Rework dependencies of all objects. We don't need to add dependencies
for headers that $(CC) is capable of generating, we only need to
include $(DEPS_INCLUDE). Some dependencies are still needed so make
knows to generate symlinks for them.
We remove the use of "vpath" for cpuid.c. While it works fine for now,
when we will convert this makefile to subdirmk, vpath will not be
usable. Also, "-iquote" is now needed to build "cpuid.o".
Replace "-I." by "-iquote .", so it applies to double-quote includes
only.
Rather than checking if a symlink exist, always regenerate the
symlink. So if the source tree changed location, the symlink is
updated.
Since we are creating a new .gitignore for the symlink, also move the
entry to it.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Luca Fancellu <luca.fancellu@arm.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Rename ELF_LIB_OBJS to LIBELF_OBJS as to have the same name as in
libs/guest/.
Replace "-I" by "-iquote".
Remove the use of "vpath". It will not works when we will convert this
makefile to subdirmk. Instead, we create symlinks to the source files.
Since we are creating a new .gitignore for the links, also move the
existing entry to it.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Luca Fancellu <luca.fancellu@arm.com>
This is specified (and asserted for in a number of places) to always be
CS. Passing this as an argument in various places is therefore
pointless. The price to pay is two simple new functions, with the
benefit of the PTWR case now gaining a more appropriate error code.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Paul Durrant <paul@xen.org>
The legacy __xen_guest section doesn't support the PHYS32_ENTRY
elfnote, so it's pointless to attempt to parse the elfnotes from that
section when called from an hvm container.
Pass an hvm boolean parameter to the elf note parsing routine, so that
the respective parsing can be suppressed in case libelf is dealing with
an hvm container.
Suggested-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Commit fd35f32b4b ("tools/x86emul: Use struct cpuid_policy in the
userspace test harnesses") didn't account for the dependencies of
cpuid-autogen.h to potentially change between incremental builds. In
particular the harness has a "run" goal which is supposed to be usable
independently of the rest of the tools sub-tree building, and both the
harness and the fuzzer code are also supposed to be buildable
independently. Therefore a re-build of the generated header needs to be
triggered first, which is achieved by introducing a new top-level target
pattern (for just the "run" part for now).
Further cpuid.o did not have any dependencies added for it.
Finally, while at it, add a "run" target to the cpu-policy test harness.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Just like for INVLPGA the HVM hook only supports PCID 0 for the time
being for individual address invalidation. It also translates the other
types to a full flush, which is architecturally permitted and
performance-wise presumably not much worse because emulation is slow
anyway.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
The hook is already in use for INVLPGA as well. Rename the hook and add
parameters. For the moment INVLPGA with a non-zero ASID remains
unsupported, but the TODO item gets pushed into the actual hook handler.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
The hook is already in use for other purposes, and emulating e.g.
CLFLUSH by issuing WBINVD is, well, not very nice. Rename the hook and
add parameters. Use lighter weight flushing insns when possible in
hvmemul_cache_op().
hvmemul_cache_op() treating x86emul_invd the same as x86emul_wbinvd is
to retain original behavior, but I'm not sure this is what we want in
the long run.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Paul Durrant <paul.durrant@citrix.com>
There is now enough complexity that a fuzzing harness is a good idea, and
enough supporting logic to implement one which AFL seems happy with.
Take the existing recalculate_synth() helper and export it as
x86_cpuid_policy_recalc_synth(), as it is needed by the fuzzing harness.
While editing the MAINTAINERS file, insert a related entry which was
accidentally missed from c/s 919ddc3c0 "tools/cpu-policy: Add unit tests", and
sort the lines.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
This is to accompany sanitize_input(). Just like for initial state we
want to have state between two emulated insns sane, at least as far as
assumptions in the main emulator go. Do minimal checking after segment
register, CR, and MSR writes, and roll back to the old value in case of
failure (raising #GP(0) at the same time).
In the particular case observed, a CR0 write clearing CR0.PE was
followed by a VEX-encoded insn, which the decoder accepts based on
guest address size, restricting things just outside of the 64-bit case
(real and virtual modes don't allow VEX-encoded insns). Subsequently
_get_fpu() would then assert that CR0.PE must be set (and EFLAGS.VM
clear) when trying to invoke YMM, ZMM, or OPMASK state.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Don't enforce any other dependencies for now, just like we don't enforce
e.g. PAE enabled as a prereq for long mode.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: George Dunlap <george.dunlap@citrix.com>
Drop it entirely for %rbp - this register is not special purpose enough
to warrant such special treatment. Add a comment to clarify the purpose
of the canonicalization of %rip and %rsp.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
This will shortly be required to pass into the emulator itself.
