KVM paravirt: Add async PF initialization to PV guest.

Enable async PF in a guest if async PF capability is discovered. Acked-by: Rik van Riel <riel@redhat.com> Signed-off-by: Gleb Natapov <gleb@redhat.com> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

KVM paravirt: Add async PF initialization to PV guest.
Enable async PF in a guest if async PF capability is discovered. Acked-by: Rik van Riel <riel@redhat.com> Signed-off-by: Gleb Natapov <gleb@redhat.com> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Gleb Natapov · Avi Kivity
1 parent 344d9588a9
Showing 3 changed files with 101 additions and 0 deletions Side-by-side Diff
Documentation/kernel-parameters.txt
arch/x86/include/asm/kvm_para.h
arch/x86/kernel/kvm.c
@@ -1707,6 +1707,9 @@
  
 	no-kvmclock	[X86,KVM] Disable paravirtualized KVM clock driver
  
+	no-kvmapf	[X86,KVM] Disable paravirtualized asynchronous page
+			fault handling.
+
 	nolapic		[X86-32,APIC] Do not enable or use the local APIC.
  
 	nolapic_timer	[X86-32,APIC] Do not use the local APIC timer.
@@ -65,6 +65,12 @@
 	__u64 pt_phys;
 };
  
+struct kvm_vcpu_pv_apf_data {
+	__u32 reason;
+	__u8 pad[60];
+	__u32 enabled;
+};
+
 #ifdef __KERNEL__
 #include <asm/processor.h>
  
@@ -27,16 +27,30 @@
 #include <linux/mm.h>
 #include <linux/highmem.h>
 #include <linux/hardirq.h>
+#include <linux/notifier.h>
+#include <linux/reboot.h>
 #include <asm/timer.h>
+#include <asm/cpu.h>
  
 #define MMU_QUEUE_SIZE 1024
  
+static int kvmapf = 1;
+
+static int parse_no_kvmapf(char *arg)
+{
+        kvmapf = 0;
+        return 0;
+}
+
+early_param("no-kvmapf", parse_no_kvmapf);
+
 struct kvm_para_state {
 	u8 mmu_queue[MMU_QUEUE_SIZE];
 	int mmu_queue_len;
 };
  
 static DEFINE_PER_CPU(struct kvm_para_state, para_state);
+static DEFINE_PER_CPU(struct kvm_vcpu_pv_apf_data, apf_reason) __aligned(64);
  
 static struct kvm_para_state *kvm_para_state(void)
 {
  
  
@@ -231,12 +245,86 @@
 #endif
 }
  
+void __cpuinit kvm_guest_cpu_init(void)
+{
+	if (!kvm_para_available())
+		return;
+
+	if (kvm_para_has_feature(KVM_FEATURE_ASYNC_PF) && kvmapf) {
+		u64 pa = __pa(&__get_cpu_var(apf_reason));
+
+		wrmsrl(MSR_KVM_ASYNC_PF_EN, pa | KVM_ASYNC_PF_ENABLED);
+		__get_cpu_var(apf_reason).enabled = 1;
+		printk(KERN_INFO"KVM setup async PF for cpu %d\n",
+		       smp_processor_id());
+	}
+}
+
+static void kvm_pv_disable_apf(void *unused)
+{
+	if (!__get_cpu_var(apf_reason).enabled)
+		return;
+
+	wrmsrl(MSR_KVM_ASYNC_PF_EN, 0);
+	__get_cpu_var(apf_reason).enabled = 0;
+
+	printk(KERN_INFO"Unregister pv shared memory for cpu %d\n",
+	       smp_processor_id());
+}
+
+static int kvm_pv_reboot_notify(struct notifier_block *nb,
+				unsigned long code, void *unused)
+{
+	if (code == SYS_RESTART)
+		on_each_cpu(kvm_pv_disable_apf, NULL, 1);
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block kvm_pv_reboot_nb = {
+	.notifier_call = kvm_pv_reboot_notify,
+};
+
 #ifdef CONFIG_SMP
 static void __init kvm_smp_prepare_boot_cpu(void)
 {
 	WARN_ON(kvm_register_clock("primary cpu clock"));
+	kvm_guest_cpu_init();
 	native_smp_prepare_boot_cpu();
 }
+
+static void kvm_guest_cpu_online(void *dummy)
+{
+	kvm_guest_cpu_init();
+}
+
+static void kvm_guest_cpu_offline(void *dummy)
+{
+	kvm_pv_disable_apf(NULL);
+}
+
+static int __cpuinit kvm_cpu_notify(struct notifier_block *self,
+				    unsigned long action, void *hcpu)
+{
+	int cpu = (unsigned long)hcpu;
+	switch (action) {
+	case CPU_ONLINE:
+	case CPU_DOWN_FAILED:
+	case CPU_ONLINE_FROZEN:
+		smp_call_function_single(cpu, kvm_guest_cpu_online, NULL, 0);
+		break;
+	case CPU_DOWN_PREPARE:
+	case CPU_DOWN_PREPARE_FROZEN:
+		smp_call_function_single(cpu, kvm_guest_cpu_offline, NULL, 1);
+		break;
+	default:
+		break;
+	}
+	return NOTIFY_OK;
+}
+
+static struct notifier_block __cpuinitdata kvm_cpu_notifier = {
+        .notifier_call  = kvm_cpu_notify,
+};
 #endif
  
