RE: RFC Proposal to enable MSI support in Linux kernel

[Nguyen, Tom L]

I am working on an updated patch to 2.5.70 to incorporate the comments from you and Jeff Garzik. Regarding the variable renaming of irq, I am open to any suggestions. Please make a recommendation and I will incorporate it into the next update.  Regarding platform_legacy_irq, I have not seen an edge-triggered interrupt failure caused by platform_legacy_irq.

           CPU0       CPU1       
  0:     380044          0    IO-APIC-edge  timer
  1:        590          0    IO-APIC-edge  i8042
  2:          0          0          XT-PIC  cascade
 12:        715          0    IO-APIC-edge  i8042
 14:       7902          0    IO-APIC-edge  ide0
 15:          2          0    IO-APIC-edge  ide1
169:          0          0   IO-APIC-level  uhci-hcd
185:          0          0   IO-APIC-level  uhci-hcd
193:         84          0   IO-APIC-level  aic79xx
201:         30          0   IO-APIC-level  aic79xx
225:         16          0   IO-APIC-level  aic7xxx
233:         16          0   IO-APIC-level  aic7xxx
NMI:          0          0 
LOC:     379962     380218 
ERR:          0
MIS:          0

Thanks,
Tom

-----Original Message-----
From: Zwane Mwaikambo [mailto:[e-mail]]
Sent: Wednesday, May 28, 2003 4:06 PM
To: Jeff Garzik
Cc: Nguyen, Tom L; Nakajima, Jun; [e-mail]; Saxena,
Sunil; Mallick, Asit K; Carbonari, Steven
Subject: Re: RFC Proposal to enable MSI support in Linux kernel

On Wed, 28 May 2003, Jeff Garzik wrote:

> > +#ifdef CONFIG_PCI_USE_VECTOR
> > +	for (i = 0; i < (NR_VECTORS - FIRST_EXTERNAL_VECTOR); i++) {
> > +#else
> >  	for (i = 0; i < NR_IRQS; i++) {
> > +#endif
> 
> This ifdef isn't necessary.  ifdef in a header somewhere.  For example, 
> make NR_IRQS (or some other constant, if changing NR_IRQS definition 
> isn't ok) condition on the CONFIG_xxx options.

Actually we can simply use;

for (i = 0; i < (NR_VECTORS - FIRST_EXTERNAL_VECTOR); i++) {
	vector = ...
	if (i >= NR_IRQS)
		break;
	...
	set_intr_gate(vector, interrupt[i]);
}

> split into two functions, maybe?  (due to the ifdef)  Your call, it's 
> mainly a programmer preference.

I agree.

> A bigger question though:  if platform_legacy_irq() returns zero, will 
> the handler _ever_ be edge-triggered?

Good question, i wouldn't think so, that would collapse that section into 
two lines.

> As I see more and more of these ifdefs, I wonder if they couldn't be 
> hidden inside wrappers?

Yes some of them are a bit easy to botch up later on.. like the following 
excerpt.

> > +#ifdef CONFIG_PCI_USE_VECTOR
> > +			if (!platform_legacy_irq(irq))
> > +				entry->irq = IO_APIC_VECTOR(irq);
> > +			else
> > +#endif
> >  			entry->irq = irq;
> >  			continue;

Thanks,
	Zwane
-- 
function.linuxpower.ca