[PATCH 0/5] cpumask: Use cpumask_intersects* APIs where applicable

[PATCH 0/5] cpumask: Use cpumask_intersects* APIs where applicable

[Shrikanth Hegde]

It was observed while working on cpumask_intersects_and, check like
below could use it instead.

if (cpumask_first_and_and() < nr_cpu_ids)

While there, did check the code for any such similar patterns of
cpumask_any_and, cpumask_first_and and explicit loops where
cpu found by these APIs are not being used. All those callsites
instead can simply use cpumask_intersects.

Note that few more places can use similar change. But, I have left
them as is. 
drivers/net/ethernet/sfc/efx_channels.c+370
drivers/net/ethernet/sfc/siena/efx_channels.c+371
drivers/platform/x86/intel/uncore-frequency/uncore-frequency.c+153

Also, these patches can combined into three patches if that is
easier for review/merge. I have kept it separate since they belong
to different subsystems.

Shrikanth Hegde (5):
  sched/rt: cpupri_find: Use cpumask_intersects where applicable
  cpu/hotplug: Use cpumask_intersects where applicable
  perf: Use cpumask_intersects where applicable
  cpuidle: Use cpumask_intersects_and where applicable
  arch_topology: Use cpumask_intersects where applicable

 drivers/base/arch_topology.c | 9 +--------
 drivers/cpuidle/coupled.c    | 4 ++--
 kernel/cpu.c                 | 3 +--
 kernel/events/core.c         | 7 ++-----
 kernel/sched/cpupri.c        | 2 +-
 5 files changed, 7 insertions(+), 18 deletions(-)

-- 
2.52.0

[PATCH 1/5] sched/rt: cpupri_find: Use cpumask_intersects where applicable

[Shrikanth Hegde]

cpumask_any_and returns, usually the first cpu in the intersection
of the two cpumasks. However, if the returned value is not used or
only to check it against nr_cpu_ids, that implies whether the two
cpumask have a bit in common. I.e cpumask_intersects.

cpumask_any_and uses ffs on (addr1[idx] & addr2[idx]) whereas
cpumask_intersects checks only (addr1[idx] & addr2[idx]).

So replace such occurrences.

Signed-off-by: Shrikanth Hegde <sshegde@linux.ibm.com>
---
 kernel/sched/cpupri.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/sched/cpupri.c b/kernel/sched/cpupri.c
index 8f2237e8b484..44d67fbc4ab2 100644
--- a/kernel/sched/cpupri.c
+++ b/kernel/sched/cpupri.c
@@ -97,7 +97,7 @@ static inline int __cpupri_find(struct cpupri *cp, struct task_struct *p,
 	if (skip)
 		return 0;
 
-	if (cpumask_any_and(&p->cpus_mask, vec->mask) >= nr_cpu_ids)
+	if (!cpumask_intersects(&p->cpus_mask, vec->mask))
 		return 0;
 
 	if (lowest_mask) {
-- 
2.52.0

[PATCH 2/5] cpu/hotplug: Use cpumask_intersects where applicable

[Shrikanth Hegde]

cpumask_any_and returns, usually the first cpu in the intersection
of the two cpumasks. However, if the returned value is not used or
only to check it against nr_cpu_ids, that implies whether the two
cpumask have a bit in common. i.e cpumask_intersects.

cpumask_any_and uses ffs on (addr1[idx] & addr2[idx]) whereas
cpumask_intersects checks only (addr1[idx] & addr2[idx]).

So replace such occurrences.

Signed-off-by: Shrikanth Hegde <sshegde@linux.ibm.com>
---
 kernel/cpu.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/kernel/cpu.c b/kernel/cpu.c
index 376d297a6292..e39245d9bffe 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -1418,8 +1418,7 @@ static int __ref _cpu_down(unsigned int cpu, int tasks_frozen,
 	 * Keep at least one housekeeping cpu onlined to avoid generating
 	 * an empty sched_domain span.
 	 */
-	if (cpumask_any_and(cpu_online_mask,
-			    housekeeping_cpumask(HK_TYPE_DOMAIN)) >= nr_cpu_ids) {
+	if (!cpumask_intersects(cpu_online_mask, housekeeping_cpumask(HK_TYPE_DOMAIN))) {
 		ret = -EBUSY;
 		goto out;
 	}
-- 
2.52.0

