Re: [PATCH 08/10] xen/mmu: Bugfix. Fill the top entry page with appropriate middle layer pointers.

[Jeremy Fitzhardinge <jeremy@goop.org>]

On 12/21/2010 01:37 PM, Konrad Rzeszutek Wilk wrote:
> If we swapped over from using an p2m_mid_identical to p2m_mid_missing
> (earlier call to set_phys_to_machine) and then started going through the
> PFNs in descending order to program a new MFN (balloon worker), we would
> end up in this code path. At that point we would set up new page filled with
> pointers to p2m_identity instead of p2m_missing. This had the disastrous
> effect that get_phys_to_machine on that PFN would return an 1-1 mapping
> instead of INVALID_P2M_ENTRY resulting in hitting a BUG check in balloon driver.
>
Are you going to fold this into the appropriate patch later?

> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> ---
>  arch/x86/xen/mmu.c |    5 ++++-
>  1 files changed, 4 insertions(+), 1 deletions(-)
>
> diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
> index 92f4fec..a917439 100644
> --- a/arch/x86/xen/mmu.c
> +++ b/arch/x86/xen/mmu.c
> @@ -480,7 +480,10 @@ static bool alloc_p2m(unsigned long pfn)
>  		if (!mid)
>  			return false;
>  
> -		p2m_mid_init(mid, p2m_identity);
> +		if (mid == p2m_mid_identity)
> +			p2m_mid_init(mid, p2m_identity);
> +		else
> +			p2m_mid_init(mid, p2m_missing);
>  
>  		if (cmpxchg(top_p, mid_orig, mid) != mid_orig)
>  			free_p2m_page(mid);