[RFC RESEND v10 02/14] preempt: Introduce __preempt_count_{sub, add}_return()

[Heiko Carstens]

On Tue, May 27, 2025 at 06:21:43PM -0400, Lyude Paul wrote:
> From: Boqun Feng <boqun.feng at gmail.com>
> 
> In order to use preempt_count() to tracking the interrupt disable
> nesting level, __preempt_count_{add,sub}_return() are introduced, as
> their name suggest, these primitives return the new value of the
> preempt_count() after changing it. The following example shows the usage
> of it in local_interrupt_disable():
> 
> 	// increase the HARDIRQ_DISABLE bit
> 	new_count = __preempt_count_add_return(HARDIRQ_DISABLE_OFFSET);
> 
> 	// if it's the first-time increment, then disable the interrupt
> 	// at hardware level.
> 	if (new_count & HARDIRQ_DISABLE_MASK == HARDIRQ_DISABLE_OFFSET) {
> 		local_irq_save(flags);
> 		raw_cpu_write(local_interrupt_disable_state.flags, flags);
> 	}
> 
> Having these primitives will avoid a read of preempt_count() after
> changing preempt_count() on certain architectures.
> 
> Signed-off-by: Boqun Feng <boqun.feng at gmail.com>
> 
> ---
> V10:
> * Add commit message I forgot
> * Rebase against latest pcpu_hot changes
> 
> Signed-off-by: Lyude Paul <lyude at redhat.com>
> ---
>  arch/arm64/include/asm/preempt.h | 18 ++++++++++++++++++
>  arch/s390/include/asm/preempt.h  | 19 +++++++++++++++++++
>  arch/x86/include/asm/preempt.h   | 10 ++++++++++
>  include/asm-generic/preempt.h    | 14 ++++++++++++++
>  4 files changed, 61 insertions(+)

...

> diff --git a/arch/s390/include/asm/preempt.h b/arch/s390/include/asm/preempt.h
> index 6ccd033acfe52..67a6e265e9fff 100644
> --- a/arch/s390/include/asm/preempt.h
> +++ b/arch/s390/include/asm/preempt.h
> @@ -98,6 +98,25 @@ static __always_inline bool should_resched(int preempt_offset)
>  	return unlikely(READ_ONCE(get_lowcore()->preempt_count) == preempt_offset);
>  }
>  
> +static __always_inline int __preempt_count_add_return(int val)
> +{
> +	/*
> +	 * With some obscure config options and CONFIG_PROFILE_ALL_BRANCHES
> +	 * enabled, gcc 12 fails to handle __builtin_constant_p().
> +	 */
> +	if (!IS_ENABLED(CONFIG_PROFILE_ALL_BRANCHES)) {
> +		if (__builtin_constant_p(val) && (val >= -128) && (val <= 127)) {
> +			return val + __atomic_add_const(val, &get_lowcore()->preempt_count);
> +		}
> +	}
> +	return val + __atomic_add(val, &get_lowcore()->preempt_count);
> +}

This is still wrong and needs to be changed to:

static __always_inline int __preempt_count_add_return(int val)
{
	return val + __atomic_add(val, &get_lowcore()->preempt_count);
}