Re: [PATCH] kvm: reset arch memslot info on memslot creation

[Takuya Yoshikawa <takuya.yoshikawa@gmail.com>]

On Wed, 10 Jul 2013 11:24:39 +0300
"Michael S. Tsirkin" <mst@redhat.com> wrote:

> On x86, kvm_arch_create_memslot assumes that rmap/lpage_info for the
> slot are zeroed out: if they weren't, error handling code after out_free
> label will free memory which wasn't allocated here.
> This always happens to be the case because on KVM_MR_DELETE we clear the
> whole arch structure.  So there's no bug, but it's cleaner not to rely
> on this here.

Yes, the assumption is that the function is called only with zero-sized slots.
Since changing the size is not allowed, DELETE-CREATE is the only case we
care about.

But isn't it possible to make it explicit that zero-sized slots have always
zero-cleared contents instead?  Otherwise, there would be many troubles.

	Takuya

> 
> Make the code more robust by clearing the rmap/lpage_info explicitly.
> 
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> ---
>  arch/x86/kvm/x86.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index e8ba99c..96e6eb4 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -6922,6 +6922,10 @@ int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
>  {
>  	int i;
>  
> +	/* Reset in case slot had some rmap/lpage_info. */
> +	memset(&slot->arch.rmap, 0, sizeof slot->arch.rmap);
> +	memset(&slot->arch.lpage_info, 0, sizeof slot->arch.lpage_info);
> +
>  	for (i = 0; i < KVM_NR_PAGE_SIZES; ++i) {
>  		unsigned long ugfn;
>  		int lpages;
> -- 
> MST
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html


-- 
Takuya Yoshikawa <takuya.yoshikawa@gmail.com>