kvm: debugfs: Export vcpu stat via debugfs

virt inclusion
category: feature
bugzilla: 46853
CVE: NA

This patch create debugfs entry for vcpu stat.
The entry path is /sys/kernel/debug/kvm/vcpu_stat.
And vcpu_stat contains partial kvm exits items of vcpu, include:
	pid, hvc_exit_stat, wfe_exit_stat, wfi_exit_stat,
	mmio_exit_user, mmio_exit_kernel, exits

Currently, The maximum vcpu limit is 1024.

From this vcpu_stat, user can get the number of these kvm exits items
over a period of time, which is helpful to monitor the virtual machine.
Signed-off-by: NZenghui Yu <yuzenghui@huawei.com>
Signed-off-by: Nchenjiajun <chenjiajun8@huawei.com>
Reviewed-by: NXiangyou Xie <xiexiangyou@huawei.com>
Acked-by: NXie XiuQi <xiexiuqi@huawei.com>
Signed-off-by: NChen Jun <chenjun102@huawei.com>

arch/arm64/include/asm/kvm_host.h

@@ -455,6 +455,7 @@ struct kvm_vm_stat {
 };
 
 struct kvm_vcpu_stat {
+	u64 pid;
 	u64 halt_successful_poll;
 	u64 halt_attempted_poll;
 	u64 halt_poll_success_ns;

arch/arm64/kvm/guest.c

@@ -42,6 +42,19 @@ struct kvm_stats_debugfs_item debugfs_entries[] = {
 	VCPU_STAT("exits", exits),
 	VCPU_STAT("halt_poll_success_ns", halt_poll_success_ns),
 	VCPU_STAT("halt_poll_fail_ns", halt_poll_fail_ns),
+	{ "vcpu_stat", 0, KVM_STAT_DFX },
+	{ NULL }
+};
+
+/* debugfs entries of Detail For vcpu stat EXtension */
+struct dfx_kvm_stats_debugfs_item dfx_debugfs_entries[] = {
+	DFX_STAT("pid", pid),
+	DFX_STAT("hvc_exit_stat", hvc_exit_stat),
+	DFX_STAT("wfe_exit_stat", wfe_exit_stat),
+	DFX_STAT("wfi_exit_stat", wfi_exit_stat),
+	DFX_STAT("mmio_exit_user", mmio_exit_user),
+	DFX_STAT("mmio_exit_kernel", mmio_exit_kernel),
+	DFX_STAT("exits", exits),
 	{ NULL }
 };
 

arch/x86/include/asm/kvm_host.h

@@ -1028,6 +1028,7 @@ struct kvm_vm_stat {
 };
 
 struct kvm_vcpu_stat {
+	u64 pid;
 	u64 pf_fixed;
 	u64 pf_guest;
 	u64 tlb_flush;

arch/x86/kvm/x86.c

@@ -245,6 +245,11 @@ struct kvm_stats_debugfs_item debugfs_entries[] = {
 	{ NULL }
 };
 
+/* debugfs entries of Detail For vcpu stat EXtension */
+struct dfx_kvm_stats_debugfs_item dfx_debugfs_entries[] = {
+	{ NULL }
+};
+
 u64 __read_mostly host_xcr0;
 u64 __read_mostly supported_xcr0;
 EXPORT_SYMBOL_GPL(supported_xcr0);

include/linux/kvm_host.h

@@ -1151,6 +1151,7 @@ static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
 enum kvm_stat_kind {
 	KVM_STAT_VM,
 	KVM_STAT_VCPU,
+	KVM_STAT_DFX,   /* Detail For vcpu stat EXtension */
 };
 
 struct kvm_stat_data {
@@ -1172,10 +1173,26 @@ struct kvm_stats_debugfs_item {
 	{ n, offsetof(struct kvm, stat.x), KVM_STAT_VM, ## __VA_ARGS__ }
 #define VCPU_STAT(n, x, ...)							\
 	{ n, offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU, ## __VA_ARGS__ }
+#define DFX_STAT(n, x, ...)							\
+	{ n, offsetof(struct kvm_vcpu_stat, x), DFX_STAT_U64, ## __VA_ARGS__ }
 
 extern struct kvm_stats_debugfs_item debugfs_entries[];
 extern struct dentry *kvm_debugfs_dir;
 
+enum dfx_stat_kind {
+	DFX_STAT_U64,
+	DFX_STAT_CPUTIME,
+};
+
+/* Detail For vcpu stat EXtension debugfs item */
+struct dfx_kvm_stats_debugfs_item {
+	const char *name;
+	int offset;
+	enum dfx_stat_kind dfx_kind;
+	struct dentry *dentry;
+};
+extern struct dfx_kvm_stats_debugfs_item dfx_debugfs_entries[];
+
 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
 {

virt/kvm/kvm_main.c

@@ -152,6 +152,11 @@ static void kvm_uevent_notify_change(unsigned int type, struct kvm *kvm);
 static unsigned long long kvm_createvm_count;
 static unsigned long long kvm_active_vms;
 
