KVM: Create debugfs dir and stat files for each VM

This patch adds a kvm debugfs subdirectory for each VM, which is named
after its pid and file descriptor. The directories contain the same
kind of files that are already in the kvm debugfs directory, but the
data exported through them is now VM specific.

This makes the debugfs kvm data a convenient alternative to the
tracepoints which already have per VM data. The debugfs data is easy
to read and low overhead.

CC: Dan Carpenter <dan.carpenter@oracle.com> [includes fixes by Dan Carpenter]
Signed-off-by: Janosch Frank <frankja@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

include/linux/kvm_host.h

@@ -412,6 +412,8 @@ struct kvm {
 #endif
 	long tlbs_dirty;
 	struct list_head devices;
+	struct dentry *debugfs_dentry;
+	struct kvm_stat_data **debugfs_stat_data;
 };
 
 #define kvm_err(fmt, ...) \
@@ -991,6 +993,11 @@ enum kvm_stat_kind {
 	KVM_STAT_VCPU,
 };
 
+struct kvm_stat_data {
+	int offset;
+	struct kvm *kvm;
+};
+
 struct kvm_stats_debugfs_item {
 	const char *name;
 	int offset;

virt/kvm/kvm_main.c

@@ -63,6 +63,9 @@
 #define CREATE_TRACE_POINTS
 #include <trace/events/kvm.h>
 
+/* Worst case buffer size needed for holding an integer. */
+#define ITOA_MAX_LEN 12
+
 MODULE_AUTHOR("Qumranet");
 MODULE_LICENSE("GPL");
 
@@ -100,6 +103,9 @@ static __read_mostly struct preempt_ops kvm_preempt_ops;
 struct dentry *kvm_debugfs_dir;
 EXPORT_SYMBOL_GPL(kvm_debugfs_dir);
 
+static int kvm_debugfs_num_entries;
+static const struct file_operations *stat_fops_per_vm[];
+
 static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
 			   unsigned long arg);
 #ifdef CONFIG_KVM_COMPAT
@@ -542,6 +548,58 @@ static void kvm_free_memslots(struct kvm *kvm, struct kvm_memslots *slots)
 	kvfree(slots);
 }
 
+static void kvm_destroy_vm_debugfs(struct kvm *kvm)
+{
+	int i;
+
+	if (!kvm->debugfs_dentry)
+		return;
+
+	debugfs_remove_recursive(kvm->debugfs_dentry);
+
+	for (i = 0; i < kvm_debugfs_num_entries; i++)
+		kfree(kvm->debugfs_stat_data[i]);
+	kfree(kvm->debugfs_stat_data);
+}
+
+static int kvm_create_vm_debugfs(struct kvm *kvm, int fd)
+{
+	char dir_name[ITOA_MAX_LEN * 2];
+	struct kvm_stat_data *stat_data;
+	struct kvm_stats_debugfs_item *p;
+
+	if (!debugfs_initialized())
+		return 0;
+
+	snprintf(dir_name, sizeof(dir_name), "%d-%d", task_pid_nr(current), fd);
+	kvm->debugfs_dentry = debugfs_create_dir(dir_name,
+						 kvm_debugfs_dir);
+	if (!kvm->debugfs_dentry)
+		return -ENOMEM;
+
+	kvm->debugfs_stat_data = kcalloc(kvm_debugfs_num_entries,
+					 sizeof(*kvm->debugfs_stat_data),
+					 GFP_KERNEL);
+	if (!kvm->debugfs_stat_data)
+		return -ENOMEM;
+
+	for (p = debugfs_entries; p->name; p++) {
+		stat_data = kzalloc(sizeof(*stat_data), GFP_KERNEL);
+		if (!stat_data)
+			return -ENOMEM;
+
+		stat_data->kvm = kvm;
+		stat_data->offset = p->offset;
+		kvm->debugfs_stat_data[p - debugfs_entries] = stat_data;
+		if (!debugfs_create_file(p->name, 0444,
+					 kvm->debugfs_dentry,
+					 stat_data,
+					 stat_fops_per_vm[p->kind]))
+			return -ENOMEM;
+	}
+	return 0;
+}
+
 static struct kvm *kvm_create_vm(unsigned long type)
 {
 	int r, i;
@@ -647,6 +705,7 @@ static void kvm_destroy_vm(struct kvm *kvm)
 	int i;
 	struct mm_struct *mm = kvm->mm;
 
+	kvm_destroy_vm_debugfs(kvm);
 	kvm_arch_sync_events(kvm);
 	spin_lock(&kvm_lock);
 	list_del(&kvm->vm_list);
@@ -2999,8 +3058,15 @@ static int kvm_dev_ioctl_create_vm(unsigned long type)
 	}
 #endif
 	r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR | O_CLOEXEC);
-	if (r < 0)
+	if (r < 0) {
 		kvm_put_kvm(kvm);
+		return r;
+	}
+
+	if (kvm_create_vm_debugfs(kvm, r) < 0) {
+		kvm_put_kvm(kvm);
+		return -ENOMEM;
+	}
 
 	return r;
 }
@@ -3425,15 +3491,114 @@ static struct notifier_block kvm_cpu_notifier = {
 	.notifier_call = kvm_cpu_hotplug,
 };
 
