@crispyoz @Lazar-Demin
Yes, it worked. I choose the omgega2-bootloader instead of omega2p-bootloader accidentally.
After soldering with a brand new Omeba2S+ module, I carefully upgraded bootloader.

Here's the pre-compiled bootloader for others' future reference.
https://github.com/dumtux/omega2-bootloader/releases/tag/tower-bootloader-v0.1

After disabling GPIO11 default behavior from Omega2's bootloader, I could see some different behavior of MCP2515 kernel driver.
I could see some short impulses while initializing MCP2515 driver and enabling can0 interface.
I didn't see such impulses before with the default bootloader.

I'll keep posting further results.

Yes, it worked. Thank you.
I had to use different pin numbering as @luz explained in this answer for my CAN bus question.

You can add as many vias as you want under the Omega2.
Just avoid placing via on bottom pads, which are used for GND.

An example open-source PCB design is available on pcbcrew-org/omega2-4g-gateway.
It is prototyped and tested in the field by PCBCrew prototyping service.

At quick look, I am seeing the VIO is floating. It hould be connected to 3.3V internal regulator of CP2102 or just 3.3V power.
Here's a working circuit you can reference.
https://github.com/pcbcrew-org/omega2-4g-gateway/blob/develop/doc/schematics-v3.0a2.pdf

In my opinion, it is spidev (package name, kmod-spidev), as it is the working module on OpenWrt 22.03 build for Omega2.
I could see hardware SPI driver is present under /dev/spi* after installing it.
Havent tested it yet by myself, though.

If you're not a regular Linux user, setting up a build environment maybe not so quick and easy.
This online image downloader also can help.

https://firmware-selector.openwrt.org/

Choose "Omega2" or "Omega2+" as per your device model.
And customize packages as your need.
It is faster than normal local building processes.

This might help, https://github.com/dumtux/agpio-switch-mt7688
I wrote this for enabling all 5 ethernet ports for Ethernet switch.
But I think you can modify this for disabling all of them to use digital I/O, by modifying AGPIO_CFG.

What does dmesg say?

That image file you're trying may not contain WiFi kernel driver module.
If you post dmesg result, I can tell if WiFi driver is pre-packaged or not.

We were able to use Omega2S on cutom PCB without any hardware problem.
The only issue was SPI driver issue (software issue).
However, we could still use the made custom boards for other purposes and ran well for a long time.

Here's our hardware design repository.
https://github.com/dumtux/omega2-4g-gateway

And this is our PCBA supplier, where we made PCBs.
https://pcbcrew.com

You can add as many vias as you want under the Omega2.
Just avoid placing via on bottom pads, which are used for GND.

An example open-source PCB design is available on pcbcrew-org/omega2-4g-gateway.
It is prototyped and tested in the field by PCBCrew prototyping service.

Thank you @luz
Right now, I gave up solving this MCP2515 issue on OpenWrt 22.03.
We decided to move to ESP32 as the CAN bus was a critical requirement for our projects.
Although this problem is not solved on my side, I hope it could be a little help for people's future projects related to Omega2 + MCP2515.

@luz
I tried p44b build script first of all.
And I got this result.

It seems like the build script doesn't support OpenWrt 22.03 yet.
It totally make sense, as many file locations and contents are updated to kernel 5.10, it would mess if I force to apply all patches globally.

I am now on the way to manually applying the GPIO patch to OpenWrt 22.03.
But it's not easy for me as the gpio-mt7621.c seems like changed.

@luz
I also would like to see your patches as @Gerhard-Bertelsmann said.
Is it somewhere in public repo?

I just did another experiment.
I disconnected external sensor.

Omega2 -> MCP2515 -> (open) -> Oscilloscope

If I enable the CAN bus like this, I can see output CAN packet from the oscilloscope.

ip link set can0 up type can bitrate 125000
cansend can0 01a#112233

Next, I enabled CAN bus with loopback mode.

ip link set can0 down
ip link set can0 up type can bitrate 125000 loopback on
cansend can0 01a#112233

No CAN packet waveform I can see!
To see it again, I should re-enable the CAN bus without loopback mode.

ip link set can0 down
ip link set can0 up type can bitrate 125000 loopback off
cansend can0 01a#112233

Any thought from this loopback mode behavior?

Can you profile digitalWrite function only?
Like this,

 while(1) {
   digitalWrite(PWM_LEFT_PIN,0);
   digitalWrite(PWM_LEFT_PIN,1);
 }

With this code, what's the wavelength of output signal on LEFT_PIN ?

@Gerhard-Bertelsmann
I experienced similar error messages when I used the wrong GPIO number.
For example, I tried this.
interrupts = <75 0x02>
And it gave very similar error as yours on kernel module loading.

As far as I used the right GPIO number 11, the kernel module loading was just successful.
I don't have a clue if this means the issue was eliminated in 22.03.

When I connect two boards via CAN bus, I could only enable CAN driver from one board.
Currently, I am in a different situation after disabling GPIO11's default behavior of the bootloader.
I can enable both side drivers. And each board can send CAN packets by cansend.
However, the interrupt is not invoking and candump still prints nothing.

@crispyoz @Lazar-Demin
Yes, it worked. I choose the omgega2-bootloader instead of omega2p-bootloader accidentally.
After soldering with a brand new Omeba2S+ module, I carefully upgraded bootloader.

Here's the pre-compiled bootloader for others' future reference.
https://github.com/dumtux/omega2-bootloader/releases/tag/tower-bootloader-v0.1

After disabling GPIO11 default behavior from Omega2's bootloader, I could see some different behavior of MCP2515 kernel driver.
I could see some short impulses while initializing MCP2515 driver and enabling can0 interface.
I didn't see such impulses before with the default bootloader.

I'll keep posting further results.