RE: flexcan missing error state transitions

[Andri Yngvason <andri.yngvason@marel.com>]

Hi Yi,

Quoting ZHU Yi (ST-FIR/ENG1-Zhu) (2017-08-29 08:49:46)
> > From: Andri Yngvason [mailto:andri.yngvason@marel.com]
> > Sent: Monday, August 28, 2017 7:23 PM
...
> > I have users who can change the baudrate but they can not change the device
> > tree. The baudrate can even be changed by service technicians. I.e. it's field
> > configurable. How would I configure the timeout to work well for rates between
> > 125k and 1M?
> Yes, the poll once approach assumes user have access to device-tree, otherwise
> a catch-all delay (based on the slowest bitrate) is required, which is less
> flexible than the periodic timer.
Actually, if it depends on the bitrate, then it would be easiest for everyone to
just set it accordingly in the code based on the bitrate without involving dt.
Anyway, that doesn't really matter, since it seems that we're going with the
"irq + suppress flooding" approach.

> 
> > 
> > Do you think that a periodic timer would load the system more than an IRQ-flood
> > due to a bad error situation?
> > 
> If not the hrtimer than most likely the periodic timer will not load the system
> more than IRQ flood, and we can configure the interval so it can even lower the
> system load than IRQ flood. However, the error irq + suppress flooding approach
> has the following benefits compare to the timer approach:
> 1. report state transitions in time
> 2. no need to handle the contention between irq and timer
> 3. no need to deal with how to specify the delay
OK, good enough.

Regards,
Andri