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;
From:   Vitaly Kuznetsov <vkuznets@redhat.com>
To:     Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
Cc:     pbonzini@redhat.com, sean.j.christopherson@intel.com,
        wanpengli@tencent.com, jmattson@google.com, joro@8bytes.org,
        tglx@linutronix.de, mingo@redhat.com, bp@alien8.de, hpa@zytor.com,
        x86@kernel.org, linux-kernel@vger.kernel.org, kvm@vger.kernel.org
Subject: Re: [PATCH] x86/kvm: do not setup pv tlb flush when not paravirtualized
In-Reply-To: <20200203101514.GG40679@calabresa>
References: <20200131155655.49812-1-cascardo@canonical.com> <87wo94ng9d.fsf@vitty.brq.redhat.com> <20200203101514.GG40679@calabresa>
Date:   Mon, 03 Feb 2020 13:42:14 +0100
Message-ID: <87r1zbona1.fsf@vitty.brq.redhat.com>
MIME-Version: 1.0
Content-Type: text/plain
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Thadeu Lima de Souza Cascardo <cascardo@canonical.com> writes:

> On Mon, Feb 03, 2020 at 10:59:10AM +0100, Vitaly Kuznetsov wrote:
>> Thadeu Lima de Souza Cascardo <cascardo@canonical.com> writes:
>> 
>> > kvm_setup_pv_tlb_flush will waste memory and print a misguiding message
>> > when KVM paravirtualization is not available.
>> >
>> > Intel SDM says that the when cpuid is used with EAX higher than the
>> > maximum supported value for basic of extended function, the data for the
>> > highest supported basic function will be returned.
>> >
>> > So, in some systems, kvm_arch_para_features will return bogus data,
>> > causing kvm_setup_pv_tlb_flush to detect support for pv tlb flush.
>> >
>> > Testing for kvm_para_available will work as it checks for the hypervisor
>> > signature.
>> >
>> > Besides, when the "nopv" command line parameter is used, it should not
>> > continue as well, as kvm_guest_init will no be called in that case.
>> >
>> > Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
>> > ---
>> >  arch/x86/kernel/kvm.c | 3 +++
>> >  1 file changed, 3 insertions(+)
>> >
>> > diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
>> > index 81045aabb6f4..d817f255aed8 100644
>> > --- a/arch/x86/kernel/kvm.c
>> > +++ b/arch/x86/kernel/kvm.c
>> > @@ -736,6 +736,9 @@ static __init int kvm_setup_pv_tlb_flush(void)
>> >  {
>> >  	int cpu;
>> >  
>> > +	if (!kvm_para_available() || nopv)
>> > +		return 0;
>> > +
>> >  	if (kvm_para_has_feature(KVM_FEATURE_PV_TLB_FLUSH) &&
>> >  	    !kvm_para_has_hint(KVM_HINTS_REALTIME) &&
>> >  	    kvm_para_has_feature(KVM_FEATURE_STEAL_TIME)) {
>> 
>> The patch will fix the immediate issue, but why kvm_setup_pv_tlb_flush()
>> is just an arch_initcall() which will be executed regardless of the fact
>> if we are running on KVM or not?
>> 
>> In Hyper-V we setup PV TLB flush from ms_hyperv_init_platform() -- which
>> only happens if Hyper-V platform was detected. Why don't we do it from
>> kvm_init_platform() in KVM?
>> 
>> -- 
>> Vitaly
>> 
>
> Because we can't call the allocator that early.
>
> Also, see the thread where this was "decided", the v6 of the original patch:
>
> https://lore.kernel.org/kvm/20171129162118.GA10661@flask/

Ok, I see, it's basically about what we prioritize: shorter boot time vs
smaller memory footprint. I'd personally vote for the former but the
opposite opinion is equally valid. Let's preserve the status quo.

-- 
Vitaly