Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH 1/3] KVM: LAPIC: Make lapic timer unpinned when timer is
 injected by posted-interrupt
To:     Wanpeng Li <kernellwp@gmail.com>, linux-kernel@vger.kernel.org,
        kvm@vger.kernel.org
Cc:     =?UTF-8?B?UmFkaW0gS3LEjW3DocWZ?= <rkrcmar@redhat.com>
References: <1559729351-20244-1-git-send-email-wanpengli@tencent.com>
 <1559729351-20244-2-git-send-email-wanpengli@tencent.com>
From:   Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <505fc388-2223-146f-ae8a-824169078a17@redhat.com>
Date:   Wed, 5 Jun 2019 15:04:04 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.6.1
MIME-Version: 1.0
In-Reply-To: <1559729351-20244-2-git-send-email-wanpengli@tencent.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 05/06/19 12:09, Wanpeng Li wrote:
> +static inline bool posted_interrupt_inject_timer(struct kvm_vcpu *vcpu)
> +{
> +	return (kvm_x86_ops->pi_inject_timer_enabled(vcpu) &&
> +		kvm_mwait_in_guest(vcpu->kvm));
> +}
> +

Here you need to check kvm_halt_in_guest, not kvm_mwait_in_guest,
because you need to go through kvm_apic_expired if the guest needs to be
woken up from kvm_vcpu_block.

There is a case when you get to kvm_vcpu_block with kvm_halt_in_guest,
which is when the guest disables asynchronous page faults.  Currently,
timer interrupts are delivered while apf.halted = true, with this change
they wouldn't.  I would just disable KVM_REQ_APF_HALT in
kvm_can_do_async_pf if kvm_halt_in_guest is true, let me send a patch
for that later.

When you do this, I think you don't need the
kvm_x86_ops->pi_inject_timer_enabled check at all, because if we know
that the vCPU cannot be asleep in kvm_vcpu_block, then we can inject the
timer interrupt immediately with __apic_accept_irq (if APICv is
disabled, it will set IRR and do kvm_make_request + kvm_vcpu_kick).

You can keep the module parameter, mostly for debugging reasons, but
please move it from kvm-intel to kvm, and add something like

diff --git a/kernel/sched/isolation.c b/kernel/sched/isolation.c
index 123ea07a3f3b..1cc7973c382e 100644
--- a/kernel/sched/isolation.c
+++ b/kernel/sched/isolation.c
@@ -14,6 +14,11 @@
 static cpumask_var_t housekeeping_mask;
 static unsigned int housekeeping_flags;

+bool housekeeping_enabled(enum hk_flags flags)
+{
+	return !!(housekeeping_flags & flags);
+}
+
 int housekeeping_any_cpu(enum hk_flags flags)
 {
 	if (static_branch_unlikely(&housekeeping_overridden))

so that the default for the module parameter can be
housekeeping_enabled(HK_FLAG_TIMER).

Thanks,

Paolo