On 08/30/2016 08:15 AM, Benjamin Herrenschmidt wrote:
> On Mon, 2016-08-29 at 10:30 -0400, David Gibson wrote:
>>
>> Possibly. In fact, I'm planning to eliminate cpu->cpu_dt_id at some
>> point, in favour of having the machine type construct the id when it
>> actually builds the dt. It's not really a cpu level construct.
>From my understanding, cs->cpu_index is becoming the main CPU identifier.
sPAPRCPUCore assigns it :
cs->cpu_index = cc->core_id + i
which I reused in PnvCPUCore to hold the real HW identifiers.
ppc_get_vcpu_by_dt_id() can also safely use cs->cpu_index I think.
So pnv mostly work without ->cpu_dt_id but there is :
> On PowerNV it is as it's equal to the PIR, the HW interrupt server,
> etc...
xics in the way ... Now that pnv uses real hw core ids, it is
interesting to see how lost it gets without a cpu with index=0 ...
The most obvious issue is the way we look for the ICPState of a cpu :
ICPState *ss = &xics->ss[cs->cpu_index];
how about introducing a helper like (this one I hacked) :
+ICPState *xics_find_icp(XICSState *xics, int cpu_index)
+{
+ int i;
+
+ for (i = 0 ; i < xics->nr_servers; i++) {
+ ICPState *ss = &xics->ss[i];
+ if (ss->cs && ss->cs->cpu_index == cpu_index)
+ return ss;
+ }
+
+ return NULL;
+}
+
That might have been already discussed on the mailing list ?
Thanks,
C.
