Re: [PATCH] Handle vma regions with no backing page (v2)

[Avi Kivity <avi@qumranet.com>]

Anthony Liguori wrote:
>>
>>  
>>>  
>>>  struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
>>>  {
>>> -    return pfn_to_page(gfn_to_pfn(kvm, gfn));
>>> +    pfn_t pfn;
>>> +
>>> +    pfn = gfn_to_pfn(kvm, gfn);
>>> +    if (pfn_valid(pfn))
>>> +        return pfn_to_page(pfn);
>>> +
>>> +    return NULL;
>>>  }
>>>       
>>
>> You're returning NULL here, not bad_page.
>>   
>
> My thinking was that bad_page indicates that the gfn is invalid.  This 
> is a different type of error though.  The problem is that the guest is 
> we are trying to kmap() a page that has no struct page associated with 
> it.  I'm not sure what the right thing to do here is.
>

It depends on what's going on?  Does a page table point to mmio?  Or the 
glommerclock?

Not sure there is a single answer.

> Perhaps we should be replacing consumers of gfn_to_page() with 
> copy_to_user() instead?

Indeed we should.  The problem is access in atomic contexts.  It's easy 
to detect failure, but not always easy to handle it.

-- 
Any sufficiently difficult bug is indistinguishable from a feature.


-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone