Pull x86 paravirt updates from Ingo Molnar:
"Two main changes:
- Remove no longer used parts of the paravirt infrastructure and put
large quantities of paravirt ops under a new config option
PARAVIRT_XXL=y, which is selected by XEN_PV only. (Joergen Gross)
- Enable PV spinlocks on Hyperv (Yi Sun)"
* 'x86-paravirt-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/hyperv: Enable PV qspinlock for Hyper-V
x86/hyperv: Add GUEST_IDLE_MSR support
x86/paravirt: Clean up native_patch()
x86/paravirt: Prevent redefinition of SAVE_FLAGS macro
x86/xen: Make xen_reservation_lock static
x86/paravirt: Remove unneeded mmu related paravirt ops bits
x86/paravirt: Move the Xen-only pv_mmu_ops under the PARAVIRT_XXL umbrella
x86/paravirt: Move the pv_irq_ops under the PARAVIRT_XXL umbrella
x86/paravirt: Move the Xen-only pv_cpu_ops under the PARAVIRT_XXL umbrella
x86/paravirt: Move items in pv_info under PARAVIRT_XXL umbrella
x86/paravirt: Introduce new config option PARAVIRT_XXL
x86/paravirt: Remove unused paravirt bits
x86/paravirt: Use a single ops structure
x86/paravirt: Remove clobbers from struct paravirt_patch_site
x86/paravirt: Remove clobbers parameter from paravirt patch functions
x86/paravirt: Make paravirt_patch_call() and paravirt_patch_jmp() static
x86/xen: Add SPDX identifier in arch/x86/xen files
x86/xen: Link platform-pci-unplug.o only if CONFIG_XEN_PVHVM
x86/xen: Move pv specific parts of arch/x86/xen/mmu.c to mmu_pv.c
x86/xen: Move pv irq related functions under CONFIG_XEN_PV umbrella
It supposed to be safe to modify static branches after jump_label_init().
But, because static key modifying code eventually calls text_poke() it can
end up accessing a struct page which has not been initialized yet.
Here is how to quickly reproduce the problem. Insert code like this
into init/main.c:
| +static DEFINE_STATIC_KEY_FALSE(__test);
| asmlinkage __visible void __init start_kernel(void)
| {
| char *command_line;
|@@ -587,6 +609,10 @@ asmlinkage __visible void __init start_kernel(void)
| vfs_caches_init_early();
| sort_main_extable();
| trap_init();
|+ {
|+ static_branch_enable(&__test);
|+ WARN_ON(!static_branch_likely(&__test));
|+ }
| mm_init();
The following warnings show-up:
WARNING: CPU: 0 PID: 0 at arch/x86/kernel/alternative.c:701 text_poke+0x20d/0x230
RIP: 0010:text_poke+0x20d/0x230
Call Trace:
? text_poke_bp+0x50/0xda
? arch_jump_label_transform+0x89/0xe0
? __jump_label_update+0x78/0xb0
? static_key_enable_cpuslocked+0x4d/0x80
? static_key_enable+0x11/0x20
? start_kernel+0x23e/0x4c8
? secondary_startup_64+0xa5/0xb0
---[ end trace abdc99c031b8a90a ]---
If the code above is moved after mm_init(), no warning is shown, as struct
pages are initialized during handover from memblock.
Use text_poke_early() in static branching until early boot IRQs are enabled
and from there switch to text_poke. Also, ensure text_poke() is never
invoked when unitialized memory access may happen by using adding a
!after_bootmem assertion.
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: steven.sistare@oracle.com
Cc: daniel.m.jordan@oracle.com
Cc: linux@armlinux.org.uk
Cc: schwidefsky@de.ibm.com
Cc: heiko.carstens@de.ibm.com
Cc: john.stultz@linaro.org
Cc: sboyd@codeaurora.org
Cc: hpa@zytor.com
Cc: douly.fnst@cn.fujitsu.com
Cc: peterz@infradead.org
Cc: prarit@redhat.com
Cc: feng.tang@intel.com
Cc: pmladek@suse.com
Cc: gnomes@lxorguk.ukuu.org.uk
Cc: linux-s390@vger.kernel.org
Cc: boris.ostrovsky@oracle.com
Cc: jgross@suse.com
Cc: pbonzini@redhat.com
Link: https://lkml.kernel.org/r/20180719205545.16512-9-pasha.tatashin@oracle.com
Time has come to switch PTI development over to a v4.15 base - we'll still
try to make sure that all PTI fixes backport cleanly to v4.14 and earlier.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull x86 pti updates from Thomas Gleixner:
"This contains:
- a PTI bugfix to avoid setting reserved CR3 bits when PCID is
disabled. This seems to cause issues on a virtual machine at least
and is incorrect according to the AMD manual.
