Xen devel - Feb 2012 - Is this a data corruption bug for p2m_ram_shared page in hvm_hap_nested_page_fault()

Hi,

In hvm_hap_nested_page_fault(), it seems that all valid write faults are now
handled
by p2m_mem_access_check(), right? Then when will mem_sharing_unshare_page()
be called? And if p2m->access_required == false, the access restrictions is
cleared
, then the data in this shared page could be corrupted by this page write
access?


Thanks,

Nai

On 2012年02月02日 18:17, nai.xia wrote:
> Hi,
>
> In hvm_hap_nested_page_fault(), it seems that all valid write faults are
now handled
> by p2m_mem_access_check(), right? Then when will mem_sharing_unshare_page()
> be called? And if p2m->access_required == false, the access restrictions
is cleared
Oh, sorry, I notice that with p2m_ram_shared, the write permissions is always
cleared.
But, still, this seems cannot lead to the call of mem_sharing_unshare_page() and
this
write fault will happen again and again?


Thanks,

Nai

> , then the data in this shared page could be corrupted by this page write
access?
>
>
> Thanks,
>
> Nai
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

At 18:25 +0800 on 02 Feb (1328207143), nai.xia wrote:
> On 2012???02???02??? 18:17, nai.xia wrote:
> >In hvm_hap_nested_page_fault(), it seems that all valid write faults
are
> >now handled
> >by p2m_mem_access_check(), right? Then when will
mem_sharing_unshare_page()
> >be called? And if p2m->access_required == false, the access
restrictions
> >is cleared
> 
> Oh, sorry, I notice that with p2m_ram_shared, the write permissions is
> always cleared.  But, still, this seems cannot lead to the call of
> mem_sharing_unshare_page() and this write fault will happen again and
> again?
There''s an explicit call of mem_sharing_unshare_page() in
hvm_hap_nested_page_fault(); the only thing that will skip that is if
the fault is caused by a p2m_access violation, in which case 
we report the fault on the p2m_access ring and wait for the consumer of
that ring to fix the problem. 

Cheers,

Tim.

On 2012年02月02日 19:19, Tim Deegan wrote:
> At 18:25 +0800 on 02 Feb (1328207143), nai.xia wrote:
>> On 2012???02???02??? 18:17, nai.xia wrote:
>>> In hvm_hap_nested_page_fault(), it seems that all valid write
faults are
>>> now handled
>>> by p2m_mem_access_check(), right? Then when will
mem_sharing_unshare_page()
>>> be called? And if p2m->access_required == false, the access
restrictions
>>> is cleared
>>
>> Oh, sorry, I notice that with p2m_ram_shared, the write permissions is
>> always cleared.  But, still, this seems cannot lead to the call of
>> mem_sharing_unshare_page() and this write fault will happen again and
>> again?
>
> There's an explicit call of mem_sharing_unshare_page() in
> hvm_hap_nested_page_fault(); the only thing that will skip that is if
> the fault is caused by a p2m_access violation, in which case
> we report the fault on the p2m_access ring and wait for the consumer of
> that ring to fix the problem.
Oh, yes! I was mistaken that the write flag of p2m_access would also be
removed for p2m_ram_shared pages just like entry->w in ept_set_entry().
I am clear about the role of the p2m_access_t now.

Thanks for the reply! :)


Regards,

Nai
>
> Cheers,
>
> Tim.
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel