Re: [PATCH v3 4/4] x86/kvm: add boot parameter for setting max number of vcpus per guest

[Juergen Gross <jgross@suse.com>]

[-- Attachment #1.1.1: Type: text/plain, Size: 7670 bytes --]

On 17.11.21 21:57, Sean Christopherson wrote:
> On Tue, Nov 16, 2021, Juergen Gross wrote:
>> Today the maximum number of vcpus of a kvm guest is set via a #define
>> in a header file.
>>
>> In order to support higher vcpu numbers for guests without generally
>> increasing the memory consumption of guests on the host especially on
>> very large systems add a boot parameter for specifying the number of
>> allowed vcpus for guests.
>>
>> The default will still be the current setting of 1024. The value 0 has
>> the special meaning to limit the number of possible vcpus to the
>> number of possible cpus of the host.
>>
>> Signed-off-by: Juergen Gross <jgross@suse.com>
>> ---
>> V3:
>> - rebase
>> ---
>>   Documentation/admin-guide/kernel-parameters.txt | 7 +++++++
>>   arch/x86/include/asm/kvm_host.h                 | 5 ++++-
>>   arch/x86/kvm/x86.c                              | 9 ++++++++-
>>   3 files changed, 19 insertions(+), 2 deletions(-)
>>
>> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
>> index e269c3f66ba4..409a72c2d91b 100644
>> --- a/Documentation/admin-guide/kernel-parameters.txt
>> +++ b/Documentation/admin-guide/kernel-parameters.txt
>> @@ -2445,6 +2445,13 @@
>>   			feature (tagged TLBs) on capable Intel chips.
>>   			Default is 1 (enabled)
>>   
>> +	kvm.max_vcpus=	[KVM,X86] Set the maximum allowed numbers of vcpus per
>> +			guest. The special value 0 sets the limit to the number
>> +			of physical cpus possible on the host (including not
>> +			yet hotplugged cpus). Higher values will result in
>> +			slightly higher memory consumption per guest.
>> +			Default: 1024
> 
> Rather than makes this a module param, I would prefer to start with the below
> patch (originally from TDX pre-enabling) and then wire up a way for userspace to
> _lower_ the max on a per-VM basis, e.g. add a capability.
> 
> VMs largely fall into two categories: (1) the max number of vCPUs is known prior
> to VM creation, or (2) the max number of vCPUs is unbounded (up to KVM's hard
> limit), e.g. for container-style use cases where "vCPUs" are created on-demand in
> response to the "guest" creating a new task.
> 
> For #1, a per-VM control lets userspace lower the limit to the bare minimum.  For
> #2, neither the module param nor the per-VM control is likely to be useful, but
> a per-VM control does let mixed environments (both #1 and #2 VMs) lower the limits
> for compatible VMs, whereas a module param must be set to the max of any potential VM.

The main reason for this whole series is a request by a partner
to enable huge VMs on huge machines (huge meaning thousands of
vcpus on thousands of physical cpus).

Making this large number a compile time setting would hurt all
the users who have more standard requirements by allocating the
needed resources even on small systems, so I've switched to a boot
parameter in order to enable those huge numbers only when required.

With Marc's series to use an xarray for the vcpu pointers only the
bitmaps for sending IRQs to vcpus are left which need to be sized
according to the max vcpu limit. Your patch below seems to be fine, but
doesn't help for that case.


