On Wed, Dec 11, 2024 at 04:40:23PM +0100, Frederic Weisbecker wrote:
> When a kthread or any other task has an affinity mask that is fully
> offline or unallowed, the scheduler reaffines the task to all possible
> CPUs as a last resort.
>
> This default decision doesn't mix up very well with nohz_full CPUs that
> are part of the possible cpumask but don't want to be disturbed by
> unbound kthreads or even detached pinned user tasks.
>
> Make the fallback affinity setting aware of nohz_full.
>
> Suggested-by: Michal Hocko <mhocko@xxxxxxxx>
> Signed-off-by: Frederic Weisbecker <frederic@xxxxxxxxxx>
> ---
> arch/arm64/include/asm/cpufeature.h | 1 +
> arch/arm64/include/asm/mmu_context.h | 2 ++
> arch/arm64/kernel/cpufeature.c | 11 +++++++++++
> include/linux/mmu_context.h | 1 +
> kernel/sched/core.c | 2 +-
> 5 files changed, 16 insertions(+), 1 deletion(-)
>
> diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h
> index 8b4e5a3cd24c..cac5efc836c0 100644
> --- a/arch/arm64/include/asm/cpufeature.h
> +++ b/arch/arm64/include/asm/cpufeature.h
> @@ -671,6 +671,7 @@ static inline bool supports_clearbhb(int scope)
> }
>
> const struct cpumask *system_32bit_el0_cpumask(void);
> +const struct cpumask *fallback_32bit_el0_cpumask(void);
> DECLARE_STATIC_KEY_FALSE(arm64_mismatched_32bit_el0);
>
> static inline bool system_supports_32bit_el0(void)
> diff --git a/arch/arm64/include/asm/mmu_context.h b/arch/arm64/include/asm/mmu_context.h
> index 48b3d9553b67..7883abd6b29a 100644
> --- a/arch/arm64/include/asm/mmu_context.h
> +++ b/arch/arm64/include/asm/mmu_context.h
> @@ -283,6 +283,8 @@ task_cpu_possible_mask(struct task_struct *p)
> }
> #define task_cpu_possible_mask task_cpu_possible_mask
>
> +const struct cpumask *task_cpu_fallback_mask(struct task_struct *p);
> +
> void verify_cpu_asid_bits(void);
> void post_ttbr_update_workaround(void);
>
> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> index 7ce1b8ab417f..2b7aa32bf436 100644
> --- a/arch/arm64/kernel/cpufeature.c
> +++ b/arch/arm64/kernel/cpufeature.c
> @@ -1642,6 +1642,17 @@ const struct cpumask *system_32bit_el0_cpumask(void)
> return cpu_possible_mask;
> }
>
> +const struct cpumask *task_cpu_fallback_mask(struct task_struct *p)
> +{
> + if (!static_branch_unlikely(&arm64_mismatched_32bit_el0))
> + return housekeeping_cpumask(HK_TYPE_TICK);
> +
> + if (!is_compat_thread(task_thread_info(p)))
> + return housekeeping_cpumask(HK_TYPE_TICK);
> +
> + return system_32bit_el0_cpumask();
> +}
I think this is correct, but damn what we really want to ask for is the
intersection of task_cpu_possible_mask(p) and
housekeeping_cpumask(HK_TYPE_TICK). It's a shame to duplicate the logic
in task_cpu_possible_mask() here because we don't want to allocate a
temporary mask.
Maybe we could have a helper to consolidate things a little?
static inline const struct cpumask *
__task_cpu_possible_mask(struct task_struct *p, const struct cpumask *mask)
{
if (!static_branch_unlikely(&arm64_mismatched_32bit_el0))
return mask;
if (!is_compat_thread(task_thread_info(p)))
return mask;
return system_32bit_el0_cpumask();
}
Then we could call that from both task_cpu_possible_mask() and
task_cpu_fallback_mask(), but passing 'cpu_possible_mask' and
housekeeping_cpumask(HK_TYPE_TICK) for the 'mask' argument respectively?
Will
