[src/trunk]: src/sys/arch Fix a huge privilege separation vulnerability in Xe...

[maxv <maxv%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/0547211a01ad
branches:  trunk
changeset: 828768:0547211a01ad
user:      maxv <maxv%NetBSD.org@localhost>
date:      Sun Dec 31 08:29:38 2017 +0000

description:
Fix a huge privilege separation vulnerability in Xen-amd64.

On amd64 the kernel runs in ring3, like userland, and therefore SEL_KPL
equals SEL_UPL. While Xen can make a distinction between usermode and
kernelmode in %cs, it can't when it comes to iopl. Since we set SEL_KPL
in iopl, Xen sees SEL_UPL, and allows (unprivileged) userland processes
to read and write to the CPU ports.

It is easy, then, to completely escalate privileges; by reprogramming the
PIC, by reading the ATA disks, by intercepting the keyboard interrupts
(keylogger), etc.

Declare IOPL_KPL, set to 1 on Xen-amd64, which allows the kernel to use
the ports but not userland. I didn't test this change on i386, but it
seems fine enough.

diffstat:

 sys/arch/amd64/amd64/machdep.c    |  6 +++---
 sys/arch/amd64/include/segments.h |  8 +++++++-
 sys/arch/i386/i386/machdep.c      |  6 +++---
 sys/arch/i386/include/segments.h  |  4 +++-
 sys/arch/x86/x86/vm_machdep.c     |  6 +++---
 5 files changed, 19 insertions(+), 11 deletions(-)

diffs (121 lines):

diff -r dbef961e827e -r 0547211a01ad sys/arch/amd64/amd64/machdep.c
--- a/sys/arch/amd64/amd64/machdep.c    Sun Dec 31 07:23:09 2017 +0000
+++ b/sys/arch/amd64/amd64/machdep.c    Sun Dec 31 08:29:38 2017 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: machdep.c,v 1.279 2017/12/01 21:22:45 maxv Exp $       */
+/*     $NetBSD: machdep.c,v 1.280 2017/12/31 08:29:38 maxv Exp $       */
 
 /*
  * Copyright (c) 1996, 1997, 1998, 2000, 2006, 2007, 2008, 2011
@@ -110,7 +110,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: machdep.c,v 1.279 2017/12/01 21:22:45 maxv Exp $");
+__KERNEL_RCSID(0, "$NetBSD: machdep.c,v 1.280 2017/12/31 08:29:38 maxv Exp $");
 
 /* #define XENDEBUG_LOW  */
 
@@ -479,7 +479,7 @@
        pcb->pcb_fs = 0;
        pcb->pcb_gs = 0;
        pcb->pcb_rsp0 = (uvm_lwp_getuarea(l) + USPACE - 16) & ~0xf;
-       pcb->pcb_iopl = SEL_KPL;
+       pcb->pcb_iopl = IOPL_KPL;
        pcb->pcb_dbregs = NULL;
        pcb->pcb_cr0 = rcr0() & ~CR0_TS;
        l->l_md.md_regs = (struct trapframe *)pcb->pcb_rsp0 - 1;
diff -r dbef961e827e -r 0547211a01ad sys/arch/amd64/include/segments.h
--- a/sys/arch/amd64/include/segments.h Sun Dec 31 07:23:09 2017 +0000
+++ b/sys/arch/amd64/include/segments.h Sun Dec 31 08:29:38 2017 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: segments.h,v 1.33 2017/11/04 08:50:47 cherry Exp $     */
+/*     $NetBSD: segments.h,v 1.34 2017/12/31 08:29:38 maxv Exp $       */
 
 /*
  * Copyright (c) 1990 The Regents of the University of California.
@@ -107,6 +107,12 @@
 #define ISLDT(s)       ((s) & SEL_LDT) /* is it local or global */
 #define SEL_LDT                4               /* local descriptor table */
 
