BraX3 touchscreen controller interrupt storm: ~4000 interrupts/s with no touch, ~half a CPU core, phone freezes

Device: BraX3, iodéOS 7.8-20260930-brax3 (also present on 7.6-20260623-brax3), kernel 5.15, MT6835

Symptom: The phone occasionally freezes completely for a while, then recovers. Looking into it over adb, the touchscreen controller (Omnivision TCM, SPI3, IRQ 157) is firing interrupts continuously whenever the screen is on, even with no finger anywhere near the screen. The kernel thread serving that IRQ plus the SPI thread together use roughly half a CPU core all the time, and load average sits at 12-18 on an idle phone.

Measurements (screen on, nothing touching the screen):

$ adb shell "grep '^ *157:' /proc/interrupts; sleep 5; grep '^ *157:' /proc/interrupts"
157:  4733178  0 0 0 0 0 0 0  mtk-eint  1 Level  omnivision_tcm
157:  4753763  0 0 0 0 0 0 0  mtk-eint  1 Level  omnivision_tcm

= 20,585 interrupts in 5 s (~4,100/s). Earlier samples went as high as 36,000 in 5 s. 4.7 million interrupts 22 minutes after a fresh boot.

$ adb shell "timeout 5 getevent -t /dev/input/event2 | grep -c ."
0

= zero input events from the touch device in the same period. The controller asserts the attention line nonstop, but the driver finds no touch data to report.

$ adb shell top -b -n 2 -d 3 -s 3 -o %CPU,TIME+,ARGS | grep -E 'omnivision|spi3'
18.3   5:05.79 [irq/157-omnivision_tcm]
10.3   2:44.50 [spi3]

What I have ruled out:

  • USB / charging: same rate plugged and unplugged (tested over a 2-minute unplug).
  • OS build: identical on 7.6 and 7.8.
  • Settings: high touch polling rate, high touch sensitivity, glove mode, touch hovering all off; wake gestures on or off makes no difference.
  • Uptime: present within 3 minutes of a fresh boot, so not something that builds up.
  • Screen off: interrupts stop completely while the screen is off and resume instantly at screen-on, so the driver’s suspend path works; the problem is in active mode.

Touch itself still works normally. The interrupt is level-triggered, so this looks like the controller signalling data that the driver never acknowledges, i.e. a firmware/driver protocol mismatch, or a hardware fault holding the line low.

Question for other BraX3 owners: could you run this with the screen on and nothing touching it, and post the two numbers?

adb shell "grep '^ *157:' /proc/interrupts; sleep 5; grep '^ *157:' /proc/interrupts"

On a healthy device the first column should grow by a few hundred at most in 5 seconds. If yours grows by tens of thousands like mine, this is a driver issue in the brax3 kernel and affects everyone’s battery and responsiveness. If yours stays low, my unit has a hardware fault.

Question for Brax / iodé: can you check the omnivision_tcm driver’s interrupt handling in active mode? The driver’s sysfs node (/sys/bus/spi/devices/spi3.0/omnivision_tcm.0/) exposes info and irq_en, but they are root-only so I cannot read the controller’s firmware mode myself. If you want a dmesg from the touch driver I can provide one if you tell me how to get it without root.

Is it possible to get the sources for the omnivision_tcm driver?

Thank you for the write up. We’re escalating this to iode devs. We will share feedback here once we get this looked into. If you don’t see the code for the driver in the repo, then it is not available. We’ve published everything we’re legally allowed to (there are restrictions by supply chain partners on some components).

This is not a unique problem just to Brax. Will wink and open_slate have similar issues? What would it take to get these companies to release the code as Free Software?

A major industry shift will take that. We’re pushing for it but it will take a lot of time.
open_slate will have open sourced drivers, as it is getting mainline linux kernel support.

My Brax3 freezes randomly regardless of APP being used. My only solution is to lock the phone and PIN in again. I’m on 7.9-20261020-brax3. Thanks.

can you run the adb shell commands from my original report, to see if you also have the same issue. It would narrow down if the issue is hardware or firmware

Here is what I got running those commands. I did not run exactly in order. Not sure what all that means either :smiley:

$ adb shell “timeout 5 getevent -t /dev/input/event2 | grep -c .” 0
grep: 0: No such file or directory
$ adb shell “timeout 5 getevent -t /dev/input/event2 | grep -c .”
0
$ adb shell “timeout 5 getevent -t /dev/input/event2 | grep -c .” 0
grep: 0: No such file or directory
$ adb shell top -b -n 2 -d 3 -s 3 -o %CPU,TIME+,ARGS | grep -E ‘omnivision|spi3’
18.5 165:09.98 [spi3]
37.0 309:51.68 [irq/157-omnivision_tcm]
16.0 165:10.03 [spi3]
33.6 309:51.78 [irq/157-omnivision_tcm]
$ adb shell “grep ‘^ *157:’ /proc/interrupts; sleep 5; grep ‘^ *157:’ /proc/interrupts”
157: 244499616 0 0 0 0 0 0 0 mtk-eint 1 Level omnivision_tcm
157: 244536765 0 0 0 0 0 0 0 mtk-eint 1 Level omnivision_tcm

Thanks @raajah, that confirms it. Your unit shows the same thing: 37,149 interrupts in 5 s (~7,400/s) with zero touch events, and the IRQ thread has burned over five hours of CPU since your last boot. Two devices on two builds (7.8 and 7.9) with identical behaviour means this is not a faulty unit, it’s the omnivision_tcm driver or controller firmware on every BraX3. Your random freezes are very likely the same cause.

@tech-support: is there any update from iodé? Since the driver source is closed, the fix needs to come from the vendor. Two concrete things to check: whether the controller firmware is being left in a mode the driver doesn’t service (the level interrupt is never deasserted, so the driver reads nothing yet the controller keeps asserting), and whether irq_en under /sys/bus/spi/devices/spi3.0/omnivision_tcm.0/ is being toggled correctly on resume. Happy to run anything you need over adb. Also, this topic is set to auto-close on 2026-11-16; could you keep it open until there is a resolution?

We’ve escalated this issue already to the solution house (the one designing the PCB). We’re awaiting a fix.

Has there been any update on a fix? My phone freezes within 5 minutes on any screen or app and a pwr button press twice is the only solution at the moment. I can’t even answer calls without using the pwr button option most times. Thanks.

When it happens to me, I get the same behaviour.