[Halium9-GSI] - Oneplus 3(T)

ernest

Bluetooth is rootfs dependent mainly. Cf post of @peat_psuwit.
Wifi - Not yet, I've found a nicer solution for 7.1 too so I'll update both. But while wait I can tell you we the two files should go. I'll see how it goes on h7.1 but seems way smoother in H9. I hope my update will make smoother H7.1 too.
As it's still under WIP and GSI is standardize, likely i'll create a patch OP3.
Flash ligth, i'm using a modified app .. an old indicator, I'll to recompile it for armhf rootfs.

I'm not sure what the firmware change does... I didn't flash specific one for the ubports version and likely they're from latest official Oneplus image. I don't see much problem.
Each time you'ill need to flash halium-boot, lineage modified and ubports GSi.

joe

Living proof (of concept):
https://www.youtube.com/watch?v=UyOcQtNDZwM

ernest

Hello, just to let some potential users know that i've updated the first post and add created a OP3_GSI_Fix_V1.X.zip to fix / compensate OP3 specific bugs.

danqo6

Can I ask about swipe to close app bug, is it still hard to close app?

ernest

@danqo6 sure, it's an issue in the GSI image, not phone dependent.

aanno

Very interesting. So interesting that I've tried - but I already failed with 'treblezing' the device - and the rest of the documentation is even more thrilling...

I still got some general questions about GSI on 3T. Perhaps somebody reading this can answer them:

Is 'treblezing' the future direction for 3T, i.e. will there be a 'vendor' system plus (device-independant) GSI images?
Are GSI images android version dependant? I.e. will there be different GSI images for android 9, android 10, and so on?
Is 'Halium9-GSI' a official Halium version? Will GSI be the direction Halium will take?
I still haven't understood what modifications are needed on boot and kernel for GSI. Can somebody elaborate on this?
[off-topic] It seems that there are other (unofficial) GSI image, e.g. lineage images. Is there a change that they run on 3T?

ernest

@aanno said in [Halium9-GSI] - Oneplus 3(T):

Very interesting. So interesting that I've tried - but I already failed with 'treblezing' the device - and the rest of the documentation is even more thrilling...

I still got some general questions about GSI on 3T. Perhaps somebody reading this can answer them:

Is 'treblezing' the future direction for 3T, i.e. will there be a 'vendor' system plus (device-independant) GSI images?

Are you refering to OP3 ubports future ? I'm not in the position to decide. I've supported halium 7.1 port, add my small participation to previous amazing work by fixing there and there bugs + enabling bluetooth over 6 months. The results is still far from perfect significant bugs remains...and some hardware issue like the fingerprint reader.
I tooks me 2 weeks to enable the device for Erfan GSI ubports....and the result was great in term of daily usage (Once i found a counter measure to the ofono/rild bug, but still not a proper fix)
Now I'm testing V10

Are GSI images android version dependant? I.e. will there be different GSI images for android 9, android 10, and so on?

Currently it's android 9 vendor depend.

Is 'Halium9-GSI' a official Halium version? Will GSI be the direction Halium will take?

Good question, I'm not the one able to answer.

I still haven't understood what modifications are needed on boot and kernel for GSI. Can somebody elaborate on this?

From the phone I'm limited in the answer but kernel requires modification to allows the android container and ubports to boot. I suggest to read the halium documentation and join telegram halium group.

[off-topic] It seems that there are other (unofficial) GSI image, e.g. lineage images. Is there a change that they run on 3T?

I didn't try, due to my lack of interest in android in general.

Stefano

@ernest hi @ernest could you shed some light on the performance, bugs, once you finish testing GSI v10? Thanks in advance.

ernest

@Stefano Sure, i'll do it, in a week or two.

A Former User

@ernest Where can I find this watchdog script you made? The PinePhone is suffering from similar problems that I have been trying to solve for a while, and (parts of) it might be useful there as well.

ernest

@kkeijzer said in [Halium9-GSI] - Oneplus 3(T):

@ernest Where can I find this watchdog script you made? The PinePhone is suffering from similar problems that I have been trying to solve for a while, and (parts of) it might be useful there as well.

I'm looking for a ratio between two output of ofono : battery_props_changed vs ril_netreg_get_signal_strength .
If there is too much battery_props_changed, I force ofonod and telepathy-ofono to restart.

I notice two issues and i've difficulty to assess if there are linked or not : ofonod doesn't mind that it is loosing the cell tower information ... Thus signaling to the stack above that everything is fine. I suspect an issue between rild and ofonod, however on H9, I didn't success to generate ril log
Then the other issue, is that somethime the dbus id of ofono is not updated then telepathy-ofono and the apps are lost (impossible to close a call even if the call is already finish).

