Re: [RFC PATCH v3 12/15] context_tracking,x86: Defer kernel text patching IPIs

[Peter Zijlstra]

On Tue, Nov 19, 2024 at 04:34:59PM +0100, Valentin Schneider wrote:

> +static void __text_poke_sync(smp_cond_func_t cond_func)
> +{
> +	on_each_cpu_cond(cond_func, do_sync_core, NULL, 1);
> +}
> +
>  void text_poke_sync(void)
>  {
> -	on_each_cpu(do_sync_core, NULL, 1);
> +	__text_poke_sync(NULL);
> +}
> +
> +void text_poke_sync_deferrable(void)
> +{
> +	__text_poke_sync(do_sync_core_defer_cond);
>  }

How about we unwrap some of that like so:

>  /*
> @@ -2257,6 +2273,8 @@ static int tp_vec_nr;
>  static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries)
>  {
>  	unsigned char int3 = INT3_INSN_OPCODE;
> +	bool force_ipi = false;
> +	void (*sync_fn)(void);

	smp_cond_func_t cond = do_sync_core_defer_cond;

>  	unsigned int i;
>  	int do_sync;
>  
> @@ -2291,11 +2309,18 @@ static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries
>  	 * First step: add a int3 trap to the address that will be patched.
>  	 */
>  	for (i = 0; i < nr_entries; i++) {
> +		/*
> +		 * Record that we need to send the IPI if at least one location
> +		 * in the batch requires it.
> +		 */
> +		force_ipi |= tp[i].force_ipi;

		if (tp[i].force_ipi)
			cond = NULL;

>  		tp[i].old = *(u8 *)text_poke_addr(&tp[i]);
>  		text_poke(text_poke_addr(&tp[i]), &int3, INT3_INSN_SIZE);
>  	}
>  
> -	text_poke_sync();
> +	sync_fn = force_ipi ? text_poke_sync : text_poke_sync_deferrable;
> +
> +	sync_fn();

	__text_poke_sync(cond);