- a PTI bugfix which disables the perf BTS facility if PTI is
enabled. The BTS AUX buffer is not globally visible and causes the
CPU to fault when the mapping disappears on switching CR3 to user
space. A full fix which restores BTS on PTI is non trivial and will
be worked on.
- PTI bugfixes for EFI and trusted boot which make sure that the user
space visible page table entries have the NX bit cleared
- removal of dead code in the PTI pagetable setup functions
- add PTI documentation
- add a selftest for vsyscall to verify that the kernel actually
implements what it advertises.
- a sysfs interface to expose vulnerability and mitigation
information so there is a coherent way for users to retrieve the
status.
- the initial spectre_v2 mitigations, aka retpoline:
+ The necessary ASM thunk and compiler support
+ The ASM variants of retpoline and the conversion of affected ASM
code
+ Make LFENCE serializing on AMD so it can be used as speculation
trap
+ The RSB fill after vmexit
- initial objtool support for retpoline
As I said in the status mail this is the most of the set of patches
which should go into 4.15 except two straight forward patches still on
hold:
- the retpoline add on of LFENCE which waits for ACKs
- the RSB fill after context switch
Both should be ready to go early next week and with that we'll have
covered the major holes of spectre_v2 and go back to normality"
* 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (28 commits)
x86,perf: Disable intel_bts when PTI
security/Kconfig: Correct the Documentation reference for PTI
x86/pti: Fix !PCID and sanitize defines
selftests/x86: Add test_vsyscall
x86/retpoline: Fill return stack buffer on vmexit
x86/retpoline/irq32: Convert assembler indirect jumps
x86/retpoline/checksum32: Convert assembler indirect jumps
x86/retpoline/xen: Convert Xen hypercall indirect jumps
x86/retpoline/hyperv: Convert assembler indirect jumps
x86/retpoline/ftrace: Convert ftrace assembler indirect jumps
x86/retpoline/entry: Convert entry assembler indirect jumps
x86/retpoline/crypto: Convert crypto assembler indirect jumps
x86/spectre: Add boot time option to select Spectre v2 mitigation
x86/retpoline: Add initial retpoline support
objtool: Allow alternatives to be ignored
objtool: Detect jumps to retpoline thunks
x86/pti: Make unpoison of pgd for trusted boot work for real
x86/alternatives: Fix optimize_nops() checking
sysfs/cpu: Fix typos in vulnerability documentation
x86/cpu/AMD: Use LFENCE_RDTSC in preference to MFENCE_RDTSC
...
We use alternatives_text_reserved() to check if the address is in
the fixed pieces of alternative reserved, but the problem is that
we don't hold the smp_alt mutex when call this function. So the list
traversal may encounter a deleted list_head if another path is doing
alternatives_smp_module_del().
One solution is that we can hold smp_alt mutex before call this
function, but the difficult point is that the callers of this
functions, arch_prepare_kprobe() and arch_prepare_optimized_kprobe(),
are called inside the text_mutex. So we must hold smp_alt mutex
before we go into these arch dependent code. But we can't now,
the smp_alt mutex is the arch dependent part, only x86 has it.
Maybe we can export another arch dependent callback to solve this.
But there is a simpler way to handle this problem. We can reuse the
text_mutex to protect smp_alt_modules instead of using another mutex.
And all the arch dependent checks of kprobes are inside the text_mutex,
so it's safe now.
Signed-off-by: Zhou Chengming <zhouchengming1@huawei.com>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Acked-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: bp@suse.de
Fixes: 2cfa197 "ftrace/alternatives: Introducing *_text_reserved functions"
Link: http://lkml.kernel.org/r/1509585501-79466-1-git-send-email-zhouchengming1@huawei.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
In the current form of the code, if a->replacementlen is 0, the reference
to *insnbuf for comparison touches potentially garbage memory. While it
doesn't affect the execution flow due to the subsequent a->replacementlen
comparison, it is (rightly) detected as use of uninitialized memory by a
runtime instrumentation currently under my development, and could be
detected as such by other tools in the future, too (e.g. KMSAN).
Fix the "false-positive" by reordering the conditions to first check the
replacement instruction length before referencing specific opcode bytes.
Signed-off-by: Mateusz Jurczyk <mjurczyk@google.com>
Reviewed-by: Borislav Petkov <bp@suse.de>
Cc: Andy Lutomirski <luto@kernel.org>
Link: http://lkml.kernel.org/r/20170524135500.27223-1-mjurczyk@google.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
We use sync_core() in the alternatives code to stop speculative
execution of prefetched instructions because we are potentially changing
them and don't want to execute stale bytes.
What it does on most machines is call CPUID which is a serializing
instruction. And that's expensive.
