Re: [PATCH] call stop_machine_text_poke() on all cpus

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

* Jason Baron (jbaron@redhat.com) wrote:
> Currently, text_poke_smp() passes a NULL as the third argument to
> __stop_machine(), which will only run stop_machine_text_poke()
> on 1 cpu. Change NULL -> cpu_online_mask, as stop_machine_text_poke()
> is intended to be run on all cpus.
> 
> I actually didn't notice any problems with stop_machine_text_poke()
> only being called on 1 cpu, but found this via code inspection.
> 
> Signed-off-by: Jason Baron <jbaron@redhat.com>

Good catch!

The side-effect of this is that the other CPUs are not executing the
memory barrier and not flushing the icache, as we would normally expect.

Thanks,

Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

> ---
>  arch/x86/kernel/alternative.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
> index 9f39a1c..3c3f26f 100644
> --- a/arch/x86/kernel/alternative.c
> +++ b/arch/x86/kernel/alternative.c
> @@ -705,6 +705,6 @@ void *__kprobes text_poke_smp(void *addr, const void *opcode, size_t len)
>  	atomic_set(&stop_machine_first, 1);
>  	wrote_text = 0;
>  	/* Use __stop_machine() because the caller already got online_cpus. */
> -	__stop_machine(stop_machine_text_poke, (void *)&tpp, NULL);
> +	__stop_machine(stop_machine_text_poke, (void *)&tpp, cpu_online_mask);
>  	return addr;
>  }
> -- 
> 1.7.1
> 

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com