+/* debugfs entries of Detail For vcpu stat EXtension */
+__weak struct dfx_kvm_stats_debugfs_item dfx_debugfs_entries[] = {
+	{ NULL }
+};
+
 __weak void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
 						   unsigned long start, unsigned long end)
 {
@@ -3230,6 +3235,9 @@ static long kvm_vcpu_ioctl(struct file *filp,
 			if (oldpid)
 				synchronize_rcu();
 			put_pid(oldpid);
+#if defined(CONFIG_X86) || defined(CONFIG_ARM64)
+			vcpu->stat.pid = current->pid;
+#endif /* defined(CONFIG_X86) || defined (CONFIG_ARM64) */
 		}
 		r = kvm_arch_vcpu_ioctl_run(vcpu);
 		trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
@@ -4488,6 +4496,9 @@ static int kvm_stat_data_get(void *data, u64 *val)
 		r = kvm_get_stat_per_vcpu(stat_data->kvm,
 					  stat_data->dbgfs_item->offset, val);
 		break;
+	case KVM_STAT_DFX:
+		r = -ENOSYS;
+		break;
 	}
 
 	return r;
@@ -4510,6 +4521,9 @@ static int kvm_stat_data_clear(void *data, u64 val)
 		r = kvm_clear_stat_per_vcpu(stat_data->kvm,
 					    stat_data->dbgfs_item->offset);
 		break;
+	case KVM_STAT_DFX:
+		r = -ENOSYS;
+		break;
 	}
 
 	return r;
@@ -4602,9 +4616,80 @@ static int vcpu_stat_clear(void *_offset, u64 val)
 DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, vcpu_stat_clear,
 			"%llu\n");
 
+#define DFX_MAX_VCPU		1024
+#define DFX_MAX_VCPU_STAT_SIZE	1024
+
+static int __dfx_vcpu_stats_get(struct seq_file *p, void *v)
+{
+	struct kvm *kvm;
+	struct kvm_vcpu *vcpu;
+	struct kvm_vcpu_stat *vcpu_stats;
+	struct dfx_kvm_stats_debugfs_item *dp;
+	int vcpu_nr = 0;
+	int i, index = 0;
+
+	mutex_lock(&kvm_lock);
+	list_for_each_entry(kvm, &vm_list, vm_list)
+		kvm_for_each_vcpu(i, vcpu, kvm)
+			vcpu_nr++;
+	mutex_unlock(&kvm_lock);
+
+	vcpu_nr = min(vcpu_nr, DFX_MAX_VCPU);
+	vcpu_stats = vmalloc(vcpu_nr * sizeof(struct kvm_vcpu_stat));
+	if (!vcpu_stats)
+		return -ENOMEM;
+
+	mutex_lock(&kvm_lock);
+	list_for_each_entry(kvm, &vm_list, vm_list)
+		kvm_for_each_vcpu(i, vcpu, kvm) {
+			if (index >= vcpu_nr)
+				break;
+			memcpy(vcpu_stats + index, &vcpu->stat,
+			       sizeof(struct kvm_vcpu_stat));
+			++index;
+		}
+	mutex_unlock(&kvm_lock);
+
+	for (i = 0; i < vcpu_nr; i++) {
+		for (dp = dfx_debugfs_entries; dp->name; ++dp) {
+			switch (dp->dfx_kind) {
+			case DFX_STAT_U64:
+				seq_put_decimal_ull(p, " ",
+						*(u64 *)((void *)&vcpu_stats[i] + dp->offset));
+				break;
+			case DFX_STAT_CPUTIME:
+				pr_warn("DFX_STAT_CPUTIME not supported currently!");
+				break;
+			default:
+				pr_warn("Bad dfx_kind in dfx_debugfs_entries!");
+				break;
+			}
+		}
+		seq_putc(p, '\n');
+	}
+
+	vfree(vcpu_stats);
+	return 0;
+}
+
+static int dfx_vcpu_stats_open(struct inode *inode, struct file *file)
+{
+	size_t size = DFX_MAX_VCPU_STAT_SIZE * (DFX_MAX_VCPU + 1);
+
+	return single_open_size(file, __dfx_vcpu_stats_get, NULL, size);
+}
+
+static const struct file_operations dfx_stat_fops = {
+	.open		= dfx_vcpu_stats_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 static const struct file_operations *stat_fops[] = {
 	[KVM_STAT_VCPU] = &vcpu_stat_fops,
 	[KVM_STAT_VM]   = &vm_stat_fops,
+	[KVM_STAT_DFX]	= &dfx_stat_fops,
 };
 
 static void kvm_uevent_notify_change(unsigned int type, struct kvm *kvm)