However, the instruction cache is serialized when we're on the local CPU
and are changing the data through the same virtual address. So then, we
don't need the serializing CPUID but a simple control flow change. Last
being accomplished with a CALL/RET which the noinline causes.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Andy Lutomirski <luto@kernel.org>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Cc: Matthew Whitehead <tedheadster@gmail.com>
Cc: One Thousand Gnomes <gnomes@lxorguk.ukuu.org.uk>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20161203150258.vwr5zzco7ctgc4pe@pd.tnic
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Richard reported the following crash:
[ 0.036000] BUG: unable to handle kernel paging request at 55501e06
[ 0.036000] IP: [<c0aae48b>] common_interrupt+0xb/0x38
[ 0.036000] Call Trace:
[ 0.036000] [<c0409c80>] ? add_nops+0x90/0xa0
[ 0.036000] [<c040a054>] apply_alternatives+0x274/0x630
Chuck decoded:
" 0: 8d 90 90 83 04 24 lea 0x24048390(%eax),%edx
6: 80 fc 0f cmp $0xf,%ah
9: a8 0f test $0xf,%al
>> b: a0 06 1e 50 55 mov 0x55501e06,%al
10: 57 push %edi
11: 56 push %esi
Interrupt 0x30 occurred while the alternatives code was replacing the
initial 0x90,0x90,0x90 NOPs (from the ASM_CLAC macro) with the
optimized version, 0x8d,0x76,0x00. Only the first byte has been
replaced so far, and it makes a mess out of the insn decoding."
optimize_nops() is buggy in two aspects:
- It's not disabling interrupts across the modification
- It's lacking a sync_core() call
Add both.
Fixes: 4fd4b6e553 'x86/alternatives: Use optimized NOPs for padding'
Reported-and-tested-by: "Richard W.M. Jones" <rjones@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Richard W.M. Jones <rjones@redhat.com>
Cc: Chuck Ebbert <cebbert.lkml@gmail.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/alpine.DEB.2.11.1509031232340.15006@nanos
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Pull x86 core updates from Ingo Molnar:
"There were so many changes in the x86/asm, x86/apic and x86/mm topics
in this cycle that the topical separation of -tip broke down somewhat -
so the result is a more traditional architecture pull request,
collected into the 'x86/core' topic.
The topics were still maintained separately as far as possible, so
bisectability and conceptual separation should still be pretty good -
but there were a handful of merge points to avoid excessive
dependencies (and conflicts) that would have been poorly tested in the
end.
The next cycle will hopefully be much more quiet (or at least will
have fewer dependencies).
The main changes in this cycle were:
* x86/apic changes, with related IRQ core changes: (Jiang Liu, Thomas
Gleixner)
- This is the second and most intrusive part of changes to the x86
interrupt handling - full conversion to hierarchical interrupt
domains:
[IOAPIC domain] -----
|
[MSI domain] --------[Remapping domain] ----- [ Vector domain ]
| (optional) |
[HPET MSI domain] ----- |
|
[DMAR domain] -----------------------------
|
[Legacy domain] -----------------------------
This now reflects the actual hardware and allowed us to distangle
the domain specific code from the underlying parent domain, which
can be optional in the case of interrupt remapping. It's a clear
separation of functionality and removes quite some duct tape
constructs which plugged the remap code between ioapic/msi/hpet
and the vector management.
- Intel IOMMU IRQ remapping enhancements, to allow direct interrupt
injection into guests (Feng Wu)
* x86/asm changes:
- Tons of cleanups and small speedups, micro-optimizations. This
is in preparation to move a good chunk of the low level entry
code from assembly to C code (Denys Vlasenko, Andy Lutomirski,
Brian Gerst)
- Moved all system entry related code to a new home under
arch/x86/entry/ (Ingo Molnar)
- Removal of the fragile and ugly CFI dwarf debuginfo annotations.
Conversion to C will reintroduce many of them - but meanwhile
they are only getting in the way, and the upstream kernel does
not rely on them (Ingo Molnar)
- NOP handling refinements. (Borislav Petkov)
* x86/mm changes:
- Big PAT and MTRR rework: making the code more robust and
preparing to phase out exposing direct MTRR interfaces to drivers -
in favor of using PAT driven interfaces (Toshi Kani, Luis R
Rodriguez, Borislav Petkov)
- New ioremap_wt()/set_memory_wt() interfaces to support
Write-Through cached memory mappings. This is especially
important for good performance on NVDIMM hardware (Toshi Kani)
* x86/ras changes:
- Add support for deferred errors on AMD (Aravind Gopalakrishnan)
This is an important RAS feature which adds hardware support for
poisoned data. That means roughly that the hardware marks data
which it has detected as corrupted but wasn't able to correct, as
poisoned data and raises an APIC interrupt to signal that in the
form of a deferred error. It is the OS's responsibility then to
take proper recovery action and thus prolonge system lifetime as
far as possible.
