search

mcci-catena / Arduino_Core_STM32

STM32 core support for Arduino
2 stars 7 forks source link

System gets stuck in hot interrupt loop from USB ERR interrupt #190

Closed terrillmoore closed 2 years ago

terrillmoore commented 2 years ago

Some units (Catena 4612 specifically) seem to have problems if USB is enabled but no cable is connected. Specifically, they lock up.

Investigation by @mhmayyan in #189 show that there is an issue with the handling of the ERR interrupt. This is always enabled when USB is enabled, but it appears to get stuck sometimes if the USB hardware shows a single-ended 1 (DP and DM both 1). The ISR software dismisses the ERR request, but makes no attempt to mask the interrupt; the interrupt is immediately presented again; so the BSP gets caught in a hot interrupt loop until the condition goes away.

The appearance of this condition appears to be chip-specific and possibly board specific -- we've not seen it on the 4610 or 4630; we see it frequently on the 4612 and 4618, possibly due to layout artifacts, or bad luck on a lot of stm32 chips that are more prone to this due to leakage currents.

However, it's easy to fix. We don't use the ERR interrupt and it can be always masked. Even manually masking on a board particularly prone to this fixed the problem:

Yes, it worked. It did not get stuck again after using the command p ((unsigned *)0x40005c40)[0] = 0x9F00

_Originally posted by @mhmayyan in https://github.com/mcci-catena/Arduino_Core_STM32/issues/189#issuecomment-1072016928_

A full log of the debug process is in the comments for #189.

terrillmoore commented 2 years ago

@mhmayyan if you're able to apply the changes from branch https://github.com/mcci-catena/Arduino_Core_STM32/tree/issue190, it should solve the problem. If you have an OTII, it would be good to compare power between a "good board" and a "bad board" -- I suspect a "good board" goes into USB suspend and drops power. So we may need a more elaborate fix, but at least this will show if we're going in the right direction.

If using a debug BSP is too much of a pain, I'll make a pre-release, so you can download via the IDE.

mhmayyan commented 2 years ago

Thank you @terrillmoore. Yes, this fix solved the problem. I used a DMM (set as an ammeter) to check the amps drainage of four Catena 4612 boards (two good and two bad ones). They all drain almost the same.

terrillmoore commented 2 years ago

Thanks for testing @mhmayyan! I'll get a patch release out shortly.

terrillmoore commented 2 years ago

I've made the release. The fix is in v3.0.5, which is available in the Arduino Boards Manager.