From 44f200d67df7bb519e2d97ca58cba67092fa8044 Mon Sep 17 00:00:00 2001
From: Rob Bradford <rbradford@rivosinc.com>
Date: Tue, 19 Sep 2023 10:30:45 +0100
Subject: [PATCH] hypervisor: Set destination vCPU TSC frequency to source

Include the TSC frequency as part of the KVM state so that it will be
restored at the destination.

This ensures migration works correctly between hosts that have a
different TSC frequency if the guest is running with TSC as the source
of timekeeping.

Fixes: #5786

Signed-off-by: Rob Bradford <rbradford@rivosinc.com>
---
 hypervisor/src/cpu.rs            | 13 +++++++++++++
 hypervisor/src/kvm/mod.rs        | 23 +++++++++++++++++++++++
 hypervisor/src/kvm/x86_64/mod.rs |  1 +
 vmm/src/seccomp_filters.rs       |  2 ++
 4 files changed, 39 insertions(+)

diff --git a/hypervisor/src/cpu.rs b/hypervisor/src/cpu.rs
index 578abe4d6..a240c17e2 100644
--- a/hypervisor/src/cpu.rs
+++ b/hypervisor/src/cpu.rs
@@ -252,6 +252,11 @@ pub enum HypervisorCpuError {
     ///
     #[error("Failed to get TSC frequency: {0}")]
     GetTscKhz(#[source] anyhow::Error),
+    ///
+    /// Error setting TSC frequency
+    ///
+    #[error("Failed to set TSC frequency: {0}")]
+    SetTscKhz(#[source] anyhow::Error),
 }
 
 #[derive(Debug)]
@@ -447,6 +452,7 @@ pub trait Vcpu: Send + Sync {
     /// Return the list of initial MSR entries for a VCPU
     ///
     fn boot_msr_entries(&self) -> Vec<MsrEntry>;
+
     #[cfg(target_arch = "x86_64")]
     ///
     /// Get the frequency of the TSC if available
@@ -454,4 +460,11 @@ pub trait Vcpu: Send + Sync {
     fn tsc_khz(&self) -> Result<Option<u32>> {
         Ok(None)
     }
+    #[cfg(target_arch = "x86_64")]
+    ///
+    /// Set the frequency of the TSC if available
+    ///
+    fn set_tsc_khz(&self, _freq: u32) -> Result<()> {
+        Ok(())
+    }
 }
diff --git a/hypervisor/src/kvm/mod.rs b/hypervisor/src/kvm/mod.rs
index c4fa9f74f..cf8cf8e23 100644
--- a/hypervisor/src/kvm/mod.rs
+++ b/hypervisor/src/kvm/mod.rs
@@ -1890,6 +1890,7 @@ impl cpu::Vcpu for KvmVcpu {
         };
 
         let vcpu_events = self.get_vcpu_events()?;
+        let tsc_khz = self.tsc_khz()?;
 
         Ok(VcpuKvmState {
             cpuid,
@@ -1902,6 +1903,7 @@ impl cpu::Vcpu for KvmVcpu {
             xsave,
             xcrs,
             mp_state,
+            tsc_khz,
         }
         .into())
     }
@@ -2004,6 +2006,10 @@ impl cpu::Vcpu for KvmVcpu {
         self.set_lapic(&state.lapic_state)?;
         self.set_fpu(&state.fpu)?;
 
+        if let Some(freq) = state.tsc_khz {
+            self.set_tsc_khz(freq)?;
+        }
+
         // Try to set all MSRs previously stored.
         // If the number of MSRs set from SET_MSRS is different from the
         // expected amount, we fallback onto a slower method by setting MSRs
@@ -2186,6 +2192,23 @@ impl cpu::Vcpu for KvmVcpu {
             Ok(v) => Ok(Some(v)),
         }
     }
+
+    #[cfg(target_arch = "x86_64")]
+    ///
+    /// Set the frequency of the TSC if available
+    ///
+    fn set_tsc_khz(&self, freq: u32) -> cpu::Result<()> {
+        match self.fd.set_tsc_khz(freq) {
+            Err(e) => {
+                if e.errno() == libc::EIO {
+                    Ok(())
+                } else {
+                    Err(cpu::HypervisorCpuError::SetTscKhz(e.into()))
+                }
+            }
+            Ok(_) => Ok(()),
+        }
+    }
 }
 
 impl KvmVcpu {
diff --git a/hypervisor/src/kvm/x86_64/mod.rs b/hypervisor/src/kvm/x86_64/mod.rs
index 2a0ed6a6d..78da26d60 100644
--- a/hypervisor/src/kvm/x86_64/mod.rs
+++ b/hypervisor/src/kvm/x86_64/mod.rs
@@ -67,6 +67,7 @@ pub struct VcpuKvmState {
     pub xsave: Xsave,
     pub xcrs: ExtendedControlRegisters,
     pub mp_state: MpState,
+    pub tsc_khz: Option<u32>,
 }
 
 impl From<StandardRegisters> for kvm_regs {
diff --git a/vmm/src/seccomp_filters.rs b/vmm/src/seccomp_filters.rs
index 7b37e4316..29d2df60f 100644
--- a/vmm/src/seccomp_filters.rs
+++ b/vmm/src/seccomp_filters.rs
@@ -366,6 +366,7 @@ fn create_vmm_ioctl_seccomp_rule_kvm() -> Result<Vec<SeccompRule>, BackendError>
     const KVM_SET_LAPIC: u64 = 0x4400_ae8f;
     const KVM_SET_MSRS: u64 = 0x4008_ae89;
     const KVM_SET_SREGS: u64 = 0x4138_ae84;
+    const KVM_SET_TSC_KHZ: u64 = 0xaea2;
     const KVM_SET_TSS_ADDR: u64 = 0xae47;
     const KVM_SET_XCRS: u64 = 0x4188_aea7;
     const KVM_SET_XSAVE: u64 = 0x5000_aea5;
@@ -392,6 +393,7 @@ fn create_vmm_ioctl_seccomp_rule_kvm() -> Result<Vec<SeccompRule>, BackendError>
         and![Cond::new(1, ArgLen::Dword, Eq, KVM_SET_IDENTITY_MAP_ADDR)?],
         and![Cond::new(1, ArgLen::Dword, Eq, KVM_SET_LAPIC)?],
         and![Cond::new(1, ArgLen::Dword, Eq, KVM_SET_SREGS)?],
+        and![Cond::new(1, ArgLen::Dword, Eq, KVM_SET_TSC_KHZ)?],
         and![Cond::new(1, ArgLen::Dword, Eq, KVM_SET_TSS_ADDR,)?],
         and![Cond::new(1, ArgLen::Dword, Eq, KVM_SET_MSRS)?],
         and![Cond::new(1, ArgLen::Dword, Eq, KVM_SET_XCRS,)?],