Simplify the shared cpu_has_* helpers by reading the correct bit out of the
CPUID policy itself.
No (intended) change in behaviour.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Workaround for compiler rejection of SSE-using always_inlines defined before
SSE is disabled.
Compiling with _FORTIFY_SOURCE or higher levels of optimization enabled
will always_inline several library fns (memset, memcpy, ...)
(with gcc 8.2.0 and glibc 2.28).
In fuzz and x86_emulator test, the compiler is instructed not
to generate SSE instructions via: #pragma GCC target("no-sse")
because those registers are needed for use by the workload.
The combination above causes compilation failure as the inline functions
use those instructions. This is resolved by reordering the inclusion of
<stdio.h> and <string.h> to after the pragma disabling SSE generation.
It would be preferable to locate the no-sse pragma within x86-emulate.h at the
top of the file, prior to including any other headers; unfortunately doing so
before <stdlib.h> causes compilation failure due to declaration of 'atof' with:
"SSE register return with SSE disabled".
Fortunately there is no (known) current dependency on any always_inline
SSE-inclined function declared in <stdlib.h> or any of its dependencies, so the
pragma is therefore issued immediately after inclusion of <stdlib.h> with a
comment introduced to explain its location there.
Add compile-time checks for unwanted prior inclusion of <string.h> and
<stdio.h>, which are the two headers that provide the library functions that
are handled with wrappers and listed within "x86-emulate.h" as ones "we think
might access any of the FPU state".
* Use standard-defined "EOF" macro to detect prior <stdio.h> inclusion.
* Use "_STRING_H" (non-standardized guard macro) as best-effort
for detection of prior <string.h> inclusion. This is non-universally
viable but will provide error output on common GLIBC systems, so
provides some defensive coverage.
Adds conditional #include <stdio.h> to x86-emulate.h because fwrite, printf,
etc. are referenced when WRAP has been defined.
Signed-off-by: Christopher Clark <christopher.clark6@baesystems.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
In the x86 test harness and the fuzzer, and instead create a link in
the tools/include directory that can be used by all the tools.
No functional change.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
In my automated SLE_11 builds I often see failures like that:
[ 74s] wrappers.c:5:25: error: x86-emulate.h: No such file or directory
[ 74s] make[6]: *** [wrappers.o] Error 1
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Move line to where it belongs.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Use hooks, just like done for other special purpose registers.
This includes moving XCR0 checks from hvmemul_get_fpu() to the emulator
itself as well as adding support for XGETBV emulation.
For now fuzzer reads will obtain the real values (minus the fuzzing of
the hook pointer itself).
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com> [tracing parts]
Currently with the native tool chain on Debian Jessie ./test_x86_emulator
yields:
Testing AVX2 256bit single native execution...okay
Testing AVX2 256bit single 64-bit code sequence...[line 933] failed!
The bug is that libc's memcpy() in read() uses %xmm8 (specifically, in
__memcpy_sse2_unaligned()), which corrupts %ymm8 behind the back of the AVX2
test code.
Introduce wrappers (and machinery to forward calls to those wrappers)
saving/restoring FPU state around certain library calls.
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-and-tested-by: Andrew Cooper <andrew.cooper3@citrix.com>
The maximum size for the input size was set to INPUT_SIZE, which is actually
the size of the data array inside the fuzz_corpus structure and so was not
abling user (or AFL) to fill in the whole structure. Changing to
sizeof(struct fuzz_corpus) correct this problem.
Signed-off-by: Paul Semel <semelpaul@gmail.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Changeset 2b1cde7783 introduced "batch mode" to afl-harness, which allowed
the handling of several inputs in sequence.
Unfortunately, it introduced a file pointer leak when the file was
larger than the maximum size. Restructure the code to always close fp
if we opened it.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Release-acked-by: Julien Grall <julien.grall@linaro.org>
Move fuzz-emul.c function prototypes into a header. Also share the
definition of the input size (rather than hard-coding it in
fuzz-emul.c).
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Finding aggregate coverage for a set of test files means running each
afl-generated test case through the harness. At the moment, this is
done by re-executing afl-harness-cov with each input file. When a
large number of test cases have been generated, this can take a
significant amonut of time; a recent test with 30k total files
generated by 4 parallel fuzzers took over 7 minutes.
The vast majority of this time is taken up with 'exec', however.
Since the harness is already designed to loop over multiple inputs for
llvm "persistent mode", just allow it to take a large number of inputs
on the same when *not* running in llvm "persistent mode".. Then the
command can be efficiently executed like this:
ls */queue/id* | xargs $path/afl-harness-cov
For the above-mentioned test on 30k files, the time to generate
coverage data was reduced from 7 minutes to under 30 seconds.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
...to generate a "normal" coverage-instrumented binary, suitable for
use with gcov or afl-cov.
