RE: [Xen-devel] [PATCH] replace rdtsc emulation-vs-native xen boot option with per-domain (hypervisor part)

[Dan Magenheimer <dan.magenheimer@xxxxxxxxxx>]

Hi Jeremy --

I think it's fair for us both to completely air our opinions,
but since you've contradicted some of my "hand-wavy" points
and since much of your position is based on the contradictions, 
I feel it's necessary to respond.

> The TSC is not, and has never been reliable.

Your data is stale.  Please discuss this with processor
and system vendors (I have) and look at the latest upstream
Linux.

I agree that this presumed reliability should be carefully
tested.  (See my proposed patch posted yesterday.)  If
you are correct and I am not, your position (that "all
apps that use rdtsc are fundamentally broken") is much
stronger.

> Except that it comes with a terrible cost...
> This is a massive regression...

It is certainly significant but "terrible" and "massive"
are a bit strong.  Based on my measurements, the examples
you cite will degrade performance by a fraction of a percent.
And this loss can be eliminated IF the administrator
understands the risks and chooses performance over
correctness by specifying an option at guest creation.
 
> The fact that you haven't named a single real app...
> Are you really arguing on the basis that "some apps
> might use tsc in a fragile way" or do you actually have a 
> specific list

I have a (small) specific list.  For various reasons,
I cannot go into further detail.

But if TSC is indeed reliable on newer processors/systems
(see above) AND on VMware, there will certainly be more.

Thanks,
Dan

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