Re: [Xen-devel] [PATCH] x86: Don't crash when mapping a page using EFI runtime page tables

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 15.05.15 at 18:08, <ross.lagerwall@xxxxxxxxxx> wrote:
> --- a/xen/arch/x86/domain_page.c
> +++ b/xen/arch/x86/domain_page.c
> @@ -32,20 +32,25 @@ static inline struct vcpu *mapcache_current_vcpu(void)
>          return NULL;
>  
>      /*
> +     * When using efi runtime page tables, we have the equivalent of the idle
> +     * domain's page tables but current may point at another domain's VCPU.
> +     * Return NULL as though current is not properly set up yet.
> +     */
> +    if ( efi_enabled && read_cr3() == efi_rs_page_table() )
> +        return NULL;

I'm okay with the patch in principle; what worries me is the CR3 read
that is now going to be necessary even in non-debug builds. With
this code being the only user of efi_rs_page_table(), I wonder if it
wouldn't make sense to alter that function to return non-zero only
when spin_is_locked(&efi_rs_lock), and then alter the code above
such that the CR3 read would happen only when we got a non-zero
value back.

And with that I would then also like to fold the potential double CR3
reads in debug builds to at most one.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel