before the .spec changes, so the build would still create those mount units19:49
piggzi am wondering if they should generally always be removed if systemd doesnt support them19:50
malif some mount fails and nothing really needs it then disable it19:50
piggzyes, i get that ... but im wondering if they should be skipped on creation anyway if systemd refuses to mount them
piggzthe erros message says:19:53
piggzCannot create mount unit for API file system19:53
malI don't understand why won't just add those to the spec so the mounts are never created and then forget about the whole thing19:53
piggzmal: i am :)19:54
malso why are we talking about this?19:54
piggzi am just saying, in general, all 'api' filesystems should be  removed19:54
malobviously if it will fail19:55
piggzmal: for ota/flash support, faq says to match to 'hardware' field from /proc/cpuinfo .... x86 doesnt have that!20:28
*** ChanServ sets mode: +v T420:31
piggzlooking at the flash_partition script, i guess i can use _any_ string that appears in /proc/cpuinfo!!20:34
r0kk3rzyeah i think its just a protection mechanism20:42
piggzr0kk3rz: a single character would match that rule ;)20:51
piggzr0kk3rz: having a shot at building android-4.14 with existing defconfig20:52
r0kk3rzwell i think its to stop BadThingsTM from happening if the config repo ends up on the wrong device20:52
malr0kk3rz: yes, that is the reason for it21:02
piggzr0kk3rz: it built....21:08
r0kk3rzdun dun dun21:10
piggzr0kk3rz: eugenio: so, it seemed to just turn off :(21:33
eugeniothe mainline kernel?21:35
piggznot mainline, android-4.1421:35
piggzi tried that first as an easy way to get a boot img21:36
piggzideally would like to use a minline next try21:36
piggzeugenio: my repos have fixes for mounting issues, and reboots21:37
r0kk3rzsome hacking required no doubt21:37
r0kk3rzcheck dts21:37
eugenionewer kernels just made the screen turn black for me, unless booting with nomodeset (but that would be pointless)21:38
eugeniopiggz, nice, I'll take a look21:38
piggzeugenio: well, screen turned black, but there was no usb activity either21:38
piggzit printed a eufi secure boot disabled message, which it doesnt usually, then went black21:38
eugeniopiggz, on which repository have you pushed the reboot fixes?21:43
piggzeugenio: config ... but you also need the mount fixes from device....21:43
piggzin the config, i disable the dsme watchdog tick21:43
piggzand with all the mount fixes, the android watchdog daemon is able to run and take care of it21:44
eugeniocould you point me at the commit? I see only the android mount points one21:45
piggzeugenio: ah, didnt commit the dsme fix yet, was testing for  bit...22:02
eugenionice, thanks!22:11
eugenioI wonder if we should create a shared org on github22:11
piggzmaybe aye22:12
piggzr0kk3rz: suggestions pls
piggzeugenio: hows sensors going? does screen brightness work for you?22:20
eugeniosensors noting yet, brightness works (haven't committed yet)... it only needed adjusting the path in mce22:22
eugenioI'll share soon™, recompiling the kernel now :)22:23
eugenio(stuck in fastboot currently)22:23
r0kk3rzpiggz: id disable the service and try and init it manually22:30

Generated by 2.17.1 by Marius Gedminas - find it at!