+static int kvm_debugfs_open(struct inode *inode, struct file *file,
+			   int (*get)(void *, u64 *), int (*set)(void *, u64),
+			   const char *fmt)
+{
+	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)
+					  inode->i_private;
+
+	/* The debugfs files are a reference to the kvm struct which
+	 * is still valid when kvm_destroy_vm is called.
+	 * To avoid the race between open and the removal of the debugfs
+	 * directory we test against the users count.
+	 */
+	if (!atomic_add_unless(&stat_data->kvm->users_count, 1, 0))
+		return -ENOENT;
+
+	if (simple_attr_open(inode, file, get, set, fmt)) {
+		kvm_put_kvm(stat_data->kvm);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static int kvm_debugfs_release(struct inode *inode, struct file *file)
+{
+	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)
+					  inode->i_private;
+
+	simple_attr_release(inode, file);
+	kvm_put_kvm(stat_data->kvm);
+
+	return 0;
+}
+
+static int vm_stat_get_per_vm(void *data, u64 *val)
+{
+	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;
+
+	*val = *(u32 *)((void *)stat_data->kvm + stat_data->offset);
+
+	return 0;
+}
+
+static int vm_stat_get_per_vm_open(struct inode *inode, struct file *file)
+{
+	__simple_attr_check_format("%llu\n", 0ull);
+	return kvm_debugfs_open(inode, file, vm_stat_get_per_vm,
+				NULL, "%llu\n");
+}
+
+static const struct file_operations vm_stat_get_per_vm_fops = {
+	.owner   = THIS_MODULE,
+	.open    = vm_stat_get_per_vm_open,
+	.release = kvm_debugfs_release,
+	.read    = simple_attr_read,
+	.write   = simple_attr_write,
+	.llseek  = generic_file_llseek,
+};
+
+static int vcpu_stat_get_per_vm(void *data, u64 *val)
+{
+	int i;
+	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;
+	struct kvm_vcpu *vcpu;
+
+	*val = 0;
+
+	kvm_for_each_vcpu(i, vcpu, stat_data->kvm)
+		*val += *(u32 *)((void *)vcpu + stat_data->offset);
+
+	return 0;
+}
+
+static int vcpu_stat_get_per_vm_open(struct inode *inode, struct file *file)
+{
+	__simple_attr_check_format("%llu\n", 0ull);
+	return kvm_debugfs_open(inode, file, vcpu_stat_get_per_vm,
+				 NULL, "%llu\n");
+}
+
+static const struct file_operations vcpu_stat_get_per_vm_fops = {
+	.owner   = THIS_MODULE,
+	.open    = vcpu_stat_get_per_vm_open,
+	.release = kvm_debugfs_release,
+	.read    = simple_attr_read,
+	.write   = simple_attr_write,
+	.llseek  = generic_file_llseek,
+};
+
+static const struct file_operations *stat_fops_per_vm[] = {
+	[KVM_STAT_VCPU] = &vcpu_stat_get_per_vm_fops,
+	[KVM_STAT_VM]   = &vm_stat_get_per_vm_fops,
+};
+
 static int vm_stat_get(void *_offset, u64 *val)
 {
 	unsigned offset = (long)_offset;
 	struct kvm *kvm;
+	struct kvm_stat_data stat_tmp = {.offset = offset};
+	u64 tmp_val;
 
 	*val = 0;
 	spin_lock(&kvm_lock);
-	list_for_each_entry(kvm, &vm_list, vm_list)
+	list_for_each_entry(kvm, &vm_list, vm_list) {
-		*val += *(u32 *)((void *)kvm + offset);
+		stat_tmp.kvm = kvm;
+		vm_stat_get_per_vm((void *)&stat_tmp, &tmp_val);
+		*val += tmp_val;
+	}
 	spin_unlock(&kvm_lock);
 	return 0;
 }
@@ -3444,15 +3609,16 @@ static int vcpu_stat_get(void *_offset, u64 *val)
 {
 	unsigned offset = (long)_offset;
 	struct kvm *kvm;
-	struct kvm_vcpu *vcpu;
+	struct kvm_stat_data stat_tmp = {.offset = offset};
-	int i;
+	u64 tmp_val;
 
 	*val = 0;
 	spin_lock(&kvm_lock);
-	list_for_each_entry(kvm, &vm_list, vm_list)
+	list_for_each_entry(kvm, &vm_list, vm_list) {
-		kvm_for_each_vcpu(i, vcpu, kvm)
+		stat_tmp.kvm = kvm;
-			*val += *(u32 *)((void *)vcpu + offset);
+		vcpu_stat_get_per_vm((void *)&stat_tmp, &tmp_val);
-
+		*val += tmp_val;
+	}
 	spin_unlock(&kvm_lock);
 	return 0;
 }
@@ -3473,7 +3639,8 @@ static int kvm_init_debug(void)
 	if (kvm_debugfs_dir == NULL)
 		goto out;
 
-	for (p = debugfs_entries; p->name; ++p) {
+	kvm_debugfs_num_entries = 0;
+	for (p = debugfs_entries; p->name; ++p, kvm_debugfs_num_entries++) {
 		if (!debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
 					 (void *)(long)p->offset,
 					 stat_fops[p->kind]))