RE: net: phy: ti: implement generic .handle_interrupt() callback

RE: net: phy: ti: implement generic .handle_interrupt() callback

[Sven Schuchmann]

Hello all,

I am working with the DP83TC811R but for now on the 5.10 kernel.
I have seen this patch on the latest dev
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/net/phy/dp83tc811.c?h=v5.11&id=1d1ae3c6ca3ff49843d73852bb2a8153ce16f432

I did not test it yet, but I wonder about one thing:
What happens when there are 2 Flags set e.g.
in INT_STAT1 and one in INT_STAT2?
What I see from the code is that it would
only read (and acknowledge) the first one and
then quit out to trigger_machine.

Any thoughts about this?


Best Regards,

   Sven

Re: net: phy: ti: implement generic .handle_interrupt() callback

[Ioana Ciornei]

On Tue, Feb 23, 2021 at 05:44:57PM +0000, Sven Schuchmann wrote:
> Hello all,
> 
> I am working with the DP83TC811R but for now on the 5.10 kernel.
> I have seen this patch on the latest dev
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/net/phy/dp83tc811.c?h=v5.11&id=1d1ae3c6ca3ff49843d73852bb2a8153ce16f432
> 
> I did not test it yet, but I wonder about one thing:
> What happens when there are 2 Flags set e.g.
> in INT_STAT1 and one in INT_STAT2?
> What I see from the code is that it would
> only read (and acknowledge) the first one and
> then quit out to trigger_machine.
> 
> Any thoughts about this?
> 

Hi Sven,

Indeed, I didn't take into consideration the fact that all the status
registers should be acknowledged since multiple interrupt sources could
be asserted. This could really break a system with an edge-triggered
interrupt.

I'll send out a fixup patch.

Thanks!

Ioana