+#ifdef XEN
+#define IOPL_KPL       1
+#else
+#define IOPL_KPL       SEL_KPL
+#endif
+
 /* Dynamically allocated TSSs and LDTs start (byte offset) */
 #define SYSSEL_START   (NGDT_MEM << 3)
 #define DYNSEL_START   (SYSSEL_START + (NGDT_SYS << 4))
diff -r dbef961e827e -r 0547211a01ad sys/arch/i386/i386/machdep.c
--- a/sys/arch/i386/i386/machdep.c      Sun Dec 31 07:23:09 2017 +0000
+++ b/sys/arch/i386/i386/machdep.c      Sun Dec 31 08:29:38 2017 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: machdep.c,v 1.799 2017/11/11 12:51:06 maxv Exp $       */
+/*     $NetBSD: machdep.c,v 1.800 2017/12/31 08:29:38 maxv Exp $       */
 
 /*
  * Copyright (c) 1996, 1997, 1998, 2000, 2004, 2006, 2008, 2009, 2017
@@ -67,7 +67,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: machdep.c,v 1.799 2017/11/11 12:51:06 maxv Exp $");
+__KERNEL_RCSID(0, "$NetBSD: machdep.c,v 1.800 2017/12/31 08:29:38 maxv Exp $");
 
 #include "opt_beep.h"
 #include "opt_compat_freebsd.h"
@@ -472,7 +472,7 @@
 
        pcb->pcb_cr0 = rcr0() & ~CR0_TS;
        pcb->pcb_esp0 = uvm_lwp_getuarea(l) + USPACE - 16;
-       pcb->pcb_iopl = SEL_KPL;
+       pcb->pcb_iopl = IOPL_KPL;
        l->l_md.md_regs = (struct trapframe *)pcb->pcb_esp0 - 1;
        memcpy(&pcb->pcb_fsd, &gdtstore[GUDATA_SEL], sizeof(pcb->pcb_fsd));
        memcpy(&pcb->pcb_gsd, &gdtstore[GUDATA_SEL], sizeof(pcb->pcb_gsd));
diff -r dbef961e827e -r 0547211a01ad sys/arch/i386/include/segments.h
--- a/sys/arch/i386/include/segments.h  Sun Dec 31 07:23:09 2017 +0000
+++ b/sys/arch/i386/include/segments.h  Sun Dec 31 08:29:38 2017 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: segments.h,v 1.63 2017/11/04 08:50:47 cherry Exp $     */
+/*     $NetBSD: segments.h,v 1.64 2017/12/31 08:29:38 maxv Exp $       */
 
 /*-
  * Copyright (c) 1990 The Regents of the University of California.
@@ -105,6 +105,8 @@
 #define ISLDT(s)       ((s) & SEL_LDT) /* is it local or global */
 #define SEL_LDT                4               /* local descriptor table */
 
+#define IOPL_KPL       SEL_KPL
+
 /* Dynamically allocated TSSs and LDTs start (byte offset) */
 #define DYNSEL_START   (NGDT << 3)
 
diff -r dbef961e827e -r 0547211a01ad sys/arch/x86/x86/vm_machdep.c
--- a/sys/arch/x86/x86/vm_machdep.c     Sun Dec 31 07:23:09 2017 +0000
+++ b/sys/arch/x86/x86/vm_machdep.c     Sun Dec 31 08:29:38 2017 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: vm_machdep.c,v 1.29 2017/06/17 07:45:13 maxv Exp $     */
+/*     $NetBSD: vm_machdep.c,v 1.30 2017/12/31 08:29:38 maxv Exp $     */
 
 /*-
  * Copyright (c) 1982, 1986 The Regents of the University of California.
@@ -80,7 +80,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: vm_machdep.c,v 1.29 2017/06/17 07:45:13 maxv Exp $");
+__KERNEL_RCSID(0, "$NetBSD: vm_machdep.c,v 1.30 2017/12/31 08:29:38 maxv Exp $");
 
 #include "opt_mtrr.h"
 
@@ -164,7 +164,7 @@
        pcb2->pcb_dbregs = NULL;
 
 #if defined(XEN)
-       pcb2->pcb_iopl = SEL_KPL;
+       pcb2->pcb_iopl = IOPL_KPL;
 #endif
 
        /*