Received-SPF: pass (google.com: domain of linux-kernel+bounces-151907-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Date: Fri, 19 Apr 2024 20:34:34 +0100
From: David Woodhouse <dwmw2@infradead.org>
To: "Chen, Zide" <zide.chen@intel.com>, Jack Allister <jalliste@amazon.com>,
 Paolo Bonzini <pbonzini@redhat.com>, Jonathan Corbet <corbet@lwn.net>,
 Sean Christopherson <seanjc@google.com>,
 Thomas Gleixner <tglx@linutronix.de>, Ingo Molnar <mingo@redhat.com>,
 Borislav Petkov <bp@alien8.de>, Dave Hansen <dave.hansen@linux.intel.com>,
 x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>,
 Shuah Khan <shuah@kernel.org>
CC: Paul Durrant <paul@xen.org>, kvm@vger.kernel.org, linux-doc@vger.kernel.org,
 linux-kernel@vger.kernel.org, linux-kselftest@vger.kernel.org
Subject: =?US-ASCII?Q?Re=3A_=5BPATCH_2/2=5D_KVM=3A_selftests=3A_Add_KVM/PV_c?=
 =?US-ASCII?Q?lock_selftest_to_prove_timer_drift_correction?=
User-Agent: K-9 Mail for Android
In-Reply-To: <3664e8ec-1fa1-48c0-a80d-546b7f6cd671@intel.com>
References: <20240408220705.7637-1-jalliste@amazon.com> <20240408220705.7637-3-jalliste@amazon.com> <3664e8ec-1fa1-48c0-a80d-546b7f6cd671@intel.com>
Message-ID: <FABCFBD0-4B76-4662-9F7B-7E1A856BBBB6@infradead.org>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable

On 19 April 2024 18:13:16 BST, "Chen, Zide" <zide=2Echen@intel=2Ecom> wrote=
:
>I'm wondering what's the underling theory that we definitely can achieve
>=C2=B11ns accuracy? I tested it on a Sapphire Rapids @2100MHz TSC frequen=
cy,
>and I can see delta_corrected=3D2 in ~2% cases=2E

Hm=2E Thanks for testing!

So the KVM clock is based on the guest TSC=2E Given a delta between the gu=
est TSC T and some reference point in time R, the KVM clock is expressed as=
 a(T-R)+r, where little r is the value of the KVM clock when the guest TSC =
was R, and (a) is the rate of the guest TSC=2E

When set the clock with KVM_SET_CLOCK_GUEST, we are changing the values of=
 R and r to a new point in time=2E Call the new ones Q and q respectively=
=2E

But we calculate precisely (within 1ns at least) what the KVM clock would =
have been with the *old* formula, and adjust our new offset (q) so that at =
our new reference TSC value Q, the formulae give exactly the same result=2E

And because the *rates* are the same, they should continue to give the sam=
e results, =C2=B11ns=2E

Or such *was* my theory, at least=2E=20

Would be interesting to see it disproven with actual numbers for the old+n=
ew pvclock structs, so I can understand where the logic goes wrong=2E

Were you using frequency scaling?