Xen Security Advisory 254 (CVE-2017-5753, CVE-2017-5715, CVE-2017-5754) - Information leak via side effects of speculative execution

[Xen.org security team <security@xen.org>]

[-- Attachment #1: Type: text/plain, Size: 7982 bytes --]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

 Xen Security Advisory CVE-2017-5753,CVE-2017-5715,CVE-2017-5754 / XSA-254
                                 version 5

        Information leak via side effects of speculative execution

UPDATES IN VERSION 5
====================

PV-in-PVH/HVM shim approach leaves *guest* vulnerable to Meltdown
attacks from its unprivileged users, even if the guest has KPTI
patches.  That is, guest userspace can use Meltdown to read all memory
in the same guest.

In Vixen shim sidecar creator script, look for qemu in some more
places, and provide a command line option to specify the
qemu-system-i386 to use in case the default doesn't find it.

ISSUE DESCRIPTION
=================

Processors give the illusion of a sequence of instructions executed
one-by-one.  However, in order to most efficiently use cpu resources,
modern superscalar processors actually begin executing many
instructions in parallel.  In cases where instructions depend on the
result of previous instructions or checks which have not yet
completed, execution happens based on guesses about what the outcome
will be.  If the guess is correct, execution has been sped up.  If the
guess is incorrect, partially-executed instructions are cancelled and
architectural state changes (to registers, memory, and so on)
reverted; but the whole process is no slower than if no guess had been
made at all.  This is sometimes called "speculative execution".

Unfortunately, although architectural state is rolled back, there are
other side effects, such as changes to TLB or cache state, which are
not rolled back.  These side effects can subsequently be detected by
an attacker to determine information about what happened during the
speculative execution phase.  If an attacker can cause speculative
execution to access sensitive memory areas, they may be able to infer
what that sensitive memory contained.

Furthermore, these guesses can often be 'poisoned', such that attacker
can cause logic to reliably 'guess' the way the attacker chooses.
This advisory discusses three ways to cause speculative execution to
access sensitive memory areas (named here according to the
discoverer's naming scheme):

"Bounds-check bypass" (aka SP1, "Variant 1", Spectre CVE-2017-5753):
Poison the branch predictor, such that victim code is speculatively
executed past boundary and security checks.  This would allow an
attacker to, for instance, cause speculative code in the normal
hypercall / emulation path to execute with wild array indexes.

"Branch Target Injection" (aka SP2, "Variant 2", Spectre CVE-2017-5715):
Poison the branch predictor.  Well-abstracted code often involves
calling function pointers via indirect branches; reading these
function pointers may involve a (slow) memory access, so the CPU
attempts to guess where indirect branches will lead.  Poisoning this
enables an attacker to speculatively branch to any code that is
executable by the victim (eg, anywhere in the hypervisor).

"Rogue Data Load" (aka SP3, "Variant 3", Meltdown, CVE-2017-5754):
On some processors, certain pagetable permission checks only happen
when the instruction is retired; effectively meaning that speculative
execution is not subject to pagetable permission checks.  On such
processors, an attacker can speculatively execute arbitrary code in
userspace with, effectively, the highest privilege level.

More information is available here:
  https://meltdownattack.com/
  https://spectreattack.com/
  https://googleprojectzero.blogspot.co.uk/2018/01/reading-privileged-memory-with-side.html

Additional Xen-specific background:

Xen hypervisors on most systems map all of physical RAM, so code
speculatively executed in a hypervisor context can read all of system
RAM.

When running PV guests, the guest and the hypervisor share the address
space; guest kernels run in a lower privilege level, and Xen runs in
the highest privilege level.  (x86 HVM and PVH guests, and ARM guests,
run in a separate address space to the hypervisor.)  However, only
64-bit PV guests can generate addresses large enough to point to
hypervisor memory.

IMPACT
======

Xen guests may be able to infer the contents of arbitrary host memory,
including memory assigned to other guests.

An attacker's choice of code to speculatively execute (and thus the
ease of extracting useful information) goes up with the numbers.  For
SP1, an attacker is limited to windows of code after bound checks of
user-supplied indexes.  For SP2, the attacker will in many cases will
be limited to executing arbitrary pre-existing code inside of Xen.
For SP3 (and other cases for SP2), an attacker can write arbitrary
code to speculatively execute.

Additionally, in general, attacks within a guest (from guest user to
guest kernel) will be the same as on real hardware.  Consult your
operating system provider for more information.

NOTE ON TIMING
==============

This vulnerability was originally scheduled to be made public on 9
January.  It was accelerated at the request of the discloser due to
one of the issues being made public.

