KVM: Add kvm trace userspace interface

This interface allows user a space application to read the trace of kvm
related events through relayfs.

Signed-off-by: Feng (Eric) Liu <eric.e.liu@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>

5 files changed, 311 insertions, 1 deletions

diff --git a/arch/x86/kvm/Kconfig b/arch/x86/kvm/Kconfig
index 76c70ab44382..8d45fabc5f3b 100644
--- a/arch/x86/kvm/Kconfig
+++ b/arch/x86/kvm/Kconfig
@@ -50,6 +50,17 @@ config KVM_AMD
           Provides support for KVM on AMD processors equipped with the AMD-V
           (SVM) extensions.
 
+config KVM_TRACE
+        bool "KVM trace support"
+        depends on KVM && MARKERS && SYSFS
+        select RELAY
+        select DEBUG_FS
+        default n
+        ---help---
+          This option allows reading a trace of kvm-related events through
+          relayfs.  Note the ABI is not considered stable and will be
+          modified in future updates.
+
 # OK, it's a little counter-intuitive to do this, but it puts it neatly under
 # the virtualization menu.
 source drivers/lguest/Kconfig
diff --git a/arch/x86/kvm/Makefile b/arch/x86/kvm/Makefile
index 4d0c22e11f1a..c97d35c218db 100644
--- a/arch/x86/kvm/Makefile
+++ b/arch/x86/kvm/Makefile
@@ -3,6 +3,9 @@
 #
 
 common-objs = $(addprefix ../../../virt/kvm/, kvm_main.o ioapic.o)
+ifeq ($(CONFIG_KVM_TRACE),y)
+common-objs += $(addprefix ../../../virt/kvm/, kvm_trace.o)
+endif
 
 EXTRA_CFLAGS += -Ivirt/kvm -Iarch/x86/kvm
 
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 578c3638bbba..bd0c2d2d840f 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -15,6 +15,7 @@
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/preempt.h>
+#include <linux/marker.h>
 #include <asm/signal.h>
 
 #include <linux/kvm.h>
@@ -309,5 +310,18 @@ struct kvm_stats_debugfs_item {
         struct dentry *dentry;
 };
 extern struct kvm_stats_debugfs_item debugfs_entries[];
+extern struct dentry *debugfs_dir;
+
+#ifdef CONFIG_KVM_TRACE
+int kvm_trace_ioctl(unsigned int ioctl, unsigned long arg);
+void kvm_trace_cleanup(void);
+#else
+static inline
+int kvm_trace_ioctl(unsigned int ioctl, unsigned long arg)
+{
+        return -EINVAL;
+}
+#define kvm_trace_cleanup() ((void)0)
+#endif
 
 #endif
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 6a52c084e068..d5911d9895c3 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -60,7 +60,7 @@ EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
 
 static __read_mostly struct preempt_ops kvm_preempt_ops;
 
-static struct dentry *debugfs_dir;
+struct dentry *debugfs_dir;
 
 static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
                            unsigned long arg);
@@ -1191,6 +1191,11 @@ static long kvm_dev_ioctl(struct file *filp,
                 r += PAGE_SIZE;    /* pio data page */
 #endif
                 break;
+        case KVM_TRACE_ENABLE:
+        case KVM_TRACE_PAUSE:
+        case KVM_TRACE_DISABLE:
+                r = kvm_trace_ioctl(ioctl, arg);
+                break;
         default:
                 return kvm_arch_dev_ioctl(filp, ioctl, arg);
         }
@@ -1519,6 +1524,7 @@ EXPORT_SYMBOL_GPL(kvm_init);
 
