Re: [PATCH] hwpoison: Fix race with changing page during offlining v2

[Andrew Morton <akpm@linux-foundation.org>]

On Mon, 30 Jun 2014 17:32:16 -0700 Andi Kleen <andi@firstfloor.org> wrote:

> From: Andi Kleen <ak@linux.intel.com>
> 
> When a hwpoison page is locked it could change state
> due to parallel modifications.  Check after the lock
> if the page is still the same compound page.
> 
> ...
>
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -1168,6 +1168,16 @@ int memory_failure(unsigned long pfn, int trapno, int flags)
>  	lock_page(hpage);
>  
>  	/*
> +	 * The page could have changed compound pages during the locking.
> +	 * If this happens just bail out.
> +	 */
> +	if (compound_head(p) != hpage) {

How can a 4k page change compound pages?  The original compound page
was torn down and then this 4k page became part of a differently-sized
compound page?

> +		action_result(pfn, "different compound page after locking", IGNORED);
> +		res = -EBUSY;
> +		goto out;
> +	}
> +
> +	/*

I don't get it.  We just go and fail the poisoning attempt?  Shouldn't
we go back, grab the new hpage and try again?