VULNERABLE SYSTEMS
==================

Systems running all versions of Xen are affected.

For SP1 and SP2, both Intel and AMD are vulnerable.  Vulnerability of
ARM processors to SP1 and SP2 varies by model and manufacturer.  ARM
has information on affected models on the following website:
   https://developer.arm.com/support/security-update

For SP3, only Intel processors are vulnerable.  (The hypervisor cannot
be attacked using SP3 on any ARM processors, even those that are
listed as affected by SP3.)

Furthermore, only 64-bit PV guests can exploit SP3 against Xen.  PVH,
HVM, and 32-bit PV guests cannot exploit SP3.

MITIGATION
==========

There is no mitigation for SP1 and SP2.

SP3 can be mitigated by running guests in HVM or PVH mode.
(Within-guest attacks are still possible unless the guest OS has also
been updated with an SP3 mitigation series such as KPTI/Kaiser.)

For guests with legacy PV kernels which cannot be run in HVM or PVH
mode directly, we have developed two "shim" hypervisors that allow PV
guests to run in HVM mode or PVH mode.  This prevents attacks on the
host, but it leaves the guest vulnerable to Meltdown attacks by its
own unprivileged processes, even if the guest OS has KPTI or similar
Meltdown mitigation.

The HVM shim (codenamed "Vixen") is available now.  We expect to have
the PVH shim (codenamed "Comet") available within a few days.  Please
read README.which-shim to determine which shim is suitable for you.

$ sha256sum xsa254*/*
2df6b811ec7a377a9cc717f7a8ed497f3a90928c21cba81182eb4a802e32ecd7  xsa254/README.vixen
4c30295513ad82debe04845248b5baac0b3d0c151b80fdca32f2df8b9aa0b541  xsa254/README.which-shim
6210615c1384e13da953452e6f47066f8837e2b2c7f671280902e32e96763b54  xsa254/pvshim-converter.pl
$

RESOLUTION
==========

There is no available resolution for SP1.  A solution may be available
in the future.

We are working on patches which mitigate SP2 but these are not
currently available.  Given that the vulnerabilities are now public,
these will be developed and published in public, initially via
xen-devel.


NOTE ON LACK OF EMBARGO
=======================

The timetable and process were set by the discloser.

After the intensive initial response period for these vulnerabilities
is over, we will prepare and publish a full timeline, as we have done
in a handful of other cases of significant public interest where we
saw opportunities for process improvement.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEcBAEBCAAGBQJaWKbzAAoJEIP+FMlX6CvZtl4H/RKmXpS1fL51efZbrhYaDBTF
nLSHxfPdmi+MLaJ8Y7hS9w061ovK7OYTcvi9xlAhE6yC0b4lX5NToc1CPkX6pjGV
atOh0q4QyxDQm9JGW1aL9pZa3ZSF/Y7ad/zv5OlU97ZmDEwuEVvOTSsGj+jMFB08
gJ+VfQ0F2R+sjdh9BIScbUedLEz+M5so2wGaOJObr/ybRfLyAobxwiIc+yPniBoi
c4eNLSdzBjmg0YrRGeMToVziNH6YXmHD+VLSj23SbVYOjgSS/vnbpRtw7DbcwGXy
jhwK8WheInGUsCe+Nz0VU54MXtRhkV+JtsB/g2h4flr49mUm8kt2VY3P0NO7dcE=
=jGQH
-----END PGP SIGNATURE-----

[-- Attachment #2: xsa254/README.vixen --]
[-- Type: application/octet-stream, Size: 2499 bytes --]

                PV-in-HVM shim with "sidecar" ISO
                =================================

Summary
-------

This README describes one of two mitigation strategies for Meltdown.

The basic principle is to run PV guests (which can read all of host
memory due to the hardware bugs) as HVM guests (which cannot, at least
not due to Meltdown).  The PV environment is still provided to the
guest by an embedded copy of Xen, the "shim".  This version of the
shim is codenamed "Vixen".

In order to boot the shim with an unmodified toolstack, you also
provide a special disk containing the shim and the guest kernel (or
pvgrub); this is called the "sidecar".

