2011-02-01 21:34:45

by Jeremy Fitzhardinge

[permalink] [raw]
Subject: Re: [PATCH 04/11] xen/mmu: BUG_ON when racing to swap middle leaf.

On 01/31/2011 02:44 PM, Konrad Rzeszutek Wilk wrote:
> The initial bootup code uses set_phys_to_machine quite a lot, and after
> bootup it would be used by the balloon driver. The balloon driver does have
> mutex lock so this should not be necessary - but just in case, add
> a BUG_ON if we do hit this scenario.
>
> [v2: Change from WARN to BUG_ON]
> [v3: Rebased on top of xen->p2m code split]
> Reviewed-by: Ian Campbell <[email protected]>
> Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
> ---
> arch/x86/xen/p2m.c | 3 ++-
> 1 files changed, 2 insertions(+), 1 deletions(-)
>
> diff --git a/arch/x86/xen/p2m.c b/arch/x86/xen/p2m.c
> index 19b0a65..fbbd2ab 100644
> --- a/arch/x86/xen/p2m.c
> +++ b/arch/x86/xen/p2m.c
> @@ -478,7 +478,8 @@ bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn)
>
> /* Swap over from MISSING to IDENTITY if needed. */
> if (p2m_top[topidx][mididx] == p2m_missing) {
> - p2m_top[topidx][mididx] = p2m_identity;
> + BUG_ON(cmpxchg(&p2m_top[topidx][mididx], p2m_missing,
> + p2m_identity) != p2m_missing);

Don't put side-effects in BUG_ONs. Why is it a bug anyway?

J