 void kvm_exit(void)
 {
+        kvm_trace_cleanup();
         misc_deregister(&kvm_dev);
         kmem_cache_destroy(kvm_vcpu_cache);
         sysdev_unregister(&kvm_sysdev);
diff --git a/virt/kvm/kvm_trace.c b/virt/kvm/kvm_trace.c
new file mode 100644
index 000000000000..5425440c54bf
--- /dev/null
+++ b/virt/kvm/kvm_trace.c
@@ -0,0 +1,276 @@
+/*
+ * kvm trace
+ *
+ * It is designed to allow debugging traces of kvm to be generated
+ * on UP / SMP machines.  Each trace entry can be timestamped so that
+ * it's possible to reconstruct a chronological record of trace events.
+ * The implementation refers to blktrace kernel support.
+ *
+ * Copyright (c) 2008 Intel Corporation
+ * Copyright (C) 2006 Jens Axboe <axboe@kernel.dk>
+ *
+ * Authors: Feng(Eric) Liu, eric.e.liu@intel.com
+ *
+ * Date:    Feb 2008
+ */
+
+#include <linux/module.h>
+#include <linux/relay.h>
+#include <linux/debugfs.h>
+
+#include <linux/kvm_host.h>
+
+#define KVM_TRACE_STATE_RUNNING         (1 << 0)
+#define KVM_TRACE_STATE_PAUSE           (1 << 1)
+#define KVM_TRACE_STATE_CLEARUP         (1 << 2)
+
+struct kvm_trace {
+        int trace_state;
+        struct rchan *rchan;
+        struct dentry *lost_file;
+        atomic_t lost_records;
+};
+static struct kvm_trace *kvm_trace;
+
+struct kvm_trace_probe {
+        const char *name;
+        const char *format;
+        u32 cycle_in;
+        marker_probe_func *probe_func;
+};
+
+static inline int calc_rec_size(int cycle, int extra)
+{
+        int rec_size = KVM_TRC_HEAD_SIZE;
+
+        rec_size += extra;
+        return cycle ? rec_size += KVM_TRC_CYCLE_SIZE : rec_size;
+}
+
+static void kvm_add_trace(void *probe_private, void *call_data,
+                          const char *format, va_list *args)
+{
+        struct kvm_trace_probe *p = probe_private;
+        struct kvm_trace *kt = kvm_trace;
+        struct kvm_trace_rec rec;
+        struct kvm_vcpu *vcpu;
+        int    i, extra, size;
+
+        if (unlikely(kt->trace_state != KVM_TRACE_STATE_RUNNING))
+                return;
+
+        rec.event       = va_arg(*args, u32);
+        vcpu            = va_arg(*args, struct kvm_vcpu *);
+        rec.pid         = current->tgid;
+        rec.vcpu_id     = vcpu->vcpu_id;
+
+        extra           = va_arg(*args, u32);
+        WARN_ON(!(extra <= KVM_TRC_EXTRA_MAX));
+        extra           = min_t(u32, extra, KVM_TRC_EXTRA_MAX);
+        rec.extra_u32   = extra;
+
+        rec.cycle_in    = p->cycle_in;
+
+        if (rec.cycle_in) {
+                u64 cycle = 0;
+
+                cycle = get_cycles();
+                rec.u.cycle.cycle_lo = (u32)cycle;
+                rec.u.cycle.cycle_hi = (u32)(cycle >> 32);
+
+                for (i = 0; i < rec.extra_u32; i++)
+                        rec.u.cycle.extra_u32[i] = va_arg(*args, u32);
+        } else {
+                for (i = 0; i < rec.extra_u32; i++)
+                        rec.u.nocycle.extra_u32[i] = va_arg(*args, u32);
+        }
+
+        size = calc_rec_size(rec.cycle_in, rec.extra_u32 * sizeof(u32));
+        relay_write(kt->rchan, &rec, size);
+}
+
+static struct kvm_trace_probe kvm_trace_probes[] = {
+        { "kvm_trace_entryexit", "%u %p %u %u %u %u %u %u", 1, kvm_add_trace },
+        { "kvm_trace_handler", "%u %p %u %u %u %u %u %u", 0, kvm_add_trace },
+};
+
+static int lost_records_get(void *data, u64 *val)
+{
+        struct kvm_trace *kt = data;
+
+        *val = atomic_read(&kt->lost_records);
+        return 0;
+}
+
+DEFINE_SIMPLE_ATTRIBUTE(kvm_trace_lost_ops, lost_records_get, NULL, "%llu\n");
+
+/*
+ *  The relay channel is used in "no-overwrite" mode, it keeps trace of how
+ *  many times we encountered a full subbuffer, to tell user space app the
+ *  lost records there were.
+ */
+static int kvm_subbuf_start_callback(struct rchan_buf *buf, void *subbuf,
+                                     void *prev_subbuf, size_t prev_padding)
+{
+        struct kvm_trace *kt;
+
+        if (!relay_buf_full(buf))
+                return 1;
+
+        kt = buf->chan->private_data;
+        atomic_inc(&kt->lost_records);
+
+        return 0;
+}
+
+static struct dentry *kvm_create_buf_file_callack(const char *filename,
+                                                 struct dentry *parent,
