From: Wanpeng Li <wanpengli@xxxxxxxxxxx>
After reverting commit 240c35a3783a (kvm: x86: Use task structs fpu field
for user), struct kvm_vcpu is 19456 bytes on my server, PAGE_ALLOC_COSTLY_ORDER(3)
is the order at which allocations are deemed costly to service. In serveless
scenario, one host can service hundreds/thoudands firecracker/kata-container
instances, howerver, new instance will fail to launch after memory is too
fragmented to allocate kvm_vcpu struct on host, this was observed in some
cloud provider product environments.
This patch dynamically allocates user_fpu, kvm_vcpu is 15168 bytes now on my
Skylake server.
Cc: Paolo Bonzini <pbonzini@xxxxxxxxxx>
Cc: Radim Krčmář <rkrcmar@xxxxxxxxxx>
Signed-off-by: Wanpeng Li <wanpengli@xxxxxxxxxxx>
---
arch/x86/include/asm/kvm_host.h | 2 +-
arch/x86/kvm/svm.c | 13 ++++++++++++-
arch/x86/kvm/vmx/vmx.c | 13 ++++++++++++-
arch/x86/kvm/x86.c | 4 ++--
4 files changed, 27 insertions(+), 5 deletions(-)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 4f938ac..7b0a4ee 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -616,7 +616,7 @@ struct kvm_vcpu_arch {
* "guest_fpu" state here contains the guest FPU context, with the
* host PRKU bits.
*/
- struct fpu user_fpu;
+ struct fpu *user_fpu;
struct fpu *guest_fpu;
u64 xcr0;
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 19f69df..7eafc69 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -2143,12 +2143,20 @@ static struct kvm_vcpu *svm_create_vcpu(struct kvm *kvm, unsigned int id)
goto out;
}
+ svm->vcpu.arch.user_fpu = kmem_cache_zalloc(x86_fpu_cache,
+ GFP_KERNEL_ACCOUNT);
+ if (!svm->vcpu.arch.user_fpu) {
+ printk(KERN_ERR "kvm: failed to allocate kvm userspace's fpu\n");
+ err = -ENOMEM;
+ goto free_partial_svm;
+ }
+
svm->vcpu.arch.guest_fpu = kmem_cache_zalloc(x86_fpu_cache,
GFP_KERNEL_ACCOUNT);
if (!svm->vcpu.arch.guest_fpu) {
printk(KERN_ERR "kvm: failed to allocate vcpu's fpu\n");
err = -ENOMEM;
- goto free_partial_svm;
+ goto free_user_fpu;
}
err = kvm_vcpu_init(&svm->vcpu, kvm, id);
@@ -2211,6 +2219,8 @@ static struct kvm_vcpu *svm_create_vcpu(struct kvm *kvm, unsigned int id)
kvm_vcpu_uninit(&svm->vcpu);
free_svm:
kmem_cache_free(x86_fpu_cache, svm->vcpu.arch.guest_fpu);
+free_user_fpu:
+ kmem_cache_free(x86_fpu_cache, svm->vcpu.arch.user_fpu);
free_partial_svm:
kmem_cache_free(kvm_vcpu_cache, svm);
out:
@@ -2241,6 +2251,7 @@ static void svm_free_vcpu(struct kvm_vcpu *vcpu)
__free_page(virt_to_page(svm->nested.hsave));
__free_pages(virt_to_page(svm->nested.msrpm), MSRPM_ALLOC_ORDER);
kvm_vcpu_uninit(vcpu);
+ kmem_cache_free(x86_fpu_cache, svm->vcpu.arch.user_fpu);
kmem_cache_free(x86_fpu_cache, svm->vcpu.arch.guest_fpu);
kmem_cache_free(kvm_vcpu_cache, svm);
}
diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index a279447..074385c 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -6598,6 +6598,7 @@ static void vmx_free_vcpu(struct kvm_vcpu *vcpu)
free_loaded_vmcs(vmx->loaded_vmcs);
kfree(vmx->guest_msrs);
kvm_vcpu_uninit(vcpu);
+ kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.user_fpu);
kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.guest_fpu);
kmem_cache_free(kvm_vcpu_cache, vmx);
}
@@ -6613,12 +6614,20 @@ static struct kvm_vcpu *vmx_create_vcpu(struct kvm *kvm, unsigned int id)
if (!vmx)
return ERR_PTR(-ENOMEM);
+ vmx->vcpu.arch.user_fpu = kmem_cache_zalloc(x86_fpu_cache,
+ GFP_KERNEL_ACCOUNT);
+ if (!vmx->vcpu.arch.user_fpu) {
+ printk(KERN_ERR "kvm: failed to allocate kvm userspace's fpu\n");
+ err = -ENOMEM;
+ goto free_partial_vcpu;
+ }
+
vmx->vcpu.arch.guest_fpu = kmem_cache_zalloc(x86_fpu_cache,
GFP_KERNEL_ACCOUNT);
if (!vmx->vcpu.arch.guest_fpu) {
printk(KERN_ERR "kvm: failed to allocate vcpu's fpu\n");
err = -ENOMEM;
- goto free_partial_vcpu;
+ goto free_user_fpu;
}
vmx->vpid = allocate_vpid();
@@ -6721,6 +6730,8 @@ static struct kvm_vcpu *vmx_create_vcpu(struct kvm *kvm, unsigned int id)
free_vcpu:
free_vpid(vmx->vpid);
kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.guest_fpu);
+free_user_fpu:
+ kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.user_fpu);
free_partial_vcpu:
kmem_cache_free(kvm_vcpu_cache, vmx);
return ERR_PTR(err);
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index bdcd250..09dbc93 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -8272,7 +8272,7 @@ static void kvm_load_guest_fpu(struct kvm_vcpu *vcpu)
{
fpregs_lock();
- copy_fpregs_to_fpstate(&vcpu->arch.user_fpu);
+ copy_fpregs_to_fpstate(vcpu->arch.user_fpu);
/* PKRU is separately restored in kvm_x86_ops->run. */
__copy_kernel_to_fpregs(&vcpu->arch.guest_fpu->state,
~XFEATURE_MASK_PKRU);
@@ -8289,7 +8289,7 @@ static void kvm_put_guest_fpu(struct kvm_vcpu *vcpu)
fpregs_lock();
copy_fpregs_to_fpstate(vcpu->arch.guest_fpu);
- copy_kernel_to_fpregs(&vcpu->arch.user_fpu.state);
+ copy_kernel_to_fpregs(&vcpu->arch.user_fpu->state);
fpregs_mark_activate();
fpregs_unlock();
--
2.7.4
