Re: [Xen-devel] [PATCH v12 1/3] xen/mem_sharing: VM forking

[Julien Grall <julien@xen.org>]

Hi,

On 25/03/2020 16:34, Tamas K Lengyel wrote:
>>> diff --git a/xen/arch/x86/mm/p2m.c b/xen/arch/x86/mm/p2m.c
>>> index 9f51370327..1ed7d13084 100644
>>> --- a/xen/arch/x86/mm/p2m.c
>>> +++ b/xen/arch/x86/mm/p2m.c
>>> @@ -509,6 +509,12 @@ mfn_t __get_gfn_type_access(struct p2m_domain *p2m, unsigned long gfn_l,
>>>
>>>       mfn = p2m->get_entry(p2m, gfn, t, a, q, page_order, NULL);
>>>
>>> +    /* Check if we need to fork the page */
>>> +    if ( (q & P2M_ALLOC) && p2m_is_hole(*t) &&
>>> +         !mem_sharing_fork_page(p2m->domain, gfn, q & P2M_UNSHARE) )
>>> +        mfn = p2m->get_entry(p2m, gfn, t, a, q, page_order, NULL);
>>> +
>>> +    /* Check if we need to unshare the page */
>>>       if ( (q & P2M_UNSHARE) && p2m_is_shared(*t) )
>>>       {
>>>           ASSERT(p2m_is_hostp2m(p2m));
>>> @@ -588,7 +594,8 @@ struct page_info *p2m_get_page_from_gfn(
>>>               return page;
>>>
>>>           /* Error path: not a suitable GFN at all */
>>> -        if ( !p2m_is_ram(*t) && !p2m_is_paging(*t) && !p2m_is_pod(*t) )
>>> +        if ( !p2m_is_ram(*t) && !p2m_is_paging(*t) && !p2m_is_pod(*t) &&
>>> +             !mem_sharing_is_fork(p2m->domain) )
>>>               return NULL;
>>>       }
>>>
>>> diff --git a/xen/common/domain.c b/xen/common/domain.c
>>> index b4eb476a9c..62aed53a16 100644
>>> --- a/xen/common/domain.c
>>> +++ b/xen/common/domain.c
>>> @@ -1270,6 +1270,9 @@ int map_vcpu_info(struct vcpu *v, unsigned long gfn, unsigned offset)
>>>
>>>       v->vcpu_info = new_info;
>>>       v->vcpu_info_mfn = page_to_mfn(page);
>>> +#ifdef CONFIG_MEM_SHARING
>>> +    v->vcpu_info_offset = offset;
>>
>> There's no need to introduce this field, you can just use v->vcpu_info
>> & ~PAGE_MASK AFAICT.
> 
> Just doing what you suggest above results in:
> 
> mem_sharing.c:1603:55: error: invalid operands to binary & (have
> ‘vcpu_info_t * const’ {aka ‘union <anonymous> * const’} and ‘long
> int’)
>                                       d_vcpu->vcpu_info & ~PAGE_MASK);
> 
> I can of course cast the vcpu_info pointer to (long int), it's just a
> bit ugly. Thoughts?

FWIW, I will also need the offset for liveupdate. I have used (unsigned 
long)v->vcpu_info & ~PAGE_MASK so far but this is not really pretty.

So I am all for either a new field or a macro hiding this uglyness.

Cheers,

-- 
Julien Grall