+                                                 int mode,
+                                                 struct rchan_buf *buf,
+                                                 int *is_global)
+{
+        return debugfs_create_file(filename, mode, parent, buf,
+                                   &relay_file_operations);
+}
+
+static int kvm_remove_buf_file_callback(struct dentry *dentry)
+{
+        debugfs_remove(dentry);
+        return 0;
+}
+
+static struct rchan_callbacks kvm_relay_callbacks = {
+        .subbuf_start           = kvm_subbuf_start_callback,
+        .create_buf_file        = kvm_create_buf_file_callack,
+        .remove_buf_file        = kvm_remove_buf_file_callback,
+};
+
+static int do_kvm_trace_enable(struct kvm_user_trace_setup *kuts)
+{
+        struct kvm_trace *kt;
+        int i, r = -ENOMEM;
+
+        if (!kuts->buf_size || !kuts->buf_nr)
+                return -EINVAL;
+
+        kt = kzalloc(sizeof(*kt), GFP_KERNEL);
+        if (!kt)
+                goto err;
+
+        r = -EIO;
+        atomic_set(&kt->lost_records, 0);
+        kt->lost_file = debugfs_create_file("lost_records", 0444, debugfs_dir,
+                                            kt, &kvm_trace_lost_ops);
+        if (!kt->lost_file)
+                goto err;
+
+        kt->rchan = relay_open("trace", debugfs_dir, kuts->buf_size,
+                                kuts->buf_nr, &kvm_relay_callbacks, kt);
+        if (!kt->rchan)
+                goto err;
+
+        kvm_trace = kt;
+
+        for (i = 0; i < ARRAY_SIZE(kvm_trace_probes); i++) {
+                struct kvm_trace_probe *p = &kvm_trace_probes[i];
+
+                r = marker_probe_register(p->name, p->format, p->probe_func, p);
+                if (r)
+                        printk(KERN_INFO "Unable to register probe %s\n",
+                               p->name);
+        }
+
+        kvm_trace->trace_state = KVM_TRACE_STATE_RUNNING;
+
+        return 0;
+err:
+        if (kt) {
+                if (kt->lost_file)
+                        debugfs_remove(kt->lost_file);
+                if (kt->rchan)
+                        relay_close(kt->rchan);
+                kfree(kt);
+        }
+        return r;
+}
+
+static int kvm_trace_enable(char __user *arg)
+{
+        struct kvm_user_trace_setup kuts;
+        int ret;
+
+        ret = copy_from_user(&kuts, arg, sizeof(kuts));
+        if (ret)
+                return -EFAULT;
+
+        ret = do_kvm_trace_enable(&kuts);
+        if (ret)
+                return ret;
+
+        return 0;
+}
+
+static int kvm_trace_pause(void)
+{
+        struct kvm_trace *kt = kvm_trace;
+        int r = -EINVAL;
+
+        if (kt == NULL)
+                return r;
+
+        if (kt->trace_state == KVM_TRACE_STATE_RUNNING) {
+                kt->trace_state = KVM_TRACE_STATE_PAUSE;
+                relay_flush(kt->rchan);
+                r = 0;
+        }
+
+        return r;
+}
+
+void kvm_trace_cleanup(void)
+{
+        struct kvm_trace *kt = kvm_trace;
+        int i;
+
+        if (kt == NULL)
+                return;
+
+        if (kt->trace_state == KVM_TRACE_STATE_RUNNING ||
+            kt->trace_state == KVM_TRACE_STATE_PAUSE) {
+
+                kt->trace_state = KVM_TRACE_STATE_CLEARUP;
+
+                for (i = 0; i < ARRAY_SIZE(kvm_trace_probes); i++) {
+                        struct kvm_trace_probe *p = &kvm_trace_probes[i];
+                        marker_probe_unregister(p->name, p->probe_func, p);
+                }
+
+                relay_close(kt->rchan);
+                debugfs_remove(kt->lost_file);
+                kfree(kt);
+        }
+}
+
+int kvm_trace_ioctl(unsigned int ioctl, unsigned long arg)
+{
+        void __user *argp = (void __user *)arg;
+        long r = -EINVAL;
+
+        if (!capable(CAP_SYS_ADMIN))
+                return -EPERM;
+
+        switch (ioctl) {
+        case KVM_TRACE_ENABLE:
+                r = kvm_trace_enable(argp);
+                break;
+        case KVM_TRACE_PAUSE:
+                r = kvm_trace_pause();
+                break;
+        case KVM_TRACE_DISABLE:
+                r = 0;
+                kvm_trace_cleanup();
+                break;
+        }
+
+        return r;
+}