[PATCH 3/5] perf: Use cpumask_intersects where applicable

[Shrikanth Hegde]

cpumask_any_and returns, usually the first cpu in the intersection
of the two cpumasks. However, if the returned value is not used or
only to check it against nr_cpu_ids, that implies whether the two
cpumask have a bit in common. i.e cpumask_intersects.

cpumask_any_and uses ffs on (addr1[idx] & addr2[idx]) whereas
cpumask_intersects checks only (addr1[idx] & addr2[idx]).

So replace such occurrences.

Signed-off-by: Shrikanth Hegde <sshegde@linux.ibm.com>
---
 kernel/events/core.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index a6c8e38a3110..f9fa1b47ddf4 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -13159,7 +13159,6 @@ static int perf_try_init_event(struct pmu *pmu, struct perf_event *event)
 	if (pmu->scope != PERF_PMU_SCOPE_NONE && event->cpu >= 0) {
 		const struct cpumask *cpumask;
 		struct cpumask *pmu_cpumask;
-		int cpu;
 
 		cpumask = perf_scope_cpu_topology_cpumask(pmu->scope, event->cpu);
 		pmu_cpumask = perf_scope_cpumask(pmu->scope);
@@ -13168,8 +13167,7 @@ static int perf_try_init_event(struct pmu *pmu, struct perf_event *event)
 		if (!pmu_cpumask || !cpumask)
 			goto err_destroy;
 
-		cpu = cpumask_any_and(pmu_cpumask, cpumask);
-		if (cpu >= nr_cpu_ids)
+		if (!cpumask_intersects(pmu_cpumask, cpumask))
 			goto err_destroy;
 
 		event->event_caps |= PERF_EV_CAP_READ_SCOPE;
@@ -15312,8 +15310,7 @@ static void perf_event_setup_cpumask(unsigned int cpu)
 		if (WARN_ON_ONCE(!pmu_cpumask || !cpumask))
 			continue;
 
-		if (!cpumask_empty(cpumask) &&
-		    cpumask_any_and(pmu_cpumask, cpumask) >= nr_cpu_ids)
+		if (!cpumask_empty(cpumask) && !cpumask_intersects(pmu_cpumask, cpumask))
 			cpumask_set_cpu(cpu, pmu_cpumask);
 	}
 end:
-- 
2.52.0

[PATCH 4/5] cpuidle: Use cpumask_intersects_and where applicable

[Shrikanth Hegde]

cpumask_first_and_and returns, usually the first CPU in the intersection
of the three cpumasks. However, if the returned CPU is not used or
only to check it against nr_cpu_ids, that implies to check whether the
three cpumasks have a bit in common. I.e use cpumask_intersects_and.

cpumask_first_and_and uses ffs on (addr1[idx] & addr2[idx] & addr3[idx])
whereas cpumask_intersects_and checks only
(addr1[idx] & addr2[idx] & addr3[idx]).

So replace such an occurrence.

Signed-off-by: Shrikanth Hegde <sshegde@linux.ibm.com>
---
 drivers/cpuidle/coupled.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/cpuidle/coupled.c b/drivers/cpuidle/coupled.c
index fb91a9e0e3c2..40db5c86a466 100644
--- a/drivers/cpuidle/coupled.c
+++ b/drivers/cpuidle/coupled.c
@@ -439,8 +439,8 @@ static int cpuidle_coupled_clear_pokes(int cpu)
 
 static bool cpuidle_coupled_any_pokes_pending(struct cpuidle_coupled *coupled)
 {
-	return cpumask_first_and_and(cpu_online_mask, &coupled->coupled_cpus,
-				     &cpuidle_coupled_poke_pending) < nr_cpu_ids;
+	return cpumask_intersects_and(cpu_online_mask, &coupled->coupled_cpus,
+				      &cpuidle_coupled_poke_pending);
 }
 
 /**
-- 
2.52.0

[PATCH 5/5] arch_topology: Use cpumask_intersects where applicable

[Shrikanth Hegde]

Looping over one cpumask and checking if any bit is set in another
cpumask is same as evaluating if two cpumasks intersects or not.

