Hardware isolation of user applications

[yiiii777]

The main idea: a separate processor for the "periphery", drivers and part of the framework, and another separate processor for "user" applications. For example, ESP32. There are bluetooth and Wi-Fi, as well as enough RAM to run software interpreters of user applications, such as lua, micropithon, espduino, j2me. Surely users will want to run some kind of emulators of other platforms, such as Z80 - this is also possible. The firmware of this microcontroller is not user changeable, but can be upgraded as whole system update.

When a user wants to launch his own custom application (one of thousands), the code for this application is taken from the SD card and sent to the RAM of the second microcontroller. Perhaps it will be some of the STM series. If an application hangs, does an incorrect operation, or something unexpected happens, the processor can be restarted without restarting the shell from the main processor.

Thus, the user has complete freedom to control the hardware. The user can choose any operating system or even use bare metal. The user is not limited by the number of device flashing and can do it on the fly.

Communication between microcontrollers can be carried out through UART, which will include interfaces for drawing UI, working with a SD card, radio modules and other peripherals.

[glitchcore]

We won't add second processor only for solving "isolation" problem, and if all user-application will run on second processor, they will have common address space again.

This idea creates a ton of tasks to support many firmwares, update it and interact between.

[yiiii777]

will have common address space again.

Not at same time

This idea creates a ton of tasks to support many firmwares, update it and interact between.

Actually, this idea leads to ONE firmware in "system" MCU, as second MCU has no firmware at all, user applications uploads to ram at every launch. But yes, you need protocol to interact with main CPU and it's peripherals

[zhovner]

Choosing every component is not so easy. We need to produce tens of thousands devices and changing every component is affects a bunch of negotiations with suppliers. For example for current STM32 we need to figure out the export permits and so on. We plan to lock BOM in a few days, so adding the second MCU is the real threat to fail delivery at the time.