Re: [Xen-devel] [RFC] x86/vm_event: Allow returning i-cache for emulation

[George Dunlap <george.dunlap@xxxxxxxxxx>]

On 12/09/16 15:31, Tamas Lengyel wrote:
> On Sep 12, 2016 08:17, "George Dunlap" <george.dunlap@xxxxxxxxxx> wrote:
>>
>> On 09/09/16 16:41, Tamas K Lengyel wrote:
>>> When emulating instructions the emulator maintains a small i-cache
> fetched
>>> from the guest memory. Under certain scenarios this memory region may
> contain
>>> instructions that a monitor subscriber would prefer to hide, namely
> INT3, and
>>> instead would prefer to emulate a different instruction in-place.
>>>
>>> This patch extends the vm_event interface to allow returning this
> i-cache via
>>> the vm_event response.
>>
>> So do you have a problem right now with stale caches (i.e., you modify
>> an INT3 back to something else in guest RAM but the emulator still
>> emulates the INT3)?  Or is the idea here that instead of doing the
>> replace-singlestep-replace loop, you just tell the emulator, "Here,
>> emulate this instead" (without removing the INT3 from guest memory at
> all)?
>>
>> (Or am I completely missing the point here?)
>>
> 
> Hi George,
> it's the latter! This would make tracing with int3s a bit more flexible on
> multi-vcpu guests as there would be no racecondition. I use altp2m right
> now to get around this problem but it's always good to have a backup in
> case altp2m is disabled.

OK -- in that case, it sounds like a good idea (particularly since
there's a race I hadn't considered).  :-)

 -George

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