search

Divested-Mobile / DivestOS-Build

Everything needed to build DivestOS, a more private and more secure aftermarket mobile operating system.
https://divestos.org/index.php?page=build
Other
226 stars 31 forks source link

[vayu]wifi is broken under 20.0 #340

Open dagithubun opened 2 months ago

dagithubun commented 2 months ago

i'm aware it's already been discussed here but i thought maybe we need one separate issue going on particularly about vayu.

here's the log dump for anyone who's interested in debugging: wifi_log.txt

owners of vayu, if you've recently found any clues on the bug, please report here. many thanks @SkewedZeppelin.

SkewedZeppelin commented 2 months ago

I still don't know what causes this.

can you confirm you're on firmware version V14.0.3.0.TJUMIXM

dagithubun commented 2 months ago

can you confirm you're on firmware version V14.0.3.0.TJUMIXM

yea i suspected the firmware at first too. so i flashed V14.0.3.0.TJUMIXM's fastboot (size 5.6GB) using MiFlash before installing divestOS 20.0. the issue still persists tho

goshuberto commented 1 month ago

I still don't know what causes this.

can you confirm you're on firmware version V14.0.3.0.TJUMIXM

Hi! I'm still new to DOS and tried to get in contact (by donating via LiberaPay, which doesn't allow to attach an telegram :-( Nor was my attempt to join via XMPP successfull, likewise my post in the XDA Forum for the pro1x device). Please remove non-working contact options, it's wasting a lot of time).

Back to topic: At lease for me the root cause is a wrong country code (for the 2.4GHz spectrum). Usually at least a european SIM should advertise the correct bandplan allowing to use channels beyond 11 - which works on LOS. But DOS can't join/see base stations transmitting beyond channel 11 (somehow e.g. Android-Hotspot selects channel 12 for no good reason - surrounding bssids use 1/5/9/13 pattern (only obsolete 802.11b had 22MHz spectrum per channel, we can use 4x 20MHz wide channels in europe with the 1/5/9/13 pattern)! The bad habit of using 1/6/11 pattern is commonly known by other vendors, but completely ignoring the environment and transmitting copmpletely out of pattern is a poor jamming feature of android).

To me, this looks like a deblobbing fallout... Just a guess!

Offtopic again: Likewise the problem with the finger print reader on my QX1050. It's working fine in LOS, while rollout fails on DOS. Even worse, it looks like the TEE keystore isn't available on DOS on my QX1050/bengal/662 from a very brief look, without inspecting any code (besides flying through deblobber.sh). Unfortunately I don't have any android dev env here, nor much time or other junctions to AOSP - I'm happy consumer of GOS out-of box on my pixels.

Anohter severe issue for me is that the Android Key Attestation Test App (https://github.com/vvb2060/KeyAttestation) fails on my QX1050. Im not an expert regarding StrongBox/TEE attestation, nor on bengal/Qualcom at all... But currently I must assume that keys can't be stored in a secure way on DOS, which seems to be the root cause for the failing FingerPrint roullout.

I'd highly appreciate if we can somehow find a way to get in contct besides an issue tracker. I also highly appreciate your DOS work!