search

stm32-rs / stm32f1xx-hal

A Rust embedded-hal HAL impl for the STM32F1 family based on japarics stm32f103xx-hal
Apache License 2.0
575 stars 180 forks source link

Understanding code which uses nb #314

Open katyo opened 3 years ago

katyo commented 3 years ago

I have a problem with understanding code which uses nb crate. In particular i2c's write_read looks like incorrect. If I understand right this acts like classic Knuth's coroutine. So it simply should be called again when WouldBlock returned. But Knuth's coroutines has well known problems with state tracking on suspend/resume, and I absolutely does not see any state handling in current code. Say when I call i2c's write_read again after WouldBlock is returned it actually starts from the beginning but do not resumes. This code looks like having dose of black magic or completely misfunctional. Can anyone explain me how it actually works?

TheZoq2 commented 3 years ago

You may be right in that it re-starts the transmission process if called multiple times, at least in some cases. I'm not familiar with Knuth's coroutines, but I see the nb stuff more as "the device is busy-" or "nothing is available right now-" "try again later" rather than "this process is not done yet, continue later".

katyo commented 3 years ago

Current implementation of blocking i2c does not works for me at all. My code interacts with RTC DS3231: It just reads several registers via i2c. But self-written implementation, which operates with registers directly, works fine. I think we should introduce some state management to hold checkpoint and correctly resume I/O after returning WouldBlock.