__supported_pte_mask |= _PAGE_IOMAP twice?

__supported_pte_mask |= _PAGE_IOMAP twice?

[Josip Rodin]

Hi,

I don't really understand this, but in xen-2.6.36/dom0/core there's this:

% git diff 0c2b1aa4c31830d3b9a40061d1b5f6c8cd837d57 057e5006127460b11e03bf80227881c528b0590d
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 167ca78..0850bd6 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -1195,12 +1195,12 @@ asmlinkage void __init xen_start_kernel(void)
 
        pgd = (pgd_t *)xen_start_info->pt_base;
 
-       if (xen_initial_domain())
-               __supported_pte_mask |= _PAGE_IOMAP;
-       else
+       if (!xen_initial_domain())
                __supported_pte_mask &= ~(_PAGE_PWT | _PAGE_PCD);
 
        __supported_pte_mask |= _PAGE_IOMAP;
+
+       __supported_pte_mask |= _PAGE_IOMAP;
        /* Don't do the full vcpu_info placement stuff until we have a
           possible map and a non-dummy shared_info. */
        per_cpu(xen_vcpu, 0) = &HYPERVISOR_shared_info->vcpu_info[0];


What's the point of doing the same thing twice at the end? Maybe it's
a mis-merge?

-- 
     2. That which causes joy or happiness.

Re: __supported_pte_mask |= _PAGE_IOMAP twice?

[Josip Rodin]

On Sun, Aug 22, 2010 at 02:06:36PM +0200, joy wrote:
> Hi,
> 
> I don't really understand this, but in xen-2.6.36/dom0/core there's this:
> 
> % git diff 0c2b1aa4c31830d3b9a40061d1b5f6c8cd837d57 057e5006127460b11e03bf80227881c528b0590d
> diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
> index 167ca78..0850bd6 100644
> --- a/arch/x86/xen/enlighten.c
> +++ b/arch/x86/xen/enlighten.c
> @@ -1195,12 +1195,12 @@ asmlinkage void __init xen_start_kernel(void)
>  
>         pgd = (pgd_t *)xen_start_info->pt_base;
>  
> -       if (xen_initial_domain())
> -               __supported_pte_mask |= _PAGE_IOMAP;
> -       else
> +       if (!xen_initial_domain())
>                 __supported_pte_mask &= ~(_PAGE_PWT | _PAGE_PCD);
>  
>         __supported_pte_mask |= _PAGE_IOMAP;
> +
> +       __supported_pte_mask |= _PAGE_IOMAP;
>         /* Don't do the full vcpu_info placement stuff until we have a
>            possible map and a non-dummy shared_info. */
>         per_cpu(xen_vcpu, 0) = &HYPERVISOR_shared_info->vcpu_info[0];
> 
> 
> What's the point of doing the same thing twice at the end? Maybe it's
> a mis-merge?

It looks like it, because in:

% git diff 3fe0344faf7fdcb158bd5c1a9aec960a8d70c8e8 xen/dom0/core

...that hunk is just:

@@ -1031,6 +1066,8 @@ asmlinkage void __init xen_start_kernel(void)
        if (!xen_initial_domain())
                __supported_pte_mask &= ~(_PAGE_PWT | _PAGE_PCD);
 
+       __supported_pte_mask |= _PAGE_IOMAP;
+
 #ifdef CONFIG_X86_64
        /* Work out if we support NX */
        check_efer();

-- 
     2. That which causes joy or happiness.

Re: __supported_pte_mask |= _PAGE_IOMAP twice?

[Ian Campbell]

On Sun, 2010-08-22 at 13:06 +0100, Josip Rodin wrote:
> Hi,
> 
> I don't really understand this, but in xen-2.6.36/dom0/core there's this:
> 
> % git diff 0c2b1aa4c31830d3b9a40061d1b5f6c8cd837d57 057e5006127460b11e03bf80227881c528b0590d
> diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
> index 167ca78..0850bd6 100644
> --- a/arch/x86/xen/enlighten.c
> +++ b/arch/x86/xen/enlighten.c
> @@ -1195,12 +1195,12 @@ asmlinkage void __init xen_start_kernel(void)
>  
>         pgd = (pgd_t *)xen_start_info->pt_base;
>  
> -       if (xen_initial_domain())
> -               __supported_pte_mask |= _PAGE_IOMAP;
> -       else
> +       if (!xen_initial_domain())
>                 __supported_pte_mask &= ~(_PAGE_PWT | _PAGE_PCD);
>  
>         __supported_pte_mask |= _PAGE_IOMAP;
> +
> +       __supported_pte_mask |= _PAGE_IOMAP;
>         /* Don't do the full vcpu_info placement stuff until we have a
>            possible map and a non-dummy shared_info. */
>         per_cpu(xen_vcpu, 0) = &HYPERVISOR_shared_info->vcpu_info[0];
> 
> 
> What's the point of doing the same thing twice at the end? Maybe it's
> a mis-merge?

Yes, I think its a mismerge. The same thing was in xen/stable-2.6.32.x
until recently as well.

Ian.