Re: [PATCH 14/33] cpuset: Update HK_TYPE_DOMAIN cpumask from cpuset

[Waiman Long <llong@redhat.com>]

On 12/25/25 9:24 PM, Waiman Long wrote:
> On 12/24/25 8:45 AM, Frederic Weisbecker wrote:
>> Until now, HK_TYPE_DOMAIN used to only include boot defined isolated
>> CPUs passed through isolcpus= boot option. Users interested in also
>> knowing the runtime defined isolated CPUs through cpuset must use
>> different APIs: cpuset_cpu_is_isolated(), cpu_is_isolated(), etc...
>>
>> There are many drawbacks to that approach:
>>
>> 1) Most interested subsystems want to know about all isolated CPUs, not
>>    just those defined on boot time.
>>
>> 2) cpuset_cpu_is_isolated() / cpu_is_isolated() are not synchronized 
>> with
>>    concurrent cpuset changes.
>>
>> 3) Further cpuset modifications are not propagated to subsystems
>>
>> Solve 1) and 2) and centralize all isolated CPUs within the
>> HK_TYPE_DOMAIN housekeeping cpumask.
>>
>> Subsystems can rely on RCU to synchronize against concurrent changes.
>>
>> The propagation mentioned in 3) will be handled in further patches.
>>
>> Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
>> ---
>>   include/linux/sched/isolation.h |  7 +++
>>   kernel/cgroup/cpuset.c          |  3 ++
>>   kernel/sched/isolation.c        | 76 ++++++++++++++++++++++++++++++---
>>   kernel/sched/sched.h            |  1 +
>>   4 files changed, 81 insertions(+), 6 deletions(-)
>>
>> diff --git a/include/linux/sched/isolation.h 
>> b/include/linux/sched/isolation.h
>> index 109a2149e21a..6842a1ba4d13 100644
>> --- a/include/linux/sched/isolation.h
>> +++ b/include/linux/sched/isolation.h
>> @@ -9,6 +9,11 @@
>>   enum hk_type {
>>       /* Revert of boot-time isolcpus= argument */
>>       HK_TYPE_DOMAIN_BOOT,
>> +    /*
>> +     * Same as HK_TYPE_DOMAIN_BOOT but also includes the
>> +     * revert of cpuset isolated partitions. As such it
>> +     * is always a subset of HK_TYPE_DOMAIN_BOOT.
>> +     */
>>       HK_TYPE_DOMAIN,
>>       /* Revert of boot-time isolcpus=managed_irq argument */
>>       HK_TYPE_MANAGED_IRQ,
>> @@ -35,6 +40,7 @@ extern const struct cpumask 
>> *housekeeping_cpumask(enum hk_type type);
>>   extern bool housekeeping_enabled(enum hk_type type);
>>   extern void housekeeping_affine(struct task_struct *t, enum hk_type 
>> type);
>>   extern bool housekeeping_test_cpu(int cpu, enum hk_type type);
>> +extern int housekeeping_update(struct cpumask *isol_mask, enum 
>> hk_type type);
>>   extern void __init housekeeping_init(void);
>>     #else
>> @@ -62,6 +68,7 @@ static inline bool housekeeping_test_cpu(int cpu, 
>> enum hk_type type)
>>       return true;
>>   }
>>   +static inline int housekeeping_update(struct cpumask *isol_mask, 
>> enum hk_type type) { return 0; }
>>   static inline void housekeeping_init(void) { }
>>   #endif /* CONFIG_CPU_ISOLATION */
>>   diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
>> index 5e2e3514c22e..e13e32491ebf 100644
>> --- a/kernel/cgroup/cpuset.c
>> +++ b/kernel/cgroup/cpuset.c
>> @@ -1490,6 +1490,9 @@ static void update_isolation_cpumasks(void)
>>       ret = tmigr_isolated_exclude_cpumask(isolated_cpus);
>>       WARN_ON_ONCE(ret < 0);
>>   +    ret = housekeeping_update(isolated_cpus, HK_TYPE_DOMAIN);
>> +    WARN_ON_ONCE(ret < 0);
>> +
>>       isolated_cpus_updating = false;
>>   }
>>   diff --git a/kernel/sched/isolation.c b/kernel/sched/isolation.c
>> index 83be49ec2b06..a124f1119f2e 100644
>> --- a/kernel/sched/isolation.c
>> +++ b/kernel/sched/isolation.c
>> @@ -29,18 +29,48 @@ static struct housekeeping housekeeping;
>>     bool housekeeping_enabled(enum hk_type type)
>>   {
>> -    return !!(housekeeping.flags & BIT(type));
>> +    return !!(READ_ONCE(housekeeping.flags) & BIT(type));
>>   }
>>   EXPORT_SYMBOL_GPL(housekeeping_enabled);
>>   +static bool housekeeping_dereference_check(enum hk_type type)
>> +{
>> +    if (IS_ENABLED(CONFIG_LOCKDEP) && type == HK_TYPE_DOMAIN) {
>
> To be more correct, we should use IS_ENABLED(CONFIG_PROVE_LOCKING) as 
> this is the real kconfig that enables most of the lockdep checking. 
> PROVE_LOCKING selects LOCKDEP but not vice versa. So for some weird 
> configs that set LOCKDEP but not PROVE_LOCKING, it can cause 
> compilation problem. 

I think I get confused too. The various lockdep* helpers should be 
defined when CONFIG_LOCKDEP is enabled even if they may not do anything 
useful. So using IS_ENABLED(CONFIG_LOCKDEP) should be fine. Sorry for 
the noise.

Reviewed-by: Waiman Long <longman@redhat.com>