On Mon, Nov 01, 2021, Vitaly Kuznetsov wrote:

Sean Christopherson [off-list ref] writes:

quoted

Refactor the "extended" path of kvm_hv_flush_tlb() to reduce the nesting
depth for the non-fast sparse path, and to make the code more similar to
the extended path in kvm_hv_send_ipi().

No functional change intended.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/hyperv.c | 40 +++++++++++++++++++++-------------------
 1 file changed, 21 insertions(+), 19 deletions(-)

diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c
index cf18aa1712bf..e68931ed27f6 100644
--- a/arch/x86/kvm/hyperv.c
+++ b/arch/x86/kvm/hyperv.c

@@ -1814,31 +1814,33 @@ static u64 kvm_hv_flush_tlb(struct kvm_vcpu *vcpu, struct kvm_hv_hcall *hc, bool
 		if (hc->var_cnt != bitmap_weight((unsigned long *)&valid_bank_mask, 64))
 			return HV_STATUS_INVALID_HYPERCALL_INPUT;
 
-		if (!hc->var_cnt && !all_cpus)
+		if (all_cpus)
+			goto do_flush;

You could've probably done:

	if (all_cpus) {
		kvm_make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH_GUEST);
		goto ret_success;
	}

to get rid on the second 'all_cpus' check (and maybe even 'do_flush'
label with some extra work) below.

Yeah, but the !ex path also uses all_cpus, and in general I'd prefer to keep the
two flush requests bundled together.