Re: Interrupt questions

[Martin Husemann <martin%duskware.de@localhost>]

On Wed, Jun 25, 2025 at 09:47:44AM -0400, Mouse wrote:
> > Is it legal to call uprintf() inside an interrupt handler?
> 
> I don't know.  It might work enough of the time to be useful even if
> it's not promised; I've used printf() on amd64 inside interrupt
> handlers often enough.  I don't know whether it's promised, but I've
> found it works reliably enough to be a useful debugging tool.  I don't
> recall trying it on any ARM, much less something like your board, but
> it might be worth investigating.
> 
> > If not, is there another way to add traces?
> 
> Sure.  For example, I've had cases where printf introduces enough delay
> to cause problems; I've dealt with them in various ways, such as by
> allocating a ring buffer of logging events and putting my logs there.
> You can print them from (eg) an every-tick callout or you can just
> leave them there for postmortem debugging or some debugging interface
> (such as an ioctl or a custom character special device)....

printf() is OK in interrupt handlers, but it will cause timing issues.

In cases like yours where I just wanted to know if an interrupt handler
ever is called I have also used Debugger() calls - but make sure you
have all watchdog timers disarmed before triggering that.

If you need low latency logging you should set up a kernhist(9) for
your device (like usbhist does for USB debugging).

Martin