Drivers: QURT I2C do not use a static mutex for all I2C drivers

[hendjoshsr71]

Solved Problem

This makes the QURT I2C driver use a mutex per I2C bus.

Previously, if one I2C bus crashes or is hung it crashes all the other buses. For example the baro bus can get stalled if another bus gets stalled.

Note: The timeout to recognize an I2C bus is being held for too long is still on the order of 6 seconds at times. Far too long for safety of others devices on the same bus.

Changelog Entry

For release notes: QURT I2C do not use a static mutex for all I2C drivers

Test coverage

Tested by forcefully hanging the I2C bus before and after the change. Before this change the baro and voxl-pm busses on my system would crash when i forcefully hang a different bus.

[dagar]

This isn't necessarily per bus though right? Each driver has an instance of this I2C, so it sounds like you need an actual per bus mutex somewhere. We don't deal with this on NuttX because per bus locking is handled lower level by the OS driver.

[katzfey]

Yes, @dagar you are absolutely correct. Needs to be per bus, not per driver instance.

[dagar]

FYI if everything I2C is a typical PX4 driver (I2CSPIDriverConfig) then you should already be fine as those are all going to run in a work queue (thread per bus).

[katzfey]

Yes, it's currently all standard I2C driver stuff so good to know that it's only ever a single thread per bus. So don't really need a mutex at all then but I think I'll be paranoid and just change it to a per bus mutex in case some custom code comes along that needs it.

[katzfey]

Made a new PR. #23531

[hendjoshsr71]

Thanks Eric!