Re: [PATCH v2 05/10] staging: axis-fifo: improve IRQ handler

[Dan Carpenter <dan.carpenter@linaro.org>]

On Sun, Mar 01, 2026 at 09:48:10PM +0000, Josh Law wrote:
> From: Josh Law <objecting@objecting.org>
> 
> Return IRQ_NONE when no interrupts were triggered to avoid spurious
> interrupt storms, and only clear the active interrupts instead of
> blindly clearing all interrupts by writing 'intr' to the ISR instead of
> XLLF_INT_CLEAR_ALL.
> 

XLLF_ISR_OFFSET is the IRQs which are triggered and XLLF_IER_OFFSET
are the IRQs that we care about.  The intr variable is the IRQs which
are both, they are triggered and we care about them.  In the original
code it clears out all the IRQs that have triggered regardless of
whether or not we care about them.  You're saying clearing all the
IRQs "blindly" is wrong, but to me it feels like the correct
behavior.  In real life it probably doesn't matter at all since
we probably care about all the IRQs.  And apparently in your testing
it makes no difference either way...

I feel like leaving the flags set feels like it would be more likely
to trigger and IRQ storm.  But we're debating the number of angels
dancing on a pin.

Let's just leave it as-is unless we can show it helps.

regards,
dan carpenter