Re: [PATCH 11/20] spapr: Fix indexing of XICS irqs

[David Gibson]

On Thu, Sep 26, 2019 at 09:21:41AM +0200, Greg Kurz wrote:
> On Thu, 26 Sep 2019 11:31:48 +1000
> David Gibson <address@hidden> wrote:
> 
> > On Wed, Sep 25, 2019 at 10:17:46PM +0200, Greg Kurz wrote:
> > > On Wed, 25 Sep 2019 16:45:25 +1000
> > > David Gibson <address@hidden> wrote:
> > > 
> > > > spapr global irq numbers are different from the source numbers on the 
> > > > ICS
> > > > when using XICS - they're offset by XICS_IRQ_BASE (0x1000).  But
> > > > spapr_irq_set_irq_xics() was passing through the global irq number to
> > > > the ICS code unmodified.
> > > > 
> > > > We only got away with this because of a counteracting bug - we were
> > > > incorrectly adjusting the qemu_irq we returned for a requested global 
> > > > irq
> > > > number.
> > > > 
> > > > That approach mostly worked but is very confusing, incorrectly relies on
> > > > the way the qemu_irq array is allocated, and undermines the intention of
> > > > having the global array of qemu_irqs for spapr have a consistent meaning
> > > > regardless of irq backend.
> > > > 
> > > > So, fix both set_irq and qemu_irq indexing.  We rename some parameters 
> > > > at
> > > > the same time to make it clear that they are referring to spapr global
> > > > irq numbers.
> > > > 
> > > > Signed-off-by: David Gibson <address@hidden>
> > > > ---
> > > 
> > > Reviewed-by: Greg Kurz <address@hidden>
> > > 
> > > Further cleanup could be to have the XICS backend to only take global
> > > irq numbers and to convert them to ICS source numbers internally. This
> > > would put an end to the confusion between srcno/irq in the frontend
> > > code.
> > 
> > Yeah, maybe.  But the local srcnos do actually make sense from within
> > the perspective of ICS, so I'm not all that keen to do that.
> > 
> 
> Not sure to understand what you mean by "within the perspective of ICS".
> 
> My concern is actually to get rid of ics->offset users in spapr_irq.c.
> 
> eg,
> 
> static void spapr_irq_set_irq_xics(void *opaque, int irq, int val)
> {
>     SpaprMachineState *spapr = opaque;
>     uint32_t srcno = irq - spapr->ics->offset;
> 
>     ics_set_irq(spapr->ics, srcno, val);
> }
> 
> It looks like we should do something like:
> 
> static void spapr_irq_set_irq_xics(void *opaque, int irq, int val)
> {
>     SpaprMachineState *spapr = opaque;
> 
>     ics_set_irq(spapr->ics, irq, val);
> }
> 
> and have ics_set_irq() do:
> 
>     uint32_t srcno = irq - spapr->ics->offset;
> 
> Are you inferring that it is better to keep the irq to srcno conversions
> in spapr_irq.c ?

Ah, I see what you mean.  So, the reason srcno appears here is that
ics_set_irq() is arguably an internal xics function that we shouldn't
be calling directly from here, but do because of.. reasons.

As it happens, I have another patch in the works which will move this
function back into xics_spapr.c and bind it more tightly to the xics
code, so I think that will address your concern.

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

signature.asc
Description: PGP signature