-----BEGIN PGP SIGNED MESSAGE-----
Xen Security Advisory CVE-2017-5753,CVE-2017-5715,CVE-2017-5754 / XSA-254
Information leak via side effects of speculative execution
UPDATES IN VERSION 11
Information provided about migitation for Spectre variant 2.
Mention whether CPU hardware virtualisation extensions are required
in the SP3 mitigations summary table.
An additional patch "x86: fix GET_STACK_END" is required to fix a
possible build failure in the PTI patches. README.pti updated
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:
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
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
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.
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:
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.
There is no mitigation for SP1.
SP2 can be mitigated by a combination of new microcode and compiler
and hypervisor changes. See Resolution below.
SP3 can be mitigated by page-table isolation ("PTI").
See Resolution below.
SP3 can, alternatively, 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
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
The HVM shim (codenamed "Vixen") is available now, as is the PVH shim
(codenamed "Comet") for Xen 4.10 and Xen 4.8. Please read
README.which-shim to determine which shim is suitable for you.
These are hardware bugs, so technically speaking they cannot be
properly fixed in software. However, it is possible in many cases to
provide patches to software to work around the problems.
There is no available resolution for SP1. A solution may be available
in the future.
SP2 can be mitigated on x86 by combinations of new CPU microcode and
new hypervisor code. The required hypervisor changes for Xen 4.6,
4.7, 4.8, 4.9 and 4.10 are detailed in the attached README.bti.
For AMD hardware, and for Intel hardware pre-dating the Skylake
microarchitecture, the hypervisor changes alone are sufficient to
mitigate the issue for Xen itself. No microcode updates are required.
For the Intel Skylake microarchitecture the hypervisor changes are
insufficient to protect Xen without appropriate new microcode.
Microcode updates are required in any event to guard against one guest
Consult Intel, your hardware vendor, or your dom0 OS distributor for the
Additionally, compiler support for `indirect thunk' is required.
Again, without appropriate compiler support, the hypervisor patches
are insufficient. Consult your compiler distributor.
SP2 is mitigated on ARM 32-bit by a set of changes to the hypervisor
alone. SP2 can be mitigated on ARM 64-bit (aarch64) by a combination
of new PSCI firmware and new hypervisor code. The required hypervisor
changes for Xen 4.6, 4.7, 4.8, 4.9 and 4.10 are detailed in the
For ARM 32-bit these changes are complete.
For ARM 64-bit the hypervisor changes are still in development and are
expected to be available soon.
SP3 can be mitigated by page-table isolation ("PTI").
We have a "stage 1" implementation. It allows 64-bit PV guests to be
run natively while restricting what can be accessed via SP3 to the Xen
stack of the current pcpu (which may contain remnants of information
from other guests, but should be much more difficult to attack
Unfortunately these "stage 1" patches incur a non-negligible
performance overhead; about equivalent to the "PV shim" approaches
above. Moving to plain HVM or PVH guests is recommended where
possible. For more information on that, see below.
Patches for the "stage-1" PTI implementation are available in the Xen
staging-NN branches for each Xen revision. See README.pti for
SP3 MITIGATION OPTIONS SUMMARY TABLE FOR 64-bit X86 PV GUESTS
Everything in this section applies to 64-bit PV x86 guests only.
Xen PTI Use PVH Use HVM PVH shim HVM shim
"stage 1" "Comet" "Vixen"
How to use README.pti type="pvh" type="hvm" README.comet README.vixen
Guest All Linux 4.11+ Most All All
Xen 4.6+ 4.10+ All 4.10, 4.8 All
Testing Limited 4.10: Good Very good Moderate Very good
status Very new 4.8: Moderate
Performance Fair Excellent Varies Fair Fair
Hypervisor Needed No need No need No need No need
SP3 guest Substantially Protected Protected Protected Protected
to host protected
SP3 within Protected Guest Guest Vulnerable Vulnerable
guest patches patches  
SP3 from Protected n/a; vuln. n/a; vuln. n/a; vuln. n/a; vuln.
dom0 user    
Device model No dm No dm Qemu No dm Qemu
Config change None type="pvh" type="hvm"/ type="pvh" Tool to rewrite
builder="hvm" pvshim=1 Needs "sidecar"
Within-guest None Should be Disks+net None None
changes? none may change
CPU hw virt Not needed Needed Needed Needed Needed
Extra RAM use V. slight None ~9Mb/guest >=~20Mb/guest >=~29Mb/guest
Migration OK OK OK OK Unsupported
Guest mem adj OK OK OK Broken Unsupported
vcpu hotplug OK OK OK OK Unsupported
Solution Indefinite Indefinite Indefinite Indefinite Limited
lifetime  
 PVH is supported in Xen 4.8 only with the 4.8 "Comet" security
 Some features in PVH/HVM shim guests are not inherently broken,
but buggy in the currently available versions. These may be fixed in
future proper releases of the same feature.
 Most unikernels have Xen support based on a version of mini-os.
mini-os master can boot PVH. But this is very recent.
 Some guests which have support for Xen PV fail to boot properly in
Xen HVM. Some such guests can made to boot HVM by disabling the
PV-on-HVM support entirely in the guest or in Xen; in that case the
guest may work but IO performance will be poor. Some PV-supporting
guests can boot as HVM, with PV drivers, but fail when migrated.
 The Comet and Vixen shim hypervisors direct-map all of their
"physical" memory, and that direct-map can be accessed using Meltdown
by unprivileged processes in the guest. So the guest is vulnerable to
within-guest Meltdown attacks and the guest operating system cannot
 "Vixen" HVM shim is not expected to be incorporated in future Xen
stable releases. At some point, support for it will be withdrawn.
However, HVM shim functionality may be available in a future Xen 4.10
stable point release and would then probably be useable with the
existing conversion script provided in this advisory.
 The lifetime of the special Comet branches is limited, but we will
not desupport them until some time after the same functionality is in
appropriate Xen stable point releases.
 The 64-bit x86 PV guest ABI precludes a guest from mapping its
kernel and userspace in the same address space. So these guests are
inherently immune to within-guest Meltdown attacks, without
within-guest patching. (This applies to 64-bit x86 PV guests only.)
 It is not possible to run dom0 as HVM. dom0 PVH is a planned
enhancement which is not yet available even in preview form.
$ sha256sum xsa254*/*
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
-----END PGP SIGNATURE-----
Xen-announce mailing list
|Free forum by Nabble||Edit this page|