What you will need
------------------

 * Your host must be able to run grub-mkrescue to generate a .iso
 * You will therefore need xorriso and mtools
 * You must be using xl and able to use an alternative your guest config

 * You will need the script "pvshim-converter.pl"
  - This relies on perl-json
 * You will need the xen.git tag 4.9.1-shim-vixen-1


Instructions
------------

1. On a suitable system (perhaps a different host)
    git clone git://xenbits.xenproject.org/xen.git xen.git
    cd xen.git
    git checkout 4.9.1-shim-vixen-1

If you need bi-directional console and don't mind a less-tested patch,
you can apply the patch found in this email:

    marc.info/?i=<1515604552-9205-1-git-send-email-srn@prgmr.com>

build a xen hypervisor binary as usual:

    make xen

If your default version of python is python 3, you may need to add the following:

    make PYTHON=python2 xen

This will build a file
    xen/xen.gz

2. Copy that file to your dom0.

Ideally someplace like /usr/lib/xen/boot/xen-vixen.gz

3. Copy the script pvshim-converter to your dom0 and make
   it executable:
      chmod +x pvshim-converter.pl

4. For each guest

  (i) if the guest is currently booted with pygrub you must first
   switch to direct kernel boot (by manually copying the kernel and
   initramfs out of the guest, and configuring the command line in the
   domain configuration file), or pvgrub.

  (ii) run
      ./pvshim-converter.pl --shim=/usr/lib/xen/boot/xen-vixen.gz /etc/xen/GUEST.cfg /etc/xen/GUEST.with-shim-cfg

  (iii) shut the guest down cleanly

  (iv) create the guest with the new config
      xl create /etc/xen/GUEST.with-shim-cfg

  (v) Check that it boots properly.  xl console should work.

  (vi) Make arrangements so that autostarting of the guest will use
     the new config file rather than the old one


[-- Attachment #3: xsa254/README.which-shim --]
[-- Type: application/octet-stream, Size: 3949 bytes --]

		   How to decide which shim to use
		   ===============================

A work-around to Meltdown (aka "SP3" or "Variant 3") on Intel
processors is to run guests in HVM or PVH mode.

NB that these approaches leave the guest vulnerable to within-guest
information leaks based on Meltdown, *even if the guest OS has
KPTI/Kaiser or a similar Meltdown migitation*.

Some guests are difficult to convert to running in HVM or PVH mode,
either due to lack of partitioning / MBR, or due to kernel
compatibilities.  As an emergency backstop, there are two approaches,
which we've codenamed "Vixen" and "Comet".  Both involve running an
embedded copy of Xen (called a "shim") within the HVM or PVH guest to
provide the native PV interface.

Below describes the properties, and who might want to use each one.

NOTE: Both shims require host patches to boot on AMD hosts.  This
shouldn't be an issue, as SP3 does not affect systems running on AMD.

Vixen
-----

Vixen has the following properties:
 * Runs the shim in an HVM guest.
 * It requires no hypervisor or toolstack changes, nor does it require
   a host reboot.
 * It has been extensively tested in Amazon's deployment for versions
   of Xen going back to 3.4
 * Guest reboots are required
 * Guest configs must be fed through a converter program
 * The converter program spits out a small guest-specific .iso
   image (we call this a "sidecar") used for booting
 * Because the result is an HVM guest, this approach involves
   running qemu as a PC emulator (this is done automatically)
 * Some common features are not supported:
  - Ballooning
  - Migration
  - vcpu hotplug
  - bidirectional console support (console is write-only)
 * Direct-boot kernels and pvgrub (both pvgrub1 and pvgrub2) are
   supported by the conversion program.  'pygrub' is not supported.
 * xl and xm domain configs can be converted; libvirt domain
   configuration arrangements are not supported.
 * Guest userspace can read all of guest memory, within each guest,
   and a guest migitation for this is not possible.

You might consider this approach if:
- You want to deploy a fix immediately
- You can tolerate the loss of within-guest security
- You can't, or would like to avoid, updating to Xen 4.8 or newer
- You'd like to avoid patching and rebooting your host
- You are able to:
 - Run a script to modify each domain config
 - Afford an extra 80MiB per guest
 - Tolerate having an extra QEMU around
- You don't need migration, memory ballooning, vcpu hotplug,
  or a bi-directional console

To use this solution, see README.vixen.

Bi-directional console is available as an extra patch, but hasn't been
widely tested:

marc.info/?i=<1515604552-9205-1-git-send-email-srn@prgmr.com>

Comet
-----

Comet has the following properties:
 * Runs the shim in a PVH guest.
 * PVH mode is available in Xen 4.10, and will be backported to Xen
   4.9 and 4.8 but no farther
 * Requires host hypervisor and toolstack patches (and host reboot),
   even for Xen 4.10
 * Requires minimal guest config changes, and no "sidecar"
 * Bootloading is identical to native PV guests; direct-boot, pvgrub,
   and pygrub all work equally well
 * Because the result is a PVH guest, this approach involves no PC emulator.
 * The following features not available in Vixen are supported:
  - Memory ballooning
  - Guest migration
  - vcpu hotplug
  - bidirectional console support
 * Guest userspace can read all of guest memory, within each guest,
   and a guest migitation for this is not possible.