 void __init kvm_guest_init(void)
  
@@ -245,8 +333,12 @@
 		return;
  
 	paravirt_ops_setup();
+	register_reboot_notifier(&kvm_pv_reboot_nb);
 #ifdef CONFIG_SMP
 	smp_ops.smp_prepare_boot_cpu = kvm_smp_prepare_boot_cpu;
+	register_cpu_notifier(&kvm_cpu_notifier);
+#else
+	kvm_guest_cpu_init();
 #endif
 }
...	...	@@ -1707,6 +1707,9 @@
1707	1707
1708	1708	no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver
1709	1709
	1710	+ no-kvmapf [X86,KVM] Disable paravirtualized asynchronous page
	1711	+ fault handling.
	1712	+
1710	1713	nolapic [X86-32,APIC] Do not enable or use the local APIC.
1711	1714
1712	1715	nolapic_timer [X86-32,APIC] Do not use the local APIC timer.
...	...	@@ -65,6 +65,12 @@
65	65	__u64 pt_phys;
66	66	};
67	67
	68	+struct kvm_vcpu_pv_apf_data {
	69	+ __u32 reason;
	70	+ __u8 pad[60];
	71	+ __u32 enabled;
	72	+};
	73	+
68	74	#ifdef __KERNEL__
69	75	#include <asm/processor.h>
70	76
...	...	@@ -27,16 +27,30 @@
27	27	#include <linux/mm.h>
28	28	#include <linux/highmem.h>
29	29	#include <linux/hardirq.h>
	30	+#include <linux/notifier.h>
	31	+#include <linux/reboot.h>
30	32	#include <asm/timer.h>
	33	+#include <asm/cpu.h>
31	34
32	35	#define MMU_QUEUE_SIZE 1024
33	36
	37	+static int kvmapf = 1;
	38	+
	39	+static int parse_no_kvmapf(char *arg)
	40	+{
	41	+ kvmapf = 0;
	42	+ return 0;
	43	+}
	44	+
	45	+early_param("no-kvmapf", parse_no_kvmapf);
	46	+
34	47	struct kvm_para_state {
35	48	u8 mmu_queue[MMU_QUEUE_SIZE];
36	49	int mmu_queue_len;
37	50	};
38	51
39	52	static DEFINE_PER_CPU(struct kvm_para_state, para_state);
	53	+static DEFINE_PER_CPU(struct kvm_vcpu_pv_apf_data, apf_reason) __aligned(64);
40	54
41	55	static struct kvm_para_state *kvm_para_state(void)
42	56	{
43	57
44	58
...	...	@@ -231,12 +245,86 @@
231	245	#endif
232	246	}
233	247
	248	+void __cpuinit kvm_guest_cpu_init(void)
	249	+{
	250	+ if (!kvm_para_available())
	251	+ return;
	252	+
	253	+ if (kvm_para_has_feature(KVM_FEATURE_ASYNC_PF) && kvmapf) {
	254	+ u64 pa = __pa(&__get_cpu_var(apf_reason));
	255	+
	256	+ wrmsrl(MSR_KVM_ASYNC_PF_EN, pa \| KVM_ASYNC_PF_ENABLED);
	257	+ __get_cpu_var(apf_reason).enabled = 1;
	258	+ printk(KERN_INFO"KVM setup async PF for cpu %d\n",
	259	+ smp_processor_id());
	260	+ }
	261	+}
	262	+
