[PATCH v4 5/9] kvm: Add host pmu to support VM introspection

[Punit Agrawal]

Add a software PMU on the host that supports tracking VM operations (in
the form of PMU events). To add support for events, users of this
functionality will provide the callbacks to configure the VM to
enable/disable the events and read a count of the event's frequency.

The host PMU events can be controlled by tools like perf using standard
kernel perf interfaces.

Signed-off-by: Punit Agrawal <punit.agrawal at arm.com>
Cc: Christoffer Dall <christoffer.dall at linaro.org>
Cc: Marc Zyngier <marc.zyngier at arm.com>
Cc: Will Deacon <will.deacon at arm.com>
Cc: Paolo Bonzini <pbonzini at redhat.com>
Cc: "Radim Krčmář" <rkrcmar at redhat.com>
Cc: Mark Rutland <mark.rutland at arm.com>
---
 include/kvm/host_pmu.h |  35 +++++++
 virt/kvm/Kconfig       |   3 +
 virt/kvm/host_pmu.c    | 246 +++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 284 insertions(+)
 create mode 100644 include/kvm/host_pmu.h
 create mode 100644 virt/kvm/host_pmu.c

diff --git a/include/kvm/host_pmu.h b/include/kvm/host_pmu.h
new file mode 100644
index 000000000000..f9dc21de8886
--- /dev/null
+++ b/include/kvm/host_pmu.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright (C) 2017 ARM Ltd.
+ * Author: Punit Agrawal <punit.agrawal at arm.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef _KVM_HOST_PMU_H
+#define _KVM_HOST_PMU_H
+
+#include <linux/kvm_host.h>
+#include <linux/perf_event.h>
+
+typedef void (*configure_event_fn)(struct kvm_vcpu *vcpu, bool enable);
+typedef u64 (*get_event_count_fn)(struct kvm_vcpu *vcpu);
+
+struct kvm_event_cb {
+	get_event_count_fn get_event_count;
+	configure_event_fn configure_event;
+};
+
+int kvm_host_pmu_register(int nr_events, struct kvm_event_cb *cbs,
+			  struct attribute **event_attrs);
+void kvm_host_pmu_unregister(void);
+
+#endif /* _KVM_HOST_PMU_H */
diff --git a/virt/kvm/Kconfig b/virt/kvm/Kconfig
index b0cc1a34db27..3fe27172e757 100644
--- a/virt/kvm/Kconfig
+++ b/virt/kvm/Kconfig
@@ -50,3 +50,6 @@ config KVM_COMPAT
 