This is slightly annoying because:
- Every object file needs to have been instrumented to work
effectively
- You generally want to have both an afl-instrumented binary and a
gcov-instrumented binary at the same time, but
- gcov instrumentation and afl instrumentation are mutually exclusive
So when making the `afl-cov` target, generate a second set of object
files and a second binary with the `-cov` suffix.
While we're here, remove the redundant x86-emulate.c dependency for
x86-emulate.o.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
When generating coverage output, by default gcov generates output
filenames based only on the coverage file and the "leaf" source file,
not the full path. As a result, it uses the same name for
x86_emulate.c and x86_emulate/x86_emulate.c, generally overwriting the
second (which we actually are about) with the first (which is just a
wrapper).
Rename the user-space wrapper helpers to x86-emulate.[ch], so
that it generates separate files.
There is actually an option to gcov, `--preserve-paths`, which will
cause the full path name to be included in the filename, properly
distinguishing between the two. However, given that the user-space
wrapper doesn't actually do any emulation (and the poor state of gcov
documentation making it difficult to find the option in the first
place), it seems to make more sense to rename the file anyway.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
- Print the symbolic name rather than the number
- Explicitly state when data_read() fails due to EOI
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Commit 849a1f10c9 was checked in inappropriately; review flagged up
that clearerr() was too big a hammer, as it would clear both the EOF
flag and stream errors.
Stream errors shouldn't be cleared; we only want the EOF and other
stream-related state reset. To do this, it is sufficient to fseek()
to zero.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Once feof() returns true for a stream, it will continue to return true
for that stream until clearerr() is called (or the stream is closed
and re-opened).
In llvm-clang-fast-mode, the same file descriptor is used for each
iteration of the loop, meaning that the "Input too large" check was
broken -- feof() would return true even if the fread() hadn't hit the
end of the file. The result is that AFL generates testcases of
arbitrary size.
Fix this by clearing the error after each iteration.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Commit c07574b reorganized the way fuzzing was done, explicitly
creating a structure that the input data would be copied into.
Unfortunately, the cpu register state used by the emulator is on the
stack; it's cleared, but data is never copied into it.
If we're explicitly setting an entirely new cpu_regs struct for each
new input anyway, there's no need to have two copies around anymore;
just point to the one in the data structure.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
fuzz_insn_fetch() is the only data access helper where it is possible
to see offsets larger than 4Gb in 16- or 32-bit modes, as we leave the
incoming rIP untouched in the emulator itself. The check is needed here
as otherwise, after successfully fetching insn bytes, we may end up
zero-extending EIP soon after complete_insn, which collides with the
X86EMUL_EXCEPTION-conditional respective ASSERT() in
x86_emulate_wrapper(). (NB: put_rep_prefix() is what allows
complete_insn to be reached with rc set to other than X86EMUL_OKAY or
X86EMUL_DONE. See also commit 53f87c03b4 ["x86emul: generalize
exception handling for rep_* hooks"].)
Add assert()-s for all other (data) access routines, as effective
address generation in the emulator ought to guarantee in-range values.
For them to not trigger, several adjustments to the emulator's address
calculations are needed: While the DstBitBase one is really mandatory,
the specification allows for either original or new behavior for two-
part accesses. Observed behavior on real hardware, however, is for such
accesses to silently wrap at the 2^^32 boundary in other than 64-bit
mode, just like they do at the 2^^64 boundary in 64-bit mode, which our
code is now being brought in line with. While adding truncate_ea()
invocations there, also convert open coded instances of it.
Reported-by: George Dunlap <george.dunlap@citrix.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Running "make uninstall" does not remove all installed files, a
situation which might cause link related issues if xen is re-installed
in a different location.
In order to make uninstall correctly remove the files it is best
the process should be done recursively by mirroring each "install"
target with an "uninstall" who removes the installed files.
An exception to this rule is uninstalling the files produced by
"qemu-xen-dir-remote" and "qemu-xen-traditional-dir", which are external
to the project. These projects do not implement an "uninstall" target so
the files have to be removed manually.
Signed-off-by: Petre Pircalabu <ppircalabu@bitdefender.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
The amount of namespace resolution is unnecessarily large, as all code deals
in terms of struct segment_register. This removes the attr.fields part of all
references, and alters attr.bytes to just attr.
Three areas of code using initialisers for segment_register are tweaked to
compile with older versions of GCC. arch_set_info_hvm_guest() has its SEG()
macros altered to use plain comma-based initialisation, while
{rm,vm86}_{cs,ds}_attr are simplified to plain numbers which matches their
description in the manuals.
