Page 2 of 2

Re: [RFC] Functional Safety in HAL

Posted: Fri Dec 20, 2019 2:14 pm
by FXCoder
Hi Andrew,
Didn't try RCC reset of the MCU I2C (although I still have some boards and could give that a go at some stage).
The root cause is the Mi___chip power measurement device which didn't meet its low temperature operating spec.
We just don't use that part any more.
--
Bob

Re: [RFC] Functional Safety in HAL

Posted: Sat Dec 21, 2019 12:40 am
by tridge
Giovanni wrote:Tridge, have you tried the SW I2C implementation? that would be the safest one. You can make it go fast enough without using much CPU.

no, and I really don't want to. We run these MCUs with a large number of sensors and I really don't want the overhead of a sw I2C. Some of our boards have 4 I2C buses active, plus 3 SPI buses, dual CAN bus and 5 UARTs all of which may have lots of activity. All of them using DMA when available. It really is not practical to use sw I2C, especially as we want the hardware oversampling that we get with the hw I2C device to improve resilience to noise.
With the interrupt storm protection (ISR quota) in place and the patches I've posted I am very confident the I2C driver is now safe. The driver in ChibiOS trunk is not safe.