Re: [PATCH v9 5/5] clocksource/timer-econet-en751221: Support EN751627 without percpu IRQ

[Caleb James DeLisle]

On 15/10/2026 09:21, Daniel Lezcano wrote:
> On 10/14/26 22:36, Caleb James DeLisle wrote:
> > Hello Daniel,
> >
> > Thank you for the review.
> >
> >
> > On 14/10/2026 20:43, Daniel Lezcano wrote:
> > > Hi Caleb,
> > >
> > > with the help of a LLM, I got the following two issues with the changes
> > >
> > > On 9/18/26 17:36, Caleb James DeLisle wrote:
> > > > EN751627 is based on the 1004Kc which uses a different interrupt 
> > > > number
> > > > for each CPU timer. Support both this and the EN751221 which uses a
> > > > single percpu interrupt.
> > > >
> > > > Signed-off-by: Caleb James DeLisle <cjd@xxxxxxxx>
> > > > ---
>
> [ ... ]
>
> > > The binding requires four interrupts and two register blocks for 
> > > EN751627, but the driver sizes its resource arrays from NR_CPUS and 
> > > rejects DTs exceeding those sizes. Consequently, a UP kernel 
> > > (NR_CPUS=1) or an SMP kernel configured for two CPUs rejects a valid 
> > > EN751627 DT before registering any timer.
> >
> > This is a fair point, and I suppose a reasonable thing to do here is 
> > instead set num_blocks = ARRAY_SIZE(econet_timer.membase) so we just 
> > don't use the other blocks. I can send an updated patchset with that.
>
> IMO, econet_timer.irqs[] and econet_timer.membase[] should just be big 
> enough to store all nodes without using NR_CPUS.
>
> Given you have num_irqs and num_blocks, why not just kzalloc them ?


That works for me, I'll re-send it like that then.


Thanks,

Caleb