I didn't yet create a repo however this is the script https://paste.ubuntu.com/p/KHS5DMNw3y/

I need to assess if that is still valid or not in V10 in arm64 (V9 is armhf).

Ps : you need to have ofonod started in debug mod : exec ofonod -ddd -P provision,udev*,dun*,smart*,hfp_bluez5,stktest,sap

A Former User

@ernest A lot seems familiar. On the PinePhone the signal indicator always seems to work, but occasionally the modem just doesn't. So calls and texts don't come in and mobile data doesn't work any more.

When I then try a dbus call like dbus-send --print-reply --system --dest=org.ofono /quectelqmi_0 org.ofono.Modem.SetProperty string:"Powered" variant:boolean:false, attempting to restart the modem, it will time out. So clearly something in oFono is stuck. (Normally that query should disable the modem and variant:boolean:true should power it up again.)

The cleanest way to fix it is sudo service ofono restart (better than pkill -9 ofonod I'd say), after which mobile data starts working again. But then I run into the problem that the dialer app will show "No signal", and outgoing calls being impossible, with the logs printing something like:

error: "org.freedesktop.DBus.Error.NameHasNoOwner" message: "Could not get owner of name 'org.freedesktop.Telepathy.Connection.ofono.ofono.connection_640ce7767735c59d6fbc2542a9a06c68': no such name"

The fix for that is running pkill telepathy-ofono (there doesn't seem to be a service to restart, and I'm not sure how this is spawned) with some sleep timers I'm still working on, but that also seems similar to what you're doing. It looks like you're sleeping 20 seconds after killing ofonod, probably to give it time to re-initialize, but upstart would only give an exit code when ofonod is back up. But still you'd probably need some time inbetween that way, because my experience is that killing telepathy-ofono right after upstart reports oFono as restarted won't work.

My problem only is that I'm still looking for a way to do a non-destructive API call to figure out whether oFono is locked up or not. I can't look at RIL logs because there is no RIL on the PinePhone.

On the PinePhone, sometimes the technology preference option disappears, and the org.ofono.RadioSettings.GetProperties call would return "operation already in progress", which is a trigger for me. But sometimes the modem also locks up without this happening, so it's not ideal.

Maybe /usr/share/ofono/scripts/get-serving-cell-info could be useful for you. You could check if this stops fluctuating and have your counter look at that?

ernest

@kkeijzer
Very interesting, it seems really similar what your are experiencing....

I've played a lot with busctl to try to extract a proper bus call response/property without much success, using the ofonod script was ending with a timeout response or very long response that was not acceptable for a watchdog.
I tried to switch on and off the modem through the ofono script and the results was not the expected one.

It seems that ofonod doesn't or dbus update it's id then telepathy is lost. I'll try to stop the service in a cleaner way, thanks.

Looking in sailfish ofonod github and forum, I didn't find a similar issue. Moreover OP3 with Halium7.1 doesn't present this bug.

danqo6

I started using v10 without modifications a week ago, most things are working fine and many rootfs issues was fixed with switching to android9 branches. Once a day I need to reboot phone due cellular network, opening camera is causing system crash, dash charging and tri-state switch probably don't work.

ernest

@danqo6 which phone do you have a 3 or a 3T ?

On the 3T the phone switch off when taking a picture...can find any relevant log so far

Look in my gdrive i've dropped a file fixing dash, phone cell and switch

Stefano

@ernest hi, how did the testing go? any major issues?

ernest

@Stefano

It seems way smoother, I've upload a fix for the GSI V10. However on the OP3T, taking a picture shutdown the phone ....
So far I've no clue as the logs doesn't seems talkative and also I'm wondering if this issue is only for the 3T.
I'm looking for help on this bug.

poVoq

Hmm, this seems interesting. If you say "way smoother", that is also compared to the normal Halium7 port?

What about the bottom edge display? Does it have this partially desktop mode like on the Halium7 port as well, where many apps show the bottom edge as a button (like on convergent desktop) and thus break the layout of the app?

danqo6

Way smoother than halium 7.1.
Bottom edge is normal, but I would like that to be configurable, I liked that button. Bottom arrow in some apps is moved a bit up, not sure why

ernest

@poVoq said in [Halium9-GSI] - Oneplus 3(T):

I mean H7 (arm64) less smoother than GSI V9 (armhf) less smoother than GSI V10 (arm64).
(I've worked on the H7 port up to May this year... and didn't performed test with an up to date phone)

Really sad, that they changed between V9 and V10 the camera back-end management, that make the android layer crashing hard