[1/2] powerpc: remove cpu_online_cores_map function

This function builds the cores online map with on-stack cpumasks which
can cause high stack usage with large NR_CPUS.

It is not used in any performance sensitive paths, so instead just check
for first thread sibling.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 arch/powerpc/include/asm/cputhreads.h     | 33 -----------------------
 arch/powerpc/platforms/powernv/idle.c     | 10 +++----
 arch/powerpc/platforms/powernv/opal-imc.c |  6 ++---
 3 files changed, 8 insertions(+), 41 deletions(-)

> On 05-Nov-2021, at 9:20 AM, Nicholas Piggin <npiggin@gmail.com> wrote:
> 
> This function builds the cores online map with on-stack cpumasks which
> can cause high stack usage with large NR_CPUS.
> 
> It is not used in any performance sensitive paths, so instead just check
> for first thread sibling.
> 
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---

Tested-by: Sachin Sant <sachinp@linux.vnet.ibm.com>

Thanks
- Sachin

> arch/powerpc/include/asm/cputhreads.h     | 33 -----------------------
> arch/powerpc/platforms/powernv/idle.c     | 10 +++----
> arch/powerpc/platforms/powernv/opal-imc.c |  6 ++---
> 3 files changed, 8 insertions(+), 41 deletions(-)
> 
> diff --git a/arch/powerpc/include/asm/cputhreads.h b/arch/powerpc/include/asm/cputhreads.h
> index b167186aaee4..f26c430f3982 100644
> --- a/arch/powerpc/include/asm/cputhreads.h
> +++ b/arch/powerpc/include/asm/cputhreads.h
> @@ -32,44 +32,11 @@ extern cpumask_t threads_core_mask;
> #define threads_core_mask	(*get_cpu_mask(0))
> #endif
> 
> -/* cpu_thread_mask_to_cores - Return a cpumask of one per cores
> - *                            hit by the argument
> - *
> - * @threads:	a cpumask of online threads
> - *
> - * This function returns a cpumask which will have one online cpu's
> - * bit set for each core that has at least one thread set in the argument.
> - *
> - * This can typically be used for things like IPI for tlb invalidations
> - * since those need to be done only once per core/TLB
> - */
> -static inline cpumask_t cpu_thread_mask_to_cores(const struct cpumask *threads)
> -{
> -	cpumask_t	tmp, res;
> -	int		i, cpu;
> -
> -	cpumask_clear(&res);
> -	for (i = 0; i < NR_CPUS; i += threads_per_core) {
> -		cpumask_shift_left(&tmp, &threads_core_mask, i);
> -		if (cpumask_intersects(threads, &tmp)) {
> -			cpu = cpumask_next_and(-1, &tmp, cpu_online_mask);
> -			if (cpu < nr_cpu_ids)
> -				cpumask_set_cpu(cpu, &res);
> -		}
> -	}
> -	return res;
> -}
> -
> static inline int cpu_nr_cores(void)
> {
> 	return nr_cpu_ids >> threads_shift;
> }
> 
> -static inline cpumask_t cpu_online_cores_map(void)
> -{
> -	return cpu_thread_mask_to_cores(cpu_online_mask);
> -}
> -
> #ifdef CONFIG_SMP
> int cpu_core_index_of_thread(int cpu);
> int cpu_first_thread_of_core(int core);
> diff --git a/arch/powerpc/platforms/powernv/idle.c b/arch/powerpc/platforms/powernv/idle.c
> index e3ffdc8e8567..70da314fd2d7 100644
> --- a/arch/powerpc/platforms/powernv/idle.c
> +++ b/arch/powerpc/platforms/powernv/idle.c
> @@ -146,9 +146,13 @@ EXPORT_SYMBOL_GPL(pnv_get_supported_cpuidle_states);
> static void pnv_fastsleep_workaround_apply(void *info)
> 
> {
> +	int cpu = smp_processor_id();
> 	int rc;
> 	int *err = info;
> 
> +	if (cpu_first_thread_sibling(cpu) != cpu)
> +		return;
> +
> 	rc = opal_config_cpu_idle_state(OPAL_CONFIG_IDLE_FASTSLEEP,
> 					OPAL_CONFIG_IDLE_APPLY);
> 	if (rc)
> @@ -175,7 +179,6 @@ static ssize_t store_fastsleep_workaround_applyonce(struct device *dev,
> 		struct device_attribute *attr, const char *buf,
> 		size_t count)
> {
> -	cpumask_t primary_thread_mask;
> 	int err;
> 	u8 val;
> 
> @@ -200,10 +203,7 @@ static ssize_t store_fastsleep_workaround_applyonce(struct device *dev,
> 	power7_fastsleep_workaround_exit = false;
> 
> 	cpus_read_lock();
> -	primary_thread_mask = cpu_online_cores_map();
> -	on_each_cpu_mask(&primary_thread_mask,
> -				pnv_fastsleep_workaround_apply,
> -				&err, 1);
> +	on_each_cpu(pnv_fastsleep_workaround_apply, &err, 1);
> 	cpus_read_unlock();
> 	if (err) {
> 		pr_err("fastsleep_workaround_applyonce change failed while running pnv_fastsleep_workaround_apply");
> diff --git a/arch/powerpc/platforms/powernv/opal-imc.c b/arch/powerpc/platforms/powernv/opal-imc.c
> index 05d3832019b9..3fea5da6d1b3 100644
> --- a/arch/powerpc/platforms/powernv/opal-imc.c
> +++ b/arch/powerpc/platforms/powernv/opal-imc.c
> @@ -200,13 +200,13 @@ static void disable_nest_pmu_counters(void)
> 
> static void disable_core_pmu_counters(void)
> {
> -	cpumask_t cores_map;
> 	int cpu, rc;
> 
> 	cpus_read_lock();
> 	/* Disable the IMC Core functions */
> -	cores_map = cpu_online_cores_map();
> -	for_each_cpu(cpu, &cores_map) {
> +	for_each_online_cpu(cpu) {
> +		if (cpu_first_thread_sibling(cpu) != cpu)
> +			continue;
> 		rc = opal_imc_counters_stop(OPAL_IMC_COUNTERS_CORE,
> 					    get_hard_smp_processor_id(cpu));
> 		if (rc)
> -- 
> 2.23.0
>

On Fri, 5 Nov 2021 13:50:41 +1000, Nicholas Piggin wrote:
> This function builds the cores online map with on-stack cpumasks which
> can cause high stack usage with large NR_CPUS.
> 
> It is not used in any performance sensitive paths, so instead just check
> for first thread sibling.
> 
> 
> [...]

Applied to powerpc/next.

[1/2] powerpc: remove cpu_online_cores_map function
      https://git.kernel.org/powerpc/c/b350111bf7b3f4a780d28c44f18f7c9fcbe6d11b
[2/2] powerpc: select CPUMASK_OFFSTACK if NR_CPUS >= 8192
      https://git.kernel.org/powerpc/c/2eafc4748bc08c5b9b6ee0b5b65ad20b30f7d704

cheers