Received-SPF: pass (google.com: domain of linux-kernel+bounces-61882-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
From: Thomas Gleixner <tglx@linutronix.de>
To: Marcelo Tosatti <mtosatti@redhat.com>, linux-kernel@vger.kernel.org
Cc: Daniel Bristot de Oliveira <bristot@kernel.org>, Juri Lelli
 <juri.lelli@redhat.com>, Valentin Schneider <vschneid@redhat.com>,
 Frederic Weisbecker <frederic@kernel.org>, Leonardo Bras
 <leobras@redhat.com>, Peter Zijlstra <peterz@infradead.org>, Marcelo
 Tosatti <mtosatti@redhat.com>
Subject: Re: [patch 11/12] x86/resctrl: use smp_call_function_single_fail
In-Reply-To: <20240206185710.116221062@redhat.com>
Date: Mon, 12 Feb 2024 16:19:19 +0100
Message-ID: <87zfw5k8w8.ffs@tglx>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain

On Tue, Feb 06 2024 at 15:49, Marcelo Tosatti wrote:
> Convert update_task_closid_rmid from smp_call_function_single
> to smp_call_func_single_fail, which will fail in case
> the target CPU is tagged as block interference CPU.

You fail again to provide a rationale for this change.

What's worse is that you fail to explain why you think that creating
inconistent state is a valid approach.

> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
>
> Index: linux-isolation/arch/x86/kernel/cpu/resctrl/rdtgroup.c
> ===================================================================
> --- linux-isolation.orig/arch/x86/kernel/cpu/resctrl/rdtgroup.c
> +++ linux-isolation/arch/x86/kernel/cpu/resctrl/rdtgroup.c
> @@ -26,6 +26,7 @@
>  #include <linux/slab.h>
>  #include <linux/task_work.h>
>  #include <linux/user_namespace.h>
> +#include <linux/sched/isolation.h>
>  
>  #include <uapi/linux/magic.h>
>  
> @@ -551,12 +552,20 @@ static void _update_task_closid_rmid(voi
>  		resctrl_sched_in(task);
>  }
>  
> -static void update_task_closid_rmid(struct task_struct *t)
> +static int update_task_closid_rmid(struct task_struct *t)
>  {
> -	if (IS_ENABLED(CONFIG_SMP) && task_curr(t))
> -		smp_call_function_single(task_cpu(t), _update_task_closid_rmid, t, 1);
> -	else
> +	int idx, ret = 0;
> +
> +	if (IS_ENABLED(CONFIG_SMP) && task_curr(t)) {
> +		idx = block_interf_srcu_read_lock();
> +		ret = smp_call_function_single_fail(task_cpu(t),
> +						    _update_task_closid_rmid,
> +						    t, 1);
> +		block_interf_srcu_read_unlock(idx);
> +	} else
>  		_update_task_closid_rmid(t);
> +
> +	return ret;

This is invoked _after_ the change has been committed to the in-memory
state so how is failing here correct?

Thanks,

        tglx