Juergen

> 
>  From 0593cb4f73a6c3f0862f9411f0e14f00671f59ae Mon Sep 17 00:00:00 2001
> From: Sean Christopherson <sean.j.christopherson@intel.com>
> Date: Fri, 2 Jul 2021 15:04:27 -0700
> Subject: [PATCH] KVM: Add max_vcpus field in common 'struct kvm'
> 
> Move arm's per-VM max_vcpus field into the generic "struct kvm", and use
> it to check vcpus_created in the generic code instead of checking only
> the hardcoded absolute KVM-wide max.  x86 TDX guests will reuse the
> generic check verbatim, as the max number of vCPUs for a TDX guest is
> user defined at VM creation and immutable thereafter.
> 
> Signed-off-by: Sean Christopherson <seanjc@google.com>
> ---
>   arch/arm64/include/asm/kvm_host.h | 3 ---
>   arch/arm64/kvm/arm.c              | 7 ++-----
>   arch/arm64/kvm/vgic/vgic-init.c   | 6 +++---
>   include/linux/kvm_host.h          | 1 +
>   virt/kvm/kvm_main.c               | 3 ++-
>   5 files changed, 8 insertions(+), 12 deletions(-)
> 
> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> index 4be8486042a7..b51e1aa6ae27 100644
> --- a/arch/arm64/include/asm/kvm_host.h
> +++ b/arch/arm64/include/asm/kvm_host.h
> @@ -108,9 +108,6 @@ struct kvm_arch {
>   	/* VTCR_EL2 value for this VM */
>   	u64    vtcr;
> 
> -	/* The maximum number of vCPUs depends on the used GIC model */
> -	int max_vcpus;
> -
>   	/* Interrupt controller */
>   	struct vgic_dist	vgic;
> 
> diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
> index f5490afe1ebf..97c3b83235b4 100644
> --- a/arch/arm64/kvm/arm.c
> +++ b/arch/arm64/kvm/arm.c
> @@ -153,7 +153,7 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
>   	kvm_vgic_early_init(kvm);
> 
>   	/* The maximum number of VCPUs is limited by the host's GIC model */
> -	kvm->arch.max_vcpus = kvm_arm_default_max_vcpus();
> +	kvm->max_vcpus = kvm_arm_default_max_vcpus();
> 
>   	set_default_spectre(kvm);
> 
> @@ -228,7 +228,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
>   	case KVM_CAP_MAX_VCPUS:
>   	case KVM_CAP_MAX_VCPU_ID:
>   		if (kvm)
> -			r = kvm->arch.max_vcpus;
> +			r = kvm->max_vcpus;
>   		else
>   			r = kvm_arm_default_max_vcpus();
>   		break;
> @@ -304,9 +304,6 @@ int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id)
>   	if (irqchip_in_kernel(kvm) && vgic_initialized(kvm))
>   		return -EBUSY;
> 
> -	if (id >= kvm->arch.max_vcpus)
> -		return -EINVAL;
> -
>   	return 0;
>   }
> 
> diff --git a/arch/arm64/kvm/vgic/vgic-init.c b/arch/arm64/kvm/vgic/vgic-init.c
> index 0a06d0648970..906aee52f2bc 100644
> --- a/arch/arm64/kvm/vgic/vgic-init.c
> +++ b/arch/arm64/kvm/vgic/vgic-init.c
> @@ -97,11 +97,11 @@ int kvm_vgic_create(struct kvm *kvm, u32 type)
>   	ret = 0;
> 
>   	if (type == KVM_DEV_TYPE_ARM_VGIC_V2)
> -		kvm->arch.max_vcpus = VGIC_V2_MAX_CPUS;
> +		kvm->max_vcpus = VGIC_V2_MAX_CPUS;
>   	else
> -		kvm->arch.max_vcpus = VGIC_V3_MAX_CPUS;
> +		kvm->max_vcpus = VGIC_V3_MAX_CPUS;
> 
> -	if (atomic_read(&kvm->online_vcpus) > kvm->arch.max_vcpus) {
> +	if (atomic_read(&kvm->online_vcpus) > kvm->max_vcpus) {
>   		ret = -E2BIG;
>   		goto out_unlock;
>   	}
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index 60a35d9fe259..5f56516e2f5a 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -566,6 +566,7 @@ struct kvm {
>   	 * and is accessed atomically.
>   	 */
>   	atomic_t online_vcpus;
> +	int max_vcpus;
>   	int created_vcpus;
>   	int last_boosted_vcpu;
>   	struct list_head vm_list;
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index 3f6d450355f0..e509b963651c 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -1052,6 +1052,7 @@ static struct kvm *kvm_create_vm(unsigned long type)
>   	rcuwait_init(&kvm->mn_memslots_update_rcuwait);
> 
>   	INIT_LIST_HEAD(&kvm->devices);
> +	kvm->max_vcpus = KVM_MAX_VCPUS;
> 
>   	BUILD_BUG_ON(KVM_MEM_SLOTS_NUM > SHRT_MAX);
> 
> @@ -3599,7 +3600,7 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
>   		return -EINVAL;
> 
>   	mutex_lock(&kvm->lock);
> -	if (kvm->created_vcpus == KVM_MAX_VCPUS) {
> +	if (kvm->created_vcpus >= kvm->max_vcpus) {
>   		mutex_unlock(&kvm->lock);
>   		return -EINVAL;
>   	}
> --
> 2.34.0.rc1.387.gb447b232ab-goog
> 


[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 3135 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 495 bytes --]