Observe radio user’s CPU usage and time of process /vendor/bin/hw/mtkfusionrild -l libmtk-ril.so
Put the phone into flight-mode. Again observe the CPU usage of the process.
EXPECTED RESULT:
The process should remain dormant when the flight mode has been enabled at minimum.
Equivalent process qcrild on Sony X10 III running SFOS remains mostly dormant in all situations.
ACTUAL RESULT:
Process mtkfusionrild remains active in all situations, constantly consuming CPU between 1 - 5 %.
MODIFICATIONS:
No modifications.
ADDITIONAL INFORMATION:
Probably safe to assume that the device cannot go into deep sleep when mtkfusionrild is running at this pace constantly.
This behavior has been reproduced at least on two separate J2 phones with three separate SIM cards.
WORKAROUND:
Update and workaround in 2026.7.28: I think this is a race condition between SFOS and radio firmware.
Enable Developer mode under SFOS Settings.
If not enabled, enable the option to show Restart button in the power button menu.
Longpress the power button and choose Restart icon in the menu (top right).
Restarting SFOS using this method does not reboot the radio firmware, which then prevents the race condition. It does not trigger again until you completely shut down the phone and power it back on manually.
Seems that this problem is a race condition between the operating system and radio firmware. Workaround added into the above bug report: Just restart the OS (and not the phone). Do not power off and power on; the full power cycle is what triggers the issue in the first place.
I had been using poweroff / power on cycle to troubleshoot for days. This confirms that the problem happens always after a full power cycle. I’m guessing that it also explains why the issue has gone unnoticed earlier; The devs are likely to more often just reboot/restart the devices and not to power them completely off.
I got the type where restart does not cut it today. I had been playing around with mobile network settings and had manually selected the network. The issue triggered and continued through the restart.
Then I chose automatic network selection and restarted again, after which the process CPU usage returned to normal levels.
There is also a constant 2-3% CPU usage by system interrupts, which is very high. This may or may not be related.
Comparing the /proc/interrupts counters over a minute, I see about 150000 interrupts from arch_timer and another 5000 interrupts from mtk_cmdq.
Furthermore, there is a lot of continuous dmesg spam. I’m not sure whether to open a separate thread for this. Here is a three second log fragment showing the repetitive messages:
After 48h noticed mtkfusionrild constantly 0.3% cpu. Power off/power on cycle mtkfusionrild mostly 0% time to time max 0.1%
network o2.cz 4G/5G volte enabled.