Re: [PATCH v6 5/5] x86/kvm: Avoid dynamic allocation of pvclock data when SEV is active

From: Sean Christopherson <sean.j.christopherson@intel.com>
To: Brijesh Singh <brijesh.singh@amd.com>, Borislav Petkov <bp@suse.de>
Cc: x86@kernel.org, linux-kernel@vger.kernel.org,
	kvm@vger.kernel.org, "Tom Lendacky" <thomas.lendacky@amd.com>,
	"Thomas Gleixner" <tglx@linutronix.de>,
	"H. Peter Anvin" <hpa@zytor.com>,
	"Paolo Bonzini" <pbonzini@redhat.com>,
	"Radim Krčmář" <rkrcmar@redhat.com>
Subject: Re: [PATCH v6 5/5] x86/kvm: Avoid dynamic allocation of pvclock data when SEV is active
Date: Mon, 10 Sep 2018 08:28:17 -0700	[thread overview]
Message-ID: <1536593297.11460.72.camel@intel.com> (raw)
In-Reply-To: <097eb5f5-2cd9-8b08-32c5-d90c8e0cbb6d@amd.com>

On Mon, 2018-09-10 at 10:10 -0500, Brijesh Singh wrote:
> 
> On 09/10/2018 08:29 AM, Sean Christopherson wrote:
> ...
> 
> > > > > + */
> > > > > +static struct pvclock_vsyscall_time_info
> > > > > +			hv_clock_aux[NR_CPUS] __decrypted_aux;
> > > > Hmm, so worst case that's 64 4K pages:
> > > > 
> > > > (8192*32)/4096 = 64 4K pages.
> > > We can minimize the worst case memory usage. The number of VCPUs
> > > supported by KVM maybe less than NR_CPUS. e.g Currently KVM_MAX_VCPUS is
> > > set to 288
> > KVM_MAX_VCPUS is a property of the host, whereas this code runs in the
> > guest, e.g. KVM_MAX_VCPUS could be 2048 in the host for all we know.
> > 
> 
> IIRC, during guest creation time qemu will check the host supported
> VCPUS count. If count is greater than KVM_MAX_VCPUS then it will
> fail to launch guest (or fail to hot plug vcpus). In other words, the
> number of vcpus in a KVM guest will never to > KVM_MAX_VCPUS.
> 
> Am I missing something ?

KVM_MAX_VCPUS is a definition for use in the *host*, it's even defined
in kvm_host.h.  The guest's pvclock code won't get magically recompiled
if KVM_MAX_VCPUS is changed in the host.  KVM_MAX_VCPUS is an arbitrary
value in the sense that there isn't a fundamental hard limit, i.e. the
value can be changed, either for a custom KVM build or in mainline,
e.g. it was bumped in 2016:

commit 682f732ecf7396e9d6fe24d44738966699fae6c0
Author: Radim Krčmář <rkrcmar@redhat.com>
Date:   Tue Jul 12 22:09:29 2016 +0200

    KVM: x86: bump MAX_VCPUS to 288
    
    288 is in high demand because of Knights Landing CPU.
    We cannot set the limit to 640k, because that would be wasting space.
    
    Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
    Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
    Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 074b5c760327..21a40dc7aad6 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -34,7 +34,7 @@
 #include <asm/asm.h>
 #include <asm/kvm_page_track.h>
 
-#define KVM_MAX_VCPUS 255
+#define KVM_MAX_VCPUS 288
 #define KVM_SOFT_MAX_VCPUS 240
 #define KVM_USER_MEM_SLOTS 509
 /* memory slots that are not exposed to userspace */