 config HAVE_KVM_IRQ_BYPASS
        bool
+
+config KVM_HOST_PMU
+	bool
\ No newline at end of file
diff --git a/virt/kvm/host_pmu.c b/virt/kvm/host_pmu.c
new file mode 100644
index 000000000000..ae7c0c3aed92
--- /dev/null
+++ b/virt/kvm/host_pmu.c
@@ -0,0 +1,246 @@
+/*
+ * Copyright (C) 2017 ARM Ltd.
+ * Author: Punit Agrawal <punit.agrawal at arm.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#include <kvm/host_pmu.h>
+#include <linux/device.h>
+#include <linux/kernel.h>
+#include <linux/kvm_host.h>
+#include <linux/list.h>
+#include <linux/perf_event.h>
+#include <linux/pid.h>
+#include <linux/slab.h>
+#include <linux/spinlock_types.h>
+#include <linux/sysfs.h>
+
+struct host_pmu {
+	int nr_events;
+	struct pmu pmu;
+	struct kvm_event_cb *cbs;
+	spinlock_t event_list_lock;
+	struct list_head event_list_head;
+} host_pmu;
+#define to_host_pmu(p) (container_of(p, struct host_pmu, pmu))
+
+struct event_data {
+	int event_id;
+	struct kvm_vcpu *vcpu;
+	struct kvm_event_cb *cb;
+	struct list_head event_list;
+};
+
+static struct attribute_group events_attr_group = {
+	.name	= "events",
+};
+
+
+#define EVENT_MASK	GENMASK_ULL(0, 7)
+#define to_event(cfg)	((cfg) & EVENT_MASK)
+
+PMU_FORMAT_ATTR(event, "config:0-7");
+
+static struct attribute *format_attrs[] = {
+	&format_attr_event.attr,
+	NULL,
+};
+
+static struct attribute_group format_attr_group = {
+	.name	= "format",
+	.attrs	= format_attrs,
+};
+
+static const struct attribute_group *attr_groups[] = {
+	&events_attr_group,
+	&format_attr_group,
+	NULL,
+};
+
+static void host_event_destroy(struct perf_event *event)
+{
+	struct host_pmu *host_pmu = to_host_pmu(event->pmu);
+	struct event_data *e_data = event->pmu_private;
+
+	spin_lock(&host_pmu->event_list_lock);
+	list_del(&e_data->event_list);
+	spin_unlock(&host_pmu->event_list_lock);
+	kfree(e_data);
+}
+
+static int host_event_init(struct perf_event *event)
+{
+	struct host_pmu *host_pmu = to_host_pmu(event->pmu);
+	int event_id = to_event(event->attr.config);
+	struct event_data *e_data, *pos;
+	struct kvm_vcpu *vcpu;
+	bool found = false;
+	struct pid *pid;
+	struct kvm *kvm;
+	int i, ret = 0;
+
+	if (event->attr.type != event->pmu->type)
+		return -ENOENT;
+
+	if (has_branch_stack(event)	||
+	    is_sampling_event(event)	||
+	    event->attr.exclude_user	||
+	    event->attr.exclude_kernel	||
+	    event->attr.exclude_hv	||
+	    event->attr.exclude_idle	||
+	    event->attr.exclude_guest) {
+		return -EINVAL;
+	}
+
+	if (event->attach_state != PERF_ATTACH_TASK)
+		return -EOPNOTSUPP;
+
+	if (event_id >= host_pmu->nr_events)
+		return -EINVAL;
+
+	pid = get_task_pid(event->hw.target, PIDTYPE_PID);
+	spin_lock(&kvm_lock);
+	list_for_each_entry(kvm, &vm_list, vm_list) {
+		kvm_for_each_vcpu(i, vcpu, kvm) {
+			if (vcpu->pid == pid) {
+				found = true;
+				goto found;
+			}
+		}
+	}
+found:
+	spin_unlock(&kvm_lock);
+	put_pid(pid);
+
+	if (!found)
+		return -EINVAL;
+
+	spin_lock(&host_pmu->event_list_lock);
+	/* Make sure we don't already have the (event_id, kvm) pair */
+	list_for_each_entry(pos, &host_pmu->event_list_head, event_list) {
+		if (pos->event_id == event_id &&
+		    pos->vcpu->pid == pid) {
+			ret = -EOPNOTSUPP;
+			goto unlock;
+		}
+	}
+
+	e_data = kzalloc(sizeof(*e_data), GFP_KERNEL);
+	e_data->event_id = event_id;
+	e_data->vcpu = vcpu;
+	e_data->cb = &host_pmu->cbs[event_id];
+
+	event->pmu_private = e_data;
+	event->destroy = host_event_destroy;
+
+	list_add_tail(&e_data->event_list, &host_pmu->event_list_head);
+
+unlock:
+	spin_unlock(&host_pmu->event_list_lock);
+
+	return ret;
+}
+
+static void host_event_update(struct perf_event *event)
+{
+	struct event_data *e_data = event->pmu_private;
+	struct kvm_event_cb *cb = e_data->cb;
+	struct kvm_vcpu *vcpu = e_data->vcpu;
+	struct hw_perf_event *hw = &event->hw;
+	u64 prev_count, new_count;
+
+	do {
+		prev_count = local64_read(&hw->prev_count);
+		new_count = cb->get_event_count(vcpu);
+	} while (local64_xchg(&hw->prev_count, new_count) != prev_count);
+
+	local64_add(new_count - prev_count, &event->count);
+}
+
+static void host_event_start(struct perf_event *event, int flags)
+{
+	struct event_data *e_data = event->pmu_private;
+	struct kvm_event_cb *cb = e_data->cb;
+	struct kvm_vcpu *vcpu = e_data->vcpu;
+	u64 val;
+
+	val = cb->get_event_count(vcpu);
+	local64_set(&event->hw.prev_count, val);
+
+	cb->configure_event(vcpu, true);
+}
+
+static void host_event_stop(struct perf_event *event, int flags)
+{
+	struct event_data *e_data = event->pmu_private;
+	struct kvm_event_cb *cb = e_data->cb;
+	struct kvm_vcpu *vcpu = e_data->vcpu;
+
+	cb->configure_event(vcpu, false);
+
+	if (flags & PERF_EF_UPDATE)
+		host_event_update(event);
+}
+
+static int host_event_add(struct perf_event *event, int flags)
+{
+	if (flags & PERF_EF_START)
+		host_event_start(event, flags);
+
+	return 0;
+}
+
+static void host_event_del(struct perf_event *event, int flags)
+{
+	host_event_stop(event, PERF_EF_UPDATE);
+}
+
+static void host_event_read(struct perf_event *event)
+{
+	host_event_update(event);
+}
+
+static void init_host_pmu(struct host_pmu *host_pmu)
+{
+	host_pmu->pmu = (struct pmu) {
+		.task_ctx_nr	= perf_sw_context,
+		.attr_groups	= attr_groups,
+		.event_init	= host_event_init,
+		.add		= host_event_add,
+		.del		= host_event_del,
+		.start		= host_event_start,
+		.stop		= host_event_stop,
+		.read		= host_event_read,
+		.capabilities	= PERF_PMU_CAP_NO_INTERRUPT,
+	};
+
+	INIT_LIST_HEAD(&host_pmu->event_list_head);
+	spin_lock_init(&host_pmu->event_list_lock);
+}
+
+int kvm_host_pmu_register(int nr_events, struct kvm_event_cb *cbs,
+			  struct attribute **event_attrs)
+{
+	host_pmu.nr_events  = nr_events;
+	host_pmu.cbs = cbs;
+	events_attr_group.attrs = event_attrs;
+
+	init_host_pmu(&host_pmu);
+
+	return perf_pmu_register(&host_pmu.pmu, "kvm", -1);
+}
+
+void kvm_host_pmu_unregister(void)
+{
+	perf_pmu_unregister(&host_pmu.pmu);
+}
-- 
2.11.0