You might consider this approach if:
- You're on 4.8 or later already
- You can tolerate the loss of within-guest security
- You can patch and reboot your host
- You don't want an extra QEMU around
- You need migration, memory ballooning, or vcpu hotplug, or a
  bidirectional console
- You need pygrub
- You need to use libvirt

Unfortunately this solution is not yet available.  We expect to have
it available within a few working days.

[-- Attachment #4: xsa254/pvshim-converter.pl --]
[-- Type: application/octet-stream, Size: 6701 bytes --]

#!/usr/bin/perl -w
#
# usage:
#   pvshim-converter [OPTIONS] OLD-CONFIG NEW-CONFIG
#
# options:
#   --qemu PATH-TO-QEMU        filename of qemu-system-i386
#   --sidecars-directory DIR   default is /var/lib/xen/pvshim-sidecars
#   --shim SHIM                overrides domain config file
#   --debug                    verbose, and leaves sidecar prep dir around
#
# What we do
#
#  read existing config file using python
#  determine kernel, ramdisk and cmdline
#  use them to produce sidecar and save it under domain name
#  mess with the things that need to be messed with
#  spit out new config file

use strict;

use Getopt::Long;
use JSON;
use IO::Handle;
use POSIX;
use Fcntl qw(:flock);

our $debug;

sub runcmd {
    print STDERR "+ @_\n" if $debug;
    $!=0; $?=0; system @_ and die "$_[0]: $! $?";
}

our $qemu;
our $shim;
our $sidecars_dir = '/var/lib/xen/pvshim-sidecars';

GetOptions('qemu=s' => \$qemu,
           'sidecars-directory=s' => \$sidecars_dir,
           'shim=s' => \$shim,
           'debug' => \$debug)
    or die "pvshim-converter: bad options\n";

@ARGV==2 or die "pvshim-converter: need old and new config filenames";

our ($in,$out) = @ARGV;

our $indata;

if ($in ne '-') {
    open I, '<', "$in" or die "open input config file: $!\n";
} else {
    open I, '<&STDIN' or die $!;
}
{
    local $/;
    $indata = <I>;
}
I->error and die $!;
close I;

open P, "-|", qw(python2 -c), <<END, $indata or die $!;
import sys
import json
l = {}
exec sys.argv[1] in l
for k in l.keys():
	if k.startswith("_"):
		del l[k]
print json.dumps(l)
END

our $c;

{
    local $/;
    $_ = <P>;
    $!=0; $?=0; close P or die "$! $?";
    $c = decode_json $_;
}

die "no domain name ?" unless exists $c->{name};
die "bootloader not yet supported" if $c->{bootloader};
die "no kernel" unless $c->{kernel};

our $sidecar = $c->{pvshim_sidecar_path} || "$sidecars_dir/$c->{name}.iso";
our $dmwrap = $c->{pvshim_sidecar_path} || "$sidecars_dir/$c->{name}.dm";

$shim ||= $c->{pvshim_path};
$shim ||= '/usr/local/lib/xen/boot/xen-shim';

our $shim_cmdline = $c->{pvshim_cmdline} || 'console=com1 com1=115200n1';
$shim_cmdline .= ' '.$c->{pvshim_extra} if $c->{pvshim_extra};

our $kernel_cmdline = $c->{cmdline} || '';
$kernel_cmdline .= ' root='.$c->{root} if $c->{root};
$kernel_cmdline .= ' '.$c->{extra} if $c->{extra};

print "pvshim-converter: creating sidecar in $sidecar\n";

runcmd qw(mkdir -m700 -p --), $sidecars_dir;

open L, ">", "$sidecar.lock" or die "$sidecar.lock: open $!";
flock L, LOCK_EX or die "$sidecar.lock: lock: $!";

my $sd = "$sidecar.dir";

system qw(rm -rf --), $sd;
mkdir $sd, 0700;

runcmd qw(cp --), $shim, "$sd/shim";
runcmd qw(cp --), $c->{kernel}, "$sd/kernel";
runcmd qw(cp --), $c->{ramdisk}, "$sd/ramdisk" if $c->{ramdisk};

my $grubcfg = <<END;
serial --unit=0 --speed=9600 --word=8 --parity=no --stop=1
terminal_input serial
terminal_output serial

set timeout=0

menuentry 'Xen shim' {
	insmod gzio
	insmod xzio
        multiboot (cd)/shim placeholder $shim_cmdline
        module (cd)/kernel placeholder $kernel_cmdline
        module (cd)/ramdisk
}
END

