Re: [PATCH v2 2/2] cpufreq: CPPC: Dont read counters for idle CPUs

[Beata Michalska <beata.michalska@arm.com>]

On Wed, Aug 06, 2025 at 05:01:48PM -0700, Prashant Malani wrote:
> Hi Beata,
> 
> On Wed, 6 Aug 2025 at 00:22, Beata Michalska <beata.michalska@arm.com> wrote:
> > Would you mind giving it a go and see whether that improves things on your end ?
> > Note that this is a quick and semi-dirty hack though.
> >
> 
> Sure.
> The provided patch doesn't appear to work as expected.
> With all cores loaded (stress_ng --cpu N), it's returning the same
> counter values
> across samples. Here are readings from multiple CPUs:
> 
> t0: del:18446603338626579088, ref:192
> t1: del:18446603338626579088, ref:192
> ref_perf:10
> delivered_perf:0
> 
> 
> t0: del:18446603338627594896, ref:192
> t1: del:18446603338627594896, ref:192
> ref_perf:10
> delivered_perf:0
> 
> t0: del:18446603338627627664, ref:192
> t1: del:18446603338627627664, ref:192
> ref_perf:10
> delivered_perf:0
> 
> I verified separately that the "burst_read" path is being used by the platform
> I am testing on.
> 
> BR,
>
> -Prashant
Right .... that's what happens when you are (I am) making last minute clean up.
That should fix it. Would you mind giving it another go ? Would appreciate it.

---
BR
Beata

diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c
index 65adb78a9a87..2a51e93fcd6c 100644
--- a/arch/arm64/kernel/topology.c
+++ b/arch/arm64/kernel/topology.c
@@ -543,7 +543,7 @@ void counters_burst_read_on_cpu(void *arg)
 
 static inline bool cpc_reg_supported(struct cpc_reg *reg)
 {
-       return !((u64)reg->address != 0x0 || (u64)reg->address != 0x1);
+       return !((u64)reg->address != 0x0 && (u64)reg->address != 0x1);
 }
 
 int cpc_burst_read_ffh(int cpu, struct cpc_reg_sample *samples, size_t count)