Commit 15ad7146 authored by Avi Kivity's avatar Avi Kivity
Browse files

KVM: Use the scheduler preemption notifiers to make kvm preemptible 



Current kvm disables preemption while the new virtualization registers are
in use.  This of course is not very good for latency sensitive workloads (one
use of virtualization is to offload user interface and other latency
insensitive stuff to a container, so that it is easier to analyze the
remaining workload).  This patch re-enables preemption for kvm; preemption
is now only disabled when switching the registers in and out, and during
the switch to guest mode and back.

Contains fixes from Shaohua Li <shaohua.li@intel.com>.

Signed-off-by: default avatarAvi Kivity <avi@qumranet.com>
parent 519ef353

drivers/kvm/Kconfig

+1 −0
Original line number Diff line number Diff line
@@ -17,6 +17,7 @@ if VIRTUALIZATION 
config KVM

	tristate "Kernel-based Virtual Machine (KVM) support"

	depends on X86 && EXPERIMENTAL

	select PREEMPT_NOTIFIERS

	select ANON_INODES

	---help---

	  Support hosting fully virtualized guest machines using hardware

drivers/kvm/kvm.h

+3 −1
Original line number Diff line number Diff line
@@ -13,6 +13,7 @@ 
#include <linux/signal.h>

#include <linux/sched.h>

#include <linux/mm.h>

#include <linux/preempt.h>

#include <asm/signal.h>



#include <linux/kvm.h>
@@ -301,6 +302,7 @@ void kvm_io_bus_register_dev(struct kvm_io_bus *bus, 


struct kvm_vcpu {

	struct kvm *kvm;

	struct preempt_notifier preempt_notifier;

	int vcpu_id;

	struct mutex mutex;

	int   cpu;
@@ -429,7 +431,7 @@ struct kvm_arch_ops { 
	struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);

	void (*vcpu_free)(struct kvm_vcpu *vcpu);



	void (*vcpu_load)(struct kvm_vcpu *vcpu);

	void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);

	void (*vcpu_put)(struct kvm_vcpu *vcpu);

	void (*vcpu_decache)(struct kvm_vcpu *vcpu);

drivers/kvm/kvm_main.c

+37 −6
Original line number Diff line number Diff line
@@ -54,6 +54,8 @@ static cpumask_t cpus_hardware_enabled; 


struct kvm_arch_ops *kvm_arch_ops;



static __read_mostly struct preempt_ops kvm_preempt_ops;



#define STAT_OFFSET(x) offsetof(struct kvm_vcpu, stat.x)



static struct kvm_stats_debugfs_item {
@@ -239,13 +241,21 @@ EXPORT_SYMBOL_GPL(kvm_put_guest_fpu); 
 */

static void vcpu_load(struct kvm_vcpu *vcpu)

{

	int cpu;



	mutex_lock(&vcpu->mutex);

	kvm_arch_ops->vcpu_load(vcpu);

	cpu = get_cpu();

	preempt_notifier_register(&vcpu->preempt_notifier);

	kvm_arch_ops->vcpu_load(vcpu, cpu);

	put_cpu();

}



static void vcpu_put(struct kvm_vcpu *vcpu)

{

	preempt_disable();

	kvm_arch_ops->vcpu_put(vcpu);

	preempt_notifier_unregister(&vcpu->preempt_notifier);

	preempt_enable();

	mutex_unlock(&vcpu->mutex);

}
@@ -1672,9 +1682,7 @@ void kvm_resched(struct kvm_vcpu *vcpu) 
{

	if (!need_resched())

		return;

	vcpu_put(vcpu);

	cond_resched();

	vcpu_load(vcpu);

}

EXPORT_SYMBOL_GPL(kvm_resched);
@@ -1722,11 +1730,9 @@ static int pio_copy_data(struct kvm_vcpu *vcpu) 
	unsigned bytes;

	int nr_pages = vcpu->pio.guest_pages[1] ? 2 : 1;



	kvm_arch_ops->vcpu_put(vcpu);

	q = vmap(vcpu->pio.guest_pages, nr_pages, VM_READ|VM_WRITE,

		 PAGE_KERNEL);

	if (!q) {

		kvm_arch_ops->vcpu_load(vcpu);

		free_pio_guest_pages(vcpu);

		return -ENOMEM;

	}
@@ -1738,7 +1744,6 @@ static int pio_copy_data(struct kvm_vcpu *vcpu) 
		memcpy(p, q, bytes);

	q -= vcpu->pio.guest_page_offset;

	vunmap(q);

	kvm_arch_ops->vcpu_load(vcpu);

	free_pio_guest_pages(vcpu);

	return 0;

}
@@ -2413,6 +2418,8 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n) 
	if (IS_ERR(vcpu))

		return PTR_ERR(vcpu);



	preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);



	vcpu_load(vcpu);

	r = kvm_mmu_setup(vcpu);

	vcpu_put(vcpu);
@@ -3145,6 +3152,27 @@ static struct sys_device kvm_sysdev = { 


hpa_t bad_page_address;



static inline

struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)

{

	return container_of(pn, struct kvm_vcpu, preempt_notifier);

}



static void kvm_sched_in(struct preempt_notifier *pn, int cpu)

{

	struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);



	kvm_arch_ops->vcpu_load(vcpu, cpu);

}



static void kvm_sched_out(struct preempt_notifier *pn,

			  struct task_struct *next)

{

	struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);



	kvm_arch_ops->vcpu_put(vcpu);

}



int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module)

{

	int r;
@@ -3191,6 +3219,9 @@ int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module) 
		goto out_free;

	}



	kvm_preempt_ops.sched_in = kvm_sched_in;

	kvm_preempt_ops.sched_out = kvm_sched_out;



	return r;



out_free:

drivers/kvm/mmu.c

+0 −2
Original line number Diff line number Diff line
@@ -276,9 +276,7 @@ static int mmu_topup_memory_caches(struct kvm_vcpu *vcpu) 
	kvm_mmu_free_some_pages(vcpu);

	if (r < 0) {

		spin_unlock(&vcpu->kvm->lock);

		kvm_arch_ops->vcpu_put(vcpu);

		r = __mmu_topup_memory_caches(vcpu, GFP_KERNEL);

		kvm_arch_ops->vcpu_load(vcpu);

		spin_lock(&vcpu->kvm->lock);

		kvm_mmu_free_some_pages(vcpu);

	}

drivers/kvm/svm.c

+2 −4
Original line number Diff line number Diff line
@@ -625,12 +625,11 @@ static void svm_free_vcpu(struct kvm_vcpu *vcpu) 
	kfree(svm);

}



static void svm_vcpu_load(struct kvm_vcpu *vcpu)

static void svm_vcpu_load(struct kvm_vcpu *vcpu, int cpu)

{

	struct vcpu_svm *svm = to_svm(vcpu);

	int cpu, i;

	int i;



	cpu = get_cpu();

	if (unlikely(cpu != vcpu->cpu)) {

		u64 tsc_this, delta;
@@ -657,7 +656,6 @@ static void svm_vcpu_put(struct kvm_vcpu *vcpu) 
		wrmsrl(host_save_user_msrs[i], svm->host_user_msrs[i]);



	rdtscll(vcpu->host_tsc);

	put_cpu();

}



static void svm_vcpu_decache(struct kvm_vcpu *vcpu)

CRA Git | Maintained and supported by SUSTech CRA and CCSE