Re: [PATCH 5/9] x86/mm: Switch to kvfree_rcu() API

[Steven Rostedt]

On Wed, 24 Nov 2021 12:03:04 +0100
"Uladzislau Rezki (Sony)" <urezki@xxxxxxxxx> wrote:

> diff --git a/arch/x86/mm/mmio-mod.c b/arch/x86/mm/mmio-mod.c
> index 933a2ebad471..e75137a06c32 100644
> --- a/arch/x86/mm/mmio-mod.c
> +++ b/arch/x86/mm/mmio-mod.c
> @@ -307,10 +307,8 @@ static void iounmap_trace_core(volatile void __iomem *addr)
>  
>  not_enabled:
>  	spin_unlock_irq(&trace_lock);
> -	if (found_trace) {
> -		synchronize_rcu(); /* unregister_kmmio_probe() requirement */
> -		kfree(found_trace);
> -	}
> +	if (found_trace)
> +		kvfree_rcu(found_trace); /* unregister_kmmio_probe() requirement */
>  }
>  

This is the first I've seen kvfree_rcu() (that I actually noticed/remember,
I'm sure I probably was Cc'd on some patches). And I find the comment
around it very confusing:

Specifically:


 *     kvfree_rcu(ptr);
 *
 * where @ptr is a pointer to kvfree().

The above suggests that you should pass a pointer to the actual function
kvfree to kvfree_rcu(), which is not what I believe is to be done.

  i.e.  kvfree_rcu(kvfree) ???

Perhaps rewrite that to say:

 * where @ptr is the pointer to be freed by kvfree().

?

Other than that, the patch looks fine to me.

Acked-by: Steven Rostedt (VMware) <rostedt@xxxxxxxxxxx>

-- Steve