runcmd qw(mkdir -p --), "$sd/boot/grub";
open G, ">", "$sd/boot/grub/grub.cfg" or die "$sd, grub.cfg: $!";
print G $grubcfg or die $!;
close G or die $!;

unlink "$sidecar.new" or $!==ENOENT or die "$sidecar.new: rm: $!";
runcmd qw(grub-mkrescue -o), "$sidecar.new", "$sidecar.dir";
if (!stat "$sidecar.new") {
    $!==ENOENT or die "$sidecar.new: stat: $!";

    print STDERR <<END;
pvshim-converter: grub-mkrescue exited with status zero but failed to make iso.
NB that grub-mkrescue has a tendency to lie in its error messages.
END
    my $missing;
    foreach my $check (qw(xorriso mformat)) {
        $missing |= system qw(sh -c), "type $check";
    }

    if ($missing) {
        print STDERR <<END;
You seem to have some program(s) missing which grub-mkrescue depends on,
see above.  ("mformat" is normally in the package "mtools".)
Installing those programs will probably help.
END
    } else {
        print STDERR <<END;
And older grub-mkrescue has a tendency not to notice certain problems.
Maybe strace will tell you what is wrong.  :-/
END
    }
    die "pvshim-converter: grub-mkrescue did not make iso\n";
}

runcmd qw(rm -rf --), "$sidecar.dir" unless $debug;

open Q, ">", "$dmwrap.new" or die "$dmwrap: $!";
print Q <<'END_DMWRAP' or die $!;
#!/bin/bash

set -x
: "$@"
set +x

newargs=()

newarg () {
    newargs+=("$1")
}

while [ $# -gt 1 ]; do
    case "$1" in
	-no-shutdown|-nodefaults|-no-user-config)
	    newarg "$1"; shift
	    ;;
	-xen-domid|-chardev|-mon|-display|-boot|-m|-machine)
	    newarg "$1"; shift
	    newarg "$1"; shift
	    ;;
        -name)
            newarg "$1"; shift
            name="$1"; shift
            newarg "$name"
            ;;
	-netdev|-cdrom)
	    : fixme
	    newarg "$1"; shift
	    newarg "$1"; shift
	    ;;
	-drive|-kernel|-initrd|-append|-vnc)
	    shift; shift
	    ;;
	-device)
	    shift
	    case "$1" in
		XXXrtl8139*)
		    newarg "-device"
		    newarg "$1"; shift
		    ;;
		*)
		    shift
		    ;;
	    esac
	    ;;
	*)
	    echo >&2 "warning: unexpected argument $1 being passed through"
	    newarg "$1"; shift
	    ;;
    esac
done

#if [ "x$name" != x ]; then
#    logdir=/var/log/xen
#    logfile="$logdir/shim-$name.log"
#    savelog "$logfile" ||:
#    newarg -serial
#    newarg "file:$logfile"
#fi
END_DMWRAP

if ($qemu) {
    printf Q <<'END_DMWRAP', $qemu or die $!;
    '%s' "${newargs[@]}"
END_DMWRAP
} else {
    print Q <<'END_DMWRAP' or die $!;
set -x
for path in /usr/local/lib/xen/bin /usr/lib/xen/bin /usr/local/bin /usr/bin; do
    $path/qemu-system-i386 "${newargs[@]}" ||:
done
echo >&2 'could not exec qemu'
exit 127
END_DMWRAP
}

chmod 0755, "$dmwrap.new" or die "$dmwrap: chmod: $!";

close Q or die $!;

rename "$sidecar.new", $sidecar or die "$sidecar: install: $!";
rename "$dmwrap.new",  $dmwrap  or die "$dmwrap: install: $!";

print STDERR <<END;
pvshim-converter: wrote qemu wrapper to $dmwrap
pvshim-converter: wrote sidecar to $sidecar
END

my $append = <<END;
builder='hvm'
type='hvm'
device_model_version='qemu-xen'
device_model_override='$dmwrap'
device_model_args_hvm=['-cdrom','$sidecar']
boot='c'
serial='pty'
END

if ($out ne '-') {
    open O, ">", "$out.tmp" or die "open output config temp: $out.tmp: $!\n";
} else {
    open O, ">&STDOUT" or die $!;
}

print O $indata, "\n", $append or die "write output: $!";
close O or die "close output: $!";

if ($out ne '-') {
    rename "$out.tmp", $out or die "install output: $!";
    print STDERR "pvshim-converter: wrote new guest config to $out\n";
} else {
    print STDERR "pvshim-converter: wrote new guest config to stdout\n";
}

[-- Attachment #5: Type: text/plain, Size: 157 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel