Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 20 Sep 2022 00:42:27 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        mlevitsk@redhat.com, vkuznets@redhat.com
Subject: Re: [PATCH v2 2/9] KVM: x86: remove return value of kvm_vcpu_block
Message-ID: <YykMc2NDOzvG+xoy@google.com>
References: <20220811210605.402337-1-pbonzini@redhat.com>
 <20220811210605.402337-3-pbonzini@redhat.com>
 <Yvwpb6ofD1S+Rqk1@google.com>
 <78616cf8-2693-72cc-c2cc-5a849116ffc7@redhat.com>
 <Yv0aHXcmuivyJDXw@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Yv0aHXcmuivyJDXw@google.com>
Precedence: bulk

On Wed, Aug 17, 2022, Sean Christopherson wrote:
> Side topic, usage of kvm_apic_accept_events() appears to be broken (though nothing
> can trigger the bug).  If kvm_apic_accept_events() were to return an -errno, then
> kvm_arch_vcpu_ioctl_run() would return '0' to userspace without updating
> vcpu->run->exit_reason.  I think an easy fix is to drop the return value entirely
> and then WARN if kvm_check_nested_events() returns something other than -EBUSY.
> 
> 	if (is_guest_mode(vcpu)) {
> 		r = kvm_check_nested_events(vcpu);
> 		if (r < 0) {
> 			WARN_ON_ONCE(r != -EBUSY);
> 			return;
> 		}

For posterity, I was wrong.  Way down the stack, vmx_complete_nested_posted_interrupt()
can return -ENXIO after filling vcpu->run->exit_reason via kvm_handle_memory_failure().
That's the entire reason why negative values from kvm_check_nested_events() and
kvm_apic_accept_events() are morphed to '0', i.e. to "exit to userspace".