On 2019/10/23 5:03, Sean Christopherson wrote:
On Tue, Oct 22, 2019 at 08:46:46PM +0800, Zhenzhong Duan wrote:
Hi Vitaly,
On 2019/10/22 19:36, Vitaly Kuznetsov wrote:
Zhenzhong Duan<zhenzhong.duan@xxxxxxxxxx> writes:
...snip
diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
index 249f14a..3945aa5 100644
--- a/arch/x86/kernel/kvm.c
+++ b/arch/x86/kernel/kvm.c
@@ -825,18 +825,36 @@ __visible bool __kvm_vcpu_is_preempted(long cpu)
*/
void __init kvm_spinlock_init(void)
{
- /* Does host kernel support KVM_FEATURE_PV_UNHALT? */
- if (!kvm_para_has_feature(KVM_FEATURE_PV_UNHALT))
+ /*
+ * In case host doesn't support KVM_FEATURE_PV_UNHALT there is still an
+ * advantage of keeping virt_spin_lock_key enabled: virt_spin_lock() is
+ * preferred over native qspinlock when vCPU is preempted.
+ */
+ if (!kvm_para_has_feature(KVM_FEATURE_PV_UNHALT)) {
+ pr_info("PV spinlocks disabled, no host support.\n");
return;
+ }
+ /*
+ * Disable PV qspinlock and use native qspinlock when dedicated pCPUs
+ * are available.
+ */
if (kvm_para_has_hint(KVM_HINTS_REALTIME)) {
- static_branch_disable(&virt_spin_lock_key);
- return;
+ pr_info("PV spinlocks disabled with KVM_HINTS_REALTIME hints.\n");
+ goto out;
}
- /* Don't use the pvqspinlock code if there is only 1 vCPU. */
- if (num_possible_cpus() == 1)
- return;
+ if (num_possible_cpus() == 1) {
+ pr_info("PV spinlocks disabled, single CPU.\n");
+ goto out;
+ }
+
+ if (nopvspin) {
+ pr_info("PV spinlocks disabled, forced by \"nopvspin\" parameter.\n");
+ goto out;
+ }
+
+ pr_info("PV spinlocks enabled\n");
__pv_init_lock_hash();
pv_ops.lock.queued_spin_lock_slowpath = __pv_queued_spin_lock_slowpath;
@@ -849,6 +867,8 @@ void __init kvm_spinlock_init(void)
pv_ops.lock.vcpu_is_preempted =
PV_CALLEE_SAVE(__kvm_vcpu_is_preempted);
}
+out:
+ static_branch_disable(&virt_spin_lock_key);
You probably need to add 'return' before 'out:' as it seems you're
disabling virt_spin_lock_key in all cases now).
virt_spin_lock_key is kept enabled in !kvm_para_has_feature(KVM_FEATURE_PV_UNHALT)
case which is the only case virt_spin_lock() optimization is used.
When PV qspinlock is enabled, virt_spin_lock() isn't called in
__pv_queued_spin_lock_slowpath() in which case we don't care
virt_spin_lock_key's value.
So adding 'return' or not are both ok, I chosed to save a line,
let me know if you prefer to add a 'return' and I'll change it.
It'd be worth adding a comment here if you end up spinning another version
to change the logging prefix. The logic is sound and I like the end
result, but I had the same knee jerk "this can't be right!?!?" reaction as
Vitaly.
Sure, will do in next version.
Thanks
Zhenzhong
