Re: [Qemu-devel] [RFC PATCH 3/3] spapr: generate a CAS reset for the XIVE exploitation mode

[David Gibson <david@gibson.dropbear.id.au>]

[-- Attachment #1: Type: text/plain, Size: 3319 bytes --]

On Sun, Sep 10, 2017 at 09:30:18PM +0200, Cédric Le Goater wrote:
> On 09/10/2017 05:19 AM, David Gibson wrote:
> > On Fri, Sep 08, 2017 at 04:33:44PM +0200, Cédric Le Goater wrote:
> >> When the platform and the guest agree on using the XIVE exploitation
> >> mode for interrupts, the "interrupt-controller" node needs to reflect
> >> the change and the device tree needs an update.
> >>
> >> Reseting the guest after the CAS negotiation makes this change
> >> possible, as the device tree is built at reset time. We use the
> >> 'ov5_cas' field to check which interrupt model was negotiated before
> >> reset and populate the tree accordingly.
> >>
> >> Signed-off-by: Cédric Le Goater <clg@kaod.org>
> > 
> > Looks reasonable, though I'm not going to apply it until we actually
> > have something to go..
> 
> Yes. I agree. A simplified XIVE patchset for sPAPR should be ready 
> soon for comments. Please see below for one question.
> 
> >> ---
> >>  hw/ppc/spapr.c       | 6 +++++-
> >>  hw/ppc/spapr_hcall.c | 6 ++++++
> >>  2 files changed, 11 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> >> index 3e3ff1fbc988..be467ab61ad0 100644
> >> --- a/hw/ppc/spapr.c
> >> +++ b/hw/ppc/spapr.c
> >> @@ -1073,7 +1073,11 @@ static void *spapr_build_fdt(sPAPRMachineState *spapr,
> >>      _FDT(fdt_setprop_cell(fdt, 0, "#size-cells", 2));
> >>  
> >>      /* /interrupt controller */
> >> -    spapr_dt_xics(xics_max_server_number(), fdt, PHANDLE_XICP);
> >> +    if (!spapr_ovec_test(spapr->ov5_cas, OV5_XIVE_EXPLOIT)) {
> >> +        spapr_dt_xics(xics_max_server_number(), fdt, PHANDLE_XICP);
> >> +    } else {
> >> +        /* populate device tree for XIVE */ ;
> > 
> > .. here.
> 
> How should we handle the PHANDLE_XICP for xive ? as it is done
> for xics ? 
> 
> The DT node is now :	"interrupt-controller@6030203180000"

Is there just one node that needs to be referenced?  If so, then yes,
the XICS approach of just using a well known phandle for it should be
fine.


> 
> >> +    }
> >>  
> >>      ret = spapr_populate_memory(spapr, fdt);
> >>      if (ret < 0) {
> >> diff --git a/hw/ppc/spapr_hcall.c b/hw/ppc/spapr_hcall.c
> >> index 92f1e21358b8..ba00b8d3fdd6 100644
> >> --- a/hw/ppc/spapr_hcall.c
> >> +++ b/hw/ppc/spapr_hcall.c
> >> @@ -1613,6 +1613,12 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
> >>              (spapr_h_cas_compose_response(spapr, args[1], args[2],
> >>                                            ov5_updates) != 0);
> >>      }
> >> +
> >> +    /* We need to rebuild the device tree for XIVE, generate a reset */
> >> +    if (!spapr->cas_reboot) {
> >> +        spapr->cas_reboot = spapr_ovec_test(ov5_updates, OV5_XIVE_EXPLOIT);
> >> +    }
> > 
> > At some point we may well want to make XIVE exploitation the default
> > for POWER9 guests, but that's a detail easily adjusted later.
> 
> ah yes. we can do that.
> 
> Thanks,
> 
> C. 
>  
> >>      spapr_ovec_cleanup(ov5_updates);
> >>  
> >>      if (spapr->cas_reboot) {
> > 
> 

-- 
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

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]