	263	+static void kvm_pv_disable_apf(void *unused)
	264	+{
	265	+ if (!__get_cpu_var(apf_reason).enabled)
	266	+ return;
	267	+
	268	+ wrmsrl(MSR_KVM_ASYNC_PF_EN, 0);
	269	+ __get_cpu_var(apf_reason).enabled = 0;
	270	+
	271	+ printk(KERN_INFO"Unregister pv shared memory for cpu %d\n",
	272	+ smp_processor_id());
	273	+}
	274	+
	275	+static int kvm_pv_reboot_notify(struct notifier_block *nb,
	276	+ unsigned long code, void *unused)
	277	+{
	278	+ if (code == SYS_RESTART)
	279	+ on_each_cpu(kvm_pv_disable_apf, NULL, 1);
	280	+ return NOTIFY_DONE;
	281	+}
	282	+
	283	+static struct notifier_block kvm_pv_reboot_nb = {
	284	+ .notifier_call = kvm_pv_reboot_notify,
	285	+};
	286	+
234	287	#ifdef CONFIG_SMP
235	288	static void __init kvm_smp_prepare_boot_cpu(void)
236	289	{
237	290	WARN_ON(kvm_register_clock("primary cpu clock"));
	291	+ kvm_guest_cpu_init();
238	292	native_smp_prepare_boot_cpu();
239	293	}
	294	+
	295	+static void kvm_guest_cpu_online(void *dummy)
	296	+{
	297	+ kvm_guest_cpu_init();
	298	+}
	299	+
	300	+static void kvm_guest_cpu_offline(void *dummy)
	301	+{
	302	+ kvm_pv_disable_apf(NULL);
	303	+}
	304	+
	305	+static int __cpuinit kvm_cpu_notify(struct notifier_block *self,
	306	+ unsigned long action, void *hcpu)
	307	+{
	308	+ int cpu = (unsigned long)hcpu;
	309	+ switch (action) {
	310	+ case CPU_ONLINE:
	311	+ case CPU_DOWN_FAILED:
	312	+ case CPU_ONLINE_FROZEN:
	313	+ smp_call_function_single(cpu, kvm_guest_cpu_online, NULL, 0);
	314	+ break;
	315	+ case CPU_DOWN_PREPARE:
	316	+ case CPU_DOWN_PREPARE_FROZEN:
	317	+ smp_call_function_single(cpu, kvm_guest_cpu_offline, NULL, 1);
	318	+ break;
	319	+ default:
	320	+ break;
	321	+ }
	322	+ return NOTIFY_OK;
	323	+}
	324	+
	325	+static struct notifier_block __cpuinitdata kvm_cpu_notifier = {
	326	+ .notifier_call = kvm_cpu_notify,
	327	+};
240	328	#endif
241	329
242	330	void __init kvm_guest_init(void)
243	331
...	...	@@ -245,8 +333,12 @@
245	333	return;
246	334
247	335	paravirt_ops_setup();
	336	+ register_reboot_notifier(&kvm_pv_reboot_nb);
248	337	#ifdef CONFIG_SMP
249	338	smp_ops.smp_prepare_boot_cpu = kvm_smp_prepare_boot_cpu;
	339	+ register_cpu_notifier(&kvm_cpu_notifier);
	340	+#else
	341	+ kvm_guest_cpu_init();
250	342	#endif
251	343	}