Re: [PATCH 2/5] KVM: New guest debug interface

[Jan Kiszka <jan.kiszka@siemens.com>]

Avi Kivity wrote:
> Jan Kiszka wrote:
>> FYI: I just passed the kgdbts test with some Linux guest on an Opteron
>> 2216. Also using hbreak with gdb inside the guest appears to work. There
>> are just a few to-be-analyzed guest lockups/kernel panics that may
>> relate to the fact that the kernels were build for Intel CPUs. On the
>> other hand, it would be far too simple if it worked already - compared
>> to the pain I went through with VMX...
>>   
> 
> So long as it doesn't crash, this functionality is far enough from
> mainstream use.

I would love to dig a bit deeper, doing more SMP tests, maybe also
adding userspace NMI support to SVM at this chance, but I was just
called away again. Hope to find some time next week, we'll see.

> 
> btw, is it possible to emulate the old interface on top of the new
> implementation?  I feel unconfortable with dropping an interface (albeit
> an unused one) without even a warning.

Would be possible, I guess - but do we really have to care? It only
worked for VMX, only on UP, qemu has no support for it (I'm planning to
submit the new interface once it's merged and settled), and I don't
think many people actually touched the old one so far. And those who may
throw the old IOCTL at a new kernel will get a proper error on return.
Is this problematic?

Jan

-- 
Siemens AG, Corporate Technology, CT SE 26
Corporate Competence Center Embedded Linux