[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH 2/5] xen/vm_access: Support for memory-content hiding



>>> On 08.06.15 at 12:02, <rcojocaru@xxxxxxxxxxxxxxx> wrote:
> On 05/08/2015 07:07 PM, Jan Beulich wrote:
>>> > --- a/xen/include/asm-x86/domain.h
>>> > +++ b/xen/include/asm-x86/domain.h
>>> > @@ -512,6 +513,7 @@ struct arch_vcpu
>>> >          uint32_t emulate_flags;
>>> >          unsigned long gpa;
>>> >          unsigned long eip;
>>> > +        struct vm_event_emul_read_data emul_read_data;
>> Considering the size of this structure I don't think this should be
>> there for all vCPU-s of all guests. IOW - please allocate this
>> dynamically only on domains where it's needed.
> 
> Looking at the code, it's not immediately clear how to differentiate
> between a domain where this is needed and one where it is not. At domain
> init time we don't know, because it might become needed as soon as a
> vm_event client becomes attached to the domain. Then again, even once a
> vm_event client attaches to the domain, it might still not be necessary
> to allocate that structure, as the client might never reply with
> MEM_ACCESS_SET_EMUL_READ_DATA.
> 
> The only place where we know for sure that it's needed is in
> p2m_mem_access_check(), when receiving a vm_event response with
> MEM_ACCESS_SET_EMUL_READ_DATA (lazy allocation). Would this be alright?

I can't really judge about that; the farther you can safely defer it,
the better. For the sake of addressing my complaint it would
probably suffice to defer it until a vm_event client attaches to the
guest.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.