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

RE: [Xen-devel] VCPU Structure



Hi Satyajeet –

 

I’m still not quite sure I understand, though your description is getting better.

 

A hypercall executes on a VCPU on a PCPU.   One can’t transfer control to a VCPU structure; a VCPU structure contains data.  Are you doing a different memory integrity hash block for each VCPU in each guest?  Or one for each guest?  Or is it one for the entire physical system?  If it is one for each VCPU in each guest, yes, you will need to store the pre-defined range and the results of the hash in the VCPU structure and your “compute hash” hypercall would store those values and your “check hash” hypercall would verify the value (and return success or failure).  If this is what you plan to do, you do not need a trap/exception/fault as the hypercall replaces that.


Dan

 

From: Nimgaonkar, Satyajeet [mailto:SatyajeetNimgaonkar@xxxxxxxxxx]
Sent: Tuesday, November 09, 2010 10:07 AM
To: Nimgaonkar, Satyajeet; Dan Magenheimer; Xen Devel
Subject: RE: [Xen-devel] VCPU Structure

 

Hi Dan,
           I have created a custom hypercall, would it be possible to transfert the control from the hypercall handler to the VCPU structure. With regards to modifying the VCPU, can you please redirect me to any code examples containing trapping/exception/faulting the CPU. Also for the memory integrity hash block that I wish to add to the VCPU, I will compute the hash for a pre-defined range of memory and match it every day i.e. 24 hours.
           Thank you very much.

Regards,
Satyajeet


From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx [xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] on behalf of Nimgaonkar, Satyajeet [SatyajeetNimgaonkar@xxxxxxxxxx]
Sent: Monday, November 08, 2010 11:58 AM
To: Dan Magenheimer; Xen Devel
Subject: RE: [Xen-devel] VCPU Structure

Hi Dan,
           I want to compute the hash for a specific memory range and not all the memory. Also I havent decided the frequency of calculating the hash and matching it. But I wonder how will that affect the functionality and modification that I intend to do?
           Thanks.

Regards,
Satyajeet


From: Dan Magenheimer [dan.magenheimer@xxxxxxxxxx]
Sent: Monday, November 08, 2010 10:59 AM
To: Nimgaonkar, Satyajeet; Xen Devel
Subject: RE: [Xen-devel] VCPU Structure

Hi Satyajeet –


No, you still will need to be much more precise.  Computing a hash of ALL memory on the system will take a very long time, and SOME memory is almost certain to change (perhaps even while computing the hash) if anything at all is running on the system, whether or not an attack is performed.  Do you want to compute a hash of some specific range of memory?  Or all guest memory for some specific guest?  And memory is not specific to one VCPU, so it is not something that you add to a VCPU.  How frequently do you want to compute the hash?  And how frequently do you want to check for a match?  And how do you determine when you want to cause the hash to be taken or checked?  And do you need to store multiple values for different ranges/guests?

 

Systems programming (in Xen or Linux or anywhere) is very complex and it is important that you describe a very precise detailed plan for what you want to do – flow chart, cause and effect, etc – before it is possible to help you.

 

From: Nimgaonkar, Satyajeet [mailto:SatyajeetNimgaonkar@xxxxxxxxxx]
Sent: Monday, November 08, 2010 9:31 AM
To: Dan Magenheimer; Xen Devel
Subject: RE: [Xen-devel] VCPU Structure

 

Hi Dan,
            A memory integrity block is a simple hash function that computes the hash of all the memory at prticular instant of time and stores it in a secure memory location. This allows to detect an attack from an adversary who is trying to alter the memory values. At a lter instant of time, the hash of memory is again computed and if the two hash values match then we conclude that no attack was performed and they don't match we conclude that an attack has been performed.
            I hope that explains the functionality that I want to add to the VCPU. Can you please tell me where should I add the trapping code with in the VPCU, I mean which file should I look at. Also please can you tell me where exactly with in xen should I add the special code. I see the VCPU structure is present in xen/include/xen/sched.h. Is it the correct place to look for.
           Thanks.

Regards,
Satyajeet


From: Dan Magenheimer [dan.magenheimer@xxxxxxxxxx]
Sent: Monday, November 08, 2010 7:42 AM
To: Nimgaonkar, Satyajeet; Xen Devel
Subject: RE: [Xen-devel] VCPU Structure

I’m not asking you to describe the Xen functionality.  I am asking you to describe in more detail the modification to the processor functionality/behavior that you are trying to achieve.  I don’t know what a “simple memory integrity block” does and searching for it doesn’t find anything helpful.


Nearly all instructions in the VCPU are executed directly in hardware (by the physical CPU == PCPU).  The only way to change the VCPU behavior is to cause some kind of trap or fault or exception to occur on the PCPU, which gets intercepted by the processor and then control is turned over to privileged software (in this case Xen).  Xen has a lot of code that handles many many different kinds of traps/faults/exceptions.  Your VCPU will need to execute an instruction that causes a trap or fault or exception and then you will need to add code to Xen to recognize your special one and do something special with it.


Does that help?


Dan

 

From: Nimgaonkar, Satyajeet [mailto:SatyajeetNimgaonkar@xxxxxxxxxx]
Sent: Sunday, November 07, 2010 10:14 PM
To: Dan Magenheimer; Xen Devel
Subject: RE: [Xen-devel] VCPU Structure

 

Hi,
   I am sorry if I have not provided provided enough information about the functionality that I want to implement. But my problem is that even I new to Xen and trying to understand my way around it. With regards to the functionality that I want to implement, I wish to modify the behavior of the VCPU in xen. As a starting point, I just want to add simple memory integrity block within the VCPU. I am really not sure what would be the best way to do this so I need some suggestions on that too.
   Thanks. I hope that helps too.

Regards,
Satyajeet


From: Dan Magenheimer [dan.magenheimer@xxxxxxxxxx]
Sent: Thursday, November 04, 2010 7:25 PM
To: Nimgaonkar, Satyajeet; Xen Devel
Subject: RE: [Xen-devel] VCPU Structure

I think nobody is responding because you haven’t provided enough information about what it is you are trying to do... and resending the same message doesn’t help.


For example, if you are “extending” the instruction set (e.g. using an illegal opcode to do something useful), that would be very different than trapping memory accesses that meet a certain criteria.

 

From: Nimgaonkar, Satyajeet [mailto:SatyajeetNimgaonkar@xxxxxxxxxx]
Sent: Thursday, November 04, 2010 2:52 PM
To: Xen Devel; Dan Magenheimer
Subject: [Xen-devel] VCPU Structure

 

 Hi,
    I want to modify the VCPU structure to implement memory integrity functionality (i.e. memory hash function), such that the VCPU functionality incorporates memory integrity. Also once this is done, I want make sure that the VMs i create (Dom0 and Domu) run on this modified VCPU itself.
    Can anyone please tell me where should I look for to implement this.
    Thank you very much.

Regards,
Satyajeet

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel

 


Rackspace

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