Re: [Qemu-devel] [PATCH] mips_malta: move i8259 initialization after piix4 initialization

[Avi Kivity]

On 09/14/2011 11:32 PM, Richard Henderson wrote:
> On 09/12/2011 06:07 AM, Avi Kivity wrote:
> >  +static void malta_isa_irq_handler(void *opaque, int n, int level)
> >  +{
> >  +    MaltaISAState *s = opaque;
> >  +
> >  +    if (s->i8259) {
> >  +        qemu_set_irq(s->i8259[n], level);
> >  +    }
> >  +}
>
> Is there any point in the IF?  I realize that there's an ordering
> problem that requires the use of the memory indirection in order
> to be able to provide *some* opaque value at the proper time, but
> AFAICT the ->i8259 value will *always* be non-null at the point
> this function is called.  Am I wrong here?

The pci bridge may toggle the irq line as part of its initialization.

> Otherwise the approach looks correct.

Yeah, but it would have been even more correct to make i8259 not an isa 
device.  Something for later on.

--
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.