No functional change. (For some reason, the old {rm,vm86}_{cs,ds}_attr causes
GCC to create variable in .rodata, whereas the new code uses immediate
operands. As a result, vmx_{get,set}_segment_register() are slightly
shorter.)
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
c/s 92cf67888 "x86/emul: Hold x86_emulate() to strict X86EMUL_EXCEPTION
requirements" was appropriate for the hypervisor, but the fuzzer stubs didn't
conform to the stricter requirements. AFL is very quick to discover this.
Extend the fuzzing harness exception logic to raise exceptions appropriately.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
x86_emulates()'s is_branch_step() performs a speculative read of
IA32_DEBUGCTL, but doesn't squash exceptions should they arise. In reality,
this MSR is always available.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
The correct prototypes for the hooks are to use enum x86_segment rather than
unsigned int. It is implementation specific as to whether this compiles.
assert() that the emulator never passes an inappropriate segment. The only
hook which may legitimately be passed x86_seg_none is invlpg().
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
The fuzzing harness conditionally disables hooks to test error paths in the
emulator. However, fuzz_emulops is a static structure.
c/s 69f4633 "tools/insn-fuzz: Support AFL's afl-clang-fast mode" introduced
persistent mode, but because fuzz_emulops is static, the clobbering of hooks
accumulates over repeated input, meaning that previous corpora influence the
execution over the current corpus.
Move the partially clobbered struct x86_emulate_ops into struct fuzz_state,
which is re-initialised from full on each call to LLVMFuzzerTestOneInput()
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
AFL has a measure of stability, where it passes the same corpus into the
fuzzing harness and observes whether the execution path changes from before.
Any instability in the fuzzing harness reduces its effectiveness, as an
observed crash may not reliably be caused by the original corpus.
In preparation to fix a stability bug, introduce struct fuzz_state, allocated
on the stack and passed around via struct x86_emulate_ctxt's data parameter.
Propagate ctxt into the helpers such as maybe_fail(), so the state can be
retrieved.
Move the previously-static data_{index,num} into struct fuzz_state.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
For control-flow changes, the emulator needs to perform a zero-length
instruction fetch at the target offset. It also passes NULL for the
destination buffer, as there is no instruction stream to collect.
This trips up UBSAN when passed to memcpy(), as passing NULL is undefined
behaviour per the C spec (irrespective of passing a size of 0).
Special case these fetches in fuzz_insn_fetch() before reaching data_read().
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: George Dunlap <george.dunlap@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Long mode (or not) influences emulation behaviour in a number of cases.
Instead of reusing the ->read_msr() hook to obtain EFER.LMA, require callers
to provide it directly.
This simplifies all long mode checks during emulation to a simple boolean
read, removing embedded msr reads. It also allows for the removal of a local
variable in the sysenter emulation block, and removes a latent bug in the
syscall emulation block where rc contains a non X86EMUL_* constant for a
period of time.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
Acked-by: Tim Deegan <tim@xen.org>
Reviewed-by: Jan Beulich <JBeulich@suse.com>
With the SVM injection logic capable of doing its own emulation, there is no
need for this hardware-specific assistance in the common emulator.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Tim Deegan <tim@xen.org>
When an FPU instruction with a memory destination fails during the
memory write, it should not affect FPU register state. Due to the way
we emulate FPU (and SIMD) instructions, we can only guarantee this by
- backing out changes to the FPU register state in such a case or
- doing a descriptor read and/or page walk up front, perhaps with the
stubs accessing the actual memory location then.
The latter would require a significant change in how the emulator does
its guest memory accessing, so for now the former variant is being
chosen.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Reviewed-by: Paul Durrant <paul.durrant@citrix.com> [hvm/emulate.c]
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
AFL has an alternative llvm-base instrumentation mode, which has much lower
overhead than the traditional afl-gcc.
One extra ability is to chose exactly where the master process gets
initialised to, before being forked for testing. This point is chosen after
the call to LLVMFuzzerInitialize(), so the stack isn't being remapped
executable for every test.
Another extra ability is to feed multiple inputs into a single test process,
to reduce the number of fork() calls required overall. Two caveats are that if
stdin is used for data, it must be unbuffered, and if input is passed via a
command line parameter, the underlying file must be opened and closed on each
iteration.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
libfuzz can perform one-time initialisation by calling LLVMFuzzerInitialize().
Move emul_test_init() into this, to avoid repeating it on every
LLVMFuzzerTestOneInput() call.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
This is rather faster for afl-fuzz to arrange than using an explicit file
parameter. Also update the README to recommend using a tmpfs for findings_dir
which reduces disk load and is more performant.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Anthony PERARD
Jan Beulich
Roger Pau Monné
Andrew Cooper
Christopher Clark
Olaf Hering
Paul Semel
George Dunlap
Petre Pircalabu