From d2ac291e7dfa2bcb373ed243fabf7a10c9c3a1a0 Mon Sep 17 00:00:00 2001
From: maxv <maxv@NetBSD.org>
Date: Sun, 27 Oct 2019 20:17:36 +0000
Subject: [PATCH] Change the way root_owner works: consider the calling process
 as root_owner not if it has root privileges, but if the /dev/nvmm device was
 opened with write permissions. Introduce the undocumented nvmm_root_init()
 function to achieve that.

The goal is to simplify the logic and have more granularity, eg if we want
a monitoring agent to access VMs but don't want to give this agent real
root access on the system.
---
 lib/libnvmm/libnvmm.c | 25 ++++++++++++++++++++++++-
 lib/libnvmm/nvmm.h    |  3 ++-
 sys/dev/nvmm/nvmm.c   | 29 ++++++++++++++---------------
 3 files changed, 40 insertions(+), 17 deletions(-)

diff --git a/lib/libnvmm/libnvmm.c b/lib/libnvmm/libnvmm.c
index 45275f2be681..e61a929b5a5f 100644
--- a/lib/libnvmm/libnvmm.c
+++ b/lib/libnvmm/libnvmm.c
@@ -1,4 +1,4 @@
-/*	$NetBSD: libnvmm.c,v 1.17 2019/10/27 07:08:15 maxv Exp $	*/
+/*	$NetBSD: libnvmm.c,v 1.18 2019/10/27 20:17:36 maxv Exp $	*/
 
 /*
  * Copyright (c) 2018-2019 The NetBSD Foundation, Inc.
@@ -179,6 +179,29 @@ nvmm_init(void)
 	return 0;
 }
 
+int
+nvmm_root_init(void)
+{
+	if (nvmm_fd != -1)
+		return 0;
+	nvmm_fd = open("/dev/nvmm", O_WRONLY | O_CLOEXEC);
+	if (nvmm_fd == -1)
+		return -1;
+	if (nvmm_capability(&__capability) == -1) {
+		close(nvmm_fd);
+		nvmm_fd = -1;
+		return -1;
+	}
+	if (__capability.version != NVMM_KERN_VERSION) {
+		close(nvmm_fd);
+		nvmm_fd = -1;
+		errno = EPROGMISMATCH;
+		return -1;
+	}
+
+	return 0;
+}
+
 int
 nvmm_capability(struct nvmm_capability *cap)
 {
diff --git a/lib/libnvmm/nvmm.h b/lib/libnvmm/nvmm.h
index 02e0be032bf5..ed4ac7783233 100644
--- a/lib/libnvmm/nvmm.h
+++ b/lib/libnvmm/nvmm.h
@@ -1,4 +1,4 @@
-/*	$NetBSD: nvmm.h,v 1.15 2019/10/27 07:08:15 maxv Exp $	*/
+/*	$NetBSD: nvmm.h,v 1.16 2019/10/27 20:17:36 maxv Exp $	*/
 
 /*
  * Copyright (c) 2018-2019 The NetBSD Foundation, Inc.
@@ -90,6 +90,7 @@ struct nvmm_mem {
 typedef uint64_t nvmm_prot_t;
 
 int nvmm_init(void);
+int nvmm_root_init(void);
 
 int nvmm_capability(struct nvmm_capability *);
 
diff --git a/sys/dev/nvmm/nvmm.c b/sys/dev/nvmm/nvmm.c
index 1a7b7d089779..411d02c25ba7 100644
--- a/sys/dev/nvmm/nvmm.c
+++ b/sys/dev/nvmm/nvmm.c
@@ -1,4 +1,4 @@
-/*	$NetBSD: nvmm.c,v 1.23 2019/10/23 07:01:11 maxv Exp $	*/
+/*	$NetBSD: nvmm.c,v 1.24 2019/10/27 20:17:36 maxv Exp $	*/
 
 /*
  * Copyright (c) 2018-2019 The NetBSD Foundation, Inc.
@@ -30,7 +30,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: nvmm.c,v 1.23 2019/10/23 07:01:11 maxv Exp $");
+__KERNEL_RCSID(0, "$NetBSD: nvmm.c,v 1.24 2019/10/27 20:17:36 maxv Exp $");
 
 #include <sys/param.h>
 #include <sys/systm.h>
@@ -889,7 +889,7 @@ out:
 /* -------------------------------------------------------------------------- */
 
 static int
-nvmm_ctl_mach_info(struct nvmm_ioc_ctl *args)
+nvmm_ctl_mach_info(struct nvmm_owner *owner, struct nvmm_ioc_ctl *args)
 {
 	struct nvmm_ctl_mach_info ctl;
 	struct nvmm_machine *mach;
@@ -903,7 +903,7 @@ nvmm_ctl_mach_info(struct nvmm_ioc_ctl *args)
 	if (error)
 		return error;
 
-	error = nvmm_machine_get(&root_owner, ctl.machid, &mach, true);
+	error = nvmm_machine_get(owner, ctl.machid, &mach, true);
 	if (error)
 		return error;
 
@@ -930,16 +930,9 @@ nvmm_ctl_mach_info(struct nvmm_ioc_ctl *args)
 static int
 nvmm_ctl(struct nvmm_owner *owner, struct nvmm_ioc_ctl *args)
 {
-	int error;
-
-	error = kauth_authorize_device(curlwp->l_cred, KAUTH_DEVICE_NVMM_CTL,
-	    NULL, NULL, NULL, NULL);
-	if (error)
-		return error;
-
 	switch (args->op) {
 	case NVMM_CTL_MACH_INFO:
-		return nvmm_ctl_mach_info(args);
+		return nvmm_ctl_mach_info(owner, args);
 	default:
 		return EINVAL;
 	}
@@ -1047,8 +1040,12 @@ nvmm_open(dev_t dev, int flags, int type, struct lwp *l)
 	if (error)
 		return error;
 
-	owner = kmem_alloc(sizeof(*owner), KM_SLEEP);
-	owner->pid = l->l_proc->p_pid;
+	if (OFLAGS(flags) & O_WRONLY) {
+		owner = &root_owner;
+	} else {
+		owner = kmem_alloc(sizeof(*owner), KM_SLEEP);
+		owner->pid = l->l_proc->p_pid;
+	}
 
 	return fd_clone(fp, fd, flags, &nvmm_fileops, owner);
 }
@@ -1060,7 +1057,9 @@ nvmm_close(file_t *fp)
 
 	KASSERT(owner != NULL);
 	nvmm_kill_machines(owner);
-	kmem_free(owner, sizeof(*owner));
+	if (owner != &root_owner) {
+		kmem_free(owner, sizeof(*owner));
+	}
 	fp->f_data = NULL;
 
    	return 0;