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

Re: [Xen-devel] [PATCH RFC v2] x86/domctl: Don't pause the whole domain if only getting vcpu state

>>> On 21.02.18 at 19:24, <dunlapg@xxxxxxxxx> wrote:
> On Fri, Oct 6, 2017 at 11:02 AM, Alexandru Isaila
> <aisaila@xxxxxxxxxxxxxxx> wrote:
>> This patch adds the hvm_save_one_cpu_ctxt() function.
>> It optimizes by only pausing the vcpu on all HVMSR_PER_VCPU save
>> callbacks where only data for one VCPU is required.
> Sorry it's taken so long to get back to you on this one.
> So first of all, a big issue with this patch in general is that
> there's way too much code duplication.  Duplicating the code in every
> case will not only increase HV code size and decrease cache
> effectiveness, it will also increase the likelihood of subtle bugs
> creeping in when the two copies don't match up.  If you were going to
> do it this way I think you'd basically have to make a *_save_*_one()
> function for each callback.
> The only other option would be to pass a cpumask instead, and set
> either a single bit or all the bits; but that seems a bit wasteful.
> (Jan, Andy -- if you don't like this solution, now is the time to say
> so.)

First of all it would need to be a vcpumask, and then it wouldn't
scale well (at least as soon as the maximum number of HVM vCPU-s
isn't limited to as little as 128 anymore, which Intel folks are
already trying to rectify, albeit sadly without first addressing all
the issues with higher vCPU counts).


Xen-devel mailing list



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