So use cpumask_intersects instead. It is probably faster too.

Signed-off-by: Shrikanth Hegde <sshegde@linux.ibm.com>
---
 drivers/base/arch_topology.c | 9 +--------
 1 file changed, 1 insertion(+), 8 deletions(-)

diff --git a/drivers/base/arch_topology.c b/drivers/base/arch_topology.c
index 8c5e47c28d9a..aba86e9c2516 100644
--- a/drivers/base/arch_topology.c
+++ b/drivers/base/arch_topology.c
@@ -34,14 +34,7 @@ EXPORT_PER_CPU_SYMBOL_GPL(capacity_freq_ref);
 
 static bool supports_scale_freq_counters(const struct cpumask *cpus)
 {
-	int i;
-
-	for_each_cpu(i, cpus) {
-		if (cpumask_test_cpu(i, &scale_freq_counters_mask))
-			return true;
-	}
-
-	return false;
+	return cpumask_intersects(cpus, &scale_freq_counters_mask);
 }
 
 bool topology_scale_freq_invariant(void)
-- 
2.52.0

Re: [PATCH 5/5] arch_topology: Use cpumask_intersects where applicable

[Sudeep Holla]

On Wed, Nov 11, 2026 at 01:29:45PM +0530, Shrikanth Hegde wrote:
> Looping over one cpumask and checking if any bit is set in another
> cpumask is same as evaluating if two cpumasks intersects or not.
> 
> So use cpumask_intersects instead. It is probably faster too.
> 

Reviewed-by: Sudeep Holla <sudeep.holla@kernel.org>

-- 
Regards,
Sudeep

Re: [PATCH 0/5] cpumask: Use cpumask_intersects* APIs where applicable

[Peter Zijlstra]

On Wed, Nov 11, 2026 at 01:29:40PM +0530, Shrikanth Hegde wrote:

> Shrikanth Hegde (5):

>   cpuidle: Use cpumask_intersects_and where applicable
>   arch_topology: Use cpumask_intersects where applicable
 
>  drivers/base/arch_topology.c | 9 +--------
>  drivers/cpuidle/coupled.c    | 4 ++--

That's not actually code I maintain. That's rjw stuff.

Re: [PATCH 0/5] cpumask: Use cpumask_intersects* APIs where applicable

[Shrikanth Hegde]

On 11/11/26 2:46 PM, Peter Zijlstra wrote:
> On Wed, Nov 11, 2026 at 01:29:40PM +0530, Shrikanth Hegde wrote:
> 
>> Shrikanth Hegde (5):
> 
>>    cpuidle: Use cpumask_intersects_and where applicable
>>    arch_topology: Use cpumask_intersects where applicable
>   
>>   drivers/base/arch_topology.c | 9 +--------
>>   drivers/cpuidle/coupled.c    | 4 ++--
> 
> That's not actually code I maintain. That's rjw stuff.

Yes. Sorry, I should have cc'ed him.

Should I split this series and send it to each subsystem instead?

Re: [PATCH 0/5] cpumask: Use cpumask_intersects* APIs where applicable

[Peter Zijlstra]

On Wed, Nov 11, 2026 at 02:52:58PM +0530, Shrikanth Hegde wrote:
> 
> 
> On 11/11/26 2:46 PM, Peter Zijlstra wrote:
> > On Wed, Nov 11, 2026 at 01:29:40PM +0530, Shrikanth Hegde wrote:
> > 
> > > Shrikanth Hegde (5):
> > 
> > >    cpuidle: Use cpumask_intersects_and where applicable
> > >    arch_topology: Use cpumask_intersects where applicable
> > >   drivers/base/arch_topology.c | 9 +--------
> > >   drivers/cpuidle/coupled.c    | 4 ++--
> > 
> > That's not actually code I maintain. That's rjw stuff.
> 
> Yes. Sorry, I should have cc'ed him.
> 
> Should I split this series and send it to each subsystem instead?

I've taken the first 3 patches since that's all tip stuff. Feel free to
resend the last two with appropriate Ccs. Its all independent patches
anyway.