Re: [PATCH V3 1/3] xen/mem_access: Support for memory-content hiding

[Razvan Cojocaru <rcojocaru@bitdefender.com>]

On 07/06/2015 07:50 PM, Lengyel, Tamas wrote:
> Handy feature, thanks for doing it!

You're very welcome, I'm quite happy you find it useful.

>     @@ -1466,6 +1466,10 @@ void p2m_mem_access_emulate_check(struct vcpu *v,
>              }
> 
>              v->arch.vm_event.emulate_flags = violation ? rsp->flags : 0;
>     +
>     +        if ( rsp->flags & MEM_ACCESS_SET_EMUL_READ_DATA &&
> 
> 
> So one of the use-cases for this feature I would have is using it in the
> vm_event response to the software breakpoint event. A little bit of
> context: I have written 0xCC into target memory locations, which are
> further protected by mem_access R/W. This setup right now would suffice
> to hide the content easily from the R mem_access events without having
> to remove it. But for X tracing I'm not using mem_access events. Here
> this feature is locked to be only in response to a mem_access events. I
> can *technically* work around that by changing the response type to the
> mem_access event, but it would be nice if this feature would be clearly
> available for non-mem_access events as well (or at least for software
> breakpoint). What do you think, does that usecase make sense here?

It does make a fair ammount of sense, but with this little time until
the feature freeze it does look like a non-trivial change, so if it
doesn't bother you very much to use the plain mem_access event while we
work towards making the response more general, I'd go for trying to get
this patch into 4.6, whatever its chances are.

If you'd prefer that I do some ground work for the future (i.e. rename
MEM_ACCESS constants, etc.), please let me know.


Thanks,
Razvan