Re: [PATCH 02/10] KVM: x86: Improve accuracy of KVM clock when TSC scaling is in force

[David Woodhouse <dwmw2@infradead.org>]

[-- Attachment #1: Type: text/plain, Size: 2027 bytes --]

On Mon, 2024-04-22 at 14:22 +0200, Paolo Bonzini wrote:
> On Thu, Apr 18, 2024 at 9:51 PM David Woodhouse <dwmw2@infradead.org> wrote:
> >          gpa_t time;
> >          struct pvclock_vcpu_time_info hv_clock;
> > -       unsigned int hw_tsc_khz;
> > +       unsigned int hw_tsc_hz;
> 
> Why not change this to u64? 4.3 GHz is scarily close to current
> processors, though I expect that it will break a lot more software
> than just KVM.

I'm not sure that just changing the variable is sufficient. I'm
concerned that we may still have places (like kvm_get_time_scale(),
although it's hard to tell because it's entirely uncommented!) which
assume that they can multiply the value by *any* 32-bit number and
it'll still fit in 64 bits.

> >   static int kvm_guest_time_update(struct kvm_vcpu *v)
> >   {
> > -       unsigned long flags, tgt_tsc_khz;
> > +       unsigned long flags;
> > +       uint64_t tgt_tsc_hz;
> 
> ... especially considering that you did use a 64-bit integer here
> (though---please use u64 not uint64_t; and BTW if you want to add a
> patch to change kvm_get_time_scale() to u64, please do.

Meh, I'm used to programming in C. Yes, I *am* old enough to have been
doing this since the last decade of the 1900s, but it *has* been a long
time since 1999, and my fingers have learned :)

That said, although I *do* write in C by default, where I'm editing
functions which already have that old anachronistic crap, I generally
manage to go back and change my additions from proper C to match their
surroundings. And I suppose there *are* some of those awful u64/s64
types already in kvm_guest_time_update(), so I would normally have done
so. I'll change that in my tree.

No way I'm regressing kvm_get_time_scale() myself though :)

Heh, looks like it was you who made it uint64_t, in 2016. In a commit
(3ae13faac) which said "Prepare for improving the precision in the next
patch"... which never came, AFAICT?


[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 5965 bytes --]