- Add support for Intel "Local MCE"s: upcoming CPUs will support
CPU-local MCE interrupts, as opposed to the traditional system-
wide broadcasted MCE interrupts (Ashok Raj)
- Misc cleanups (Borislav Petkov)
* x86/platform changes:
- Intel Atom SoC updates
... and lots of other cleanups, fixlets and other changes - see the
shortlog and the Git log for details"
* 'x86-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (222 commits)
x86/hpet: Use proper hpet device number for MSI allocation
x86/hpet: Check for irq==0 when allocating hpet MSI interrupts
x86/mm/pat, drivers/infiniband/ipath: Use arch_phys_wc_add() and require PAT disabled
x86/mm/pat, drivers/media/ivtv: Use arch_phys_wc_add() and require PAT disabled
x86/platform/intel/baytrail: Add comments about why we disabled HPET on Baytrail
genirq: Prevent crash in irq_move_irq()
genirq: Enhance irq_data_to_desc() to support hierarchy irqdomain
iommu, x86: Properly handle posted interrupts for IOMMU hotplug
iommu, x86: Provide irq_remapping_cap() interface
iommu, x86: Setup Posted-Interrupts capability for Intel iommu
iommu, x86: Add cap_pi_support() to detect VT-d PI capability
iommu, x86: Avoid migrating VT-d posted interrupts
iommu, x86: Save the mode (posted or remapped) of an IRTE
iommu, x86: Implement irq_set_vcpu_affinity for intel_ir_chip
iommu: dmar: Provide helper to copy shared irte fields
iommu: dmar: Extend struct irte for VT-d Posted-Interrupts
iommu: Add new member capability to struct irq_remap_ops
x86/asm/entry/64: Disentangle error_entry/exit gsbase/ebx/usermode code
x86/asm/entry/32: Shorten __audit_syscall_entry() args preparation
x86/asm/entry/32: Explain reloading of registers after __audit_syscall_entry()
...
Take a look at the first instruction byte before optimizing the NOP -
there might be something else there already, like the ALTERNATIVE_2()
in rdtsc_barrier() which NOPs out on AMD even though we just
patched in an MFENCE.
This happens because the alternatives sees X86_FEATURE_MFENCE_RDTSC,
AMD CPUs set it, we patch in the MFENCE and right afterwards it sees
X86_FEATURE_LFENCE_RDTSC which AMD CPUs don't set and we blindly
optimize the NOP.
Checking whether at least the first byte is 0x90 prevents that.
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1428181662-18020-1-git-send-email-bp@alien8.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Quentin caught a corner case with the generation of instruction
padding in the ALTERNATIVE_2 macro: if len(orig_insn) <
len(alt1) < len(alt2), then not enough padding gets added and
that is not good(tm) as we could overwrite the beginning of the
next instruction.
Luckily, at the time of this writing, we don't have
ALTERNATIVE_2() invocations which have that problem and even if
we did, a simple fix would be to prepend the instructions with
enough prefixes so that that corner case doesn't happen.
However, best it would be if we fixed it properly. See below for
a simple, abstracted example of what we're doing.
So what we ended up doing is, we compute the
max(len(alt1), len(alt2)) - len(orig_insn)
and feed that value to the .skip gas directive. The max() cannot
have conditionals due to gas limitations, thus the fancy integer
math.
With this patch, all ALTERNATIVE_2 sites get padded correctly;
generating obscure test cases pass too:
#define alt_max_short(a, b) ((a) ^ (((a) ^ (b)) & -(-((a) < (b)))))
#define gen_skip(orig, alt1, alt2, marker) \
.skip -((alt_max_short(alt1, alt2) - (orig)) > 0) * \
(alt_max_short(alt1, alt2) - (orig)),marker
.pushsection .text, "ax"
.globl main
main:
gen_skip(1, 2, 4, 0x09)
gen_skip(4, 1, 2, 0x10)
...
.popsection
Thanks to Quentin for catching it and double-checking the fix!
Reported-by: Quentin Casasnovas <quentin.casasnovas@oracle.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20150404133443.GE21152@pd.tnic
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Alternatives allow now for an empty old instruction. In this case we go
and pad the space with NOPs at assembly time. However, there are the
optimal, longer NOPs which should be used. Do that at patching time by
adding alt_instr.padlen-sized NOPs at the old instruction address.
Cc: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Borislav Petkov <bp@suse.de>
Up until now we had to pay attention to relative JMPs in alternatives
about how their relative offset gets computed so that the jump target
is still correct. Or, as it is the case for near CALLs (opcode e8), we
still have to go and readjust the offset at patching time.
What is more, the static_cpu_has_safe() facility had to forcefully
generate 5-byte JMPs since we couldn't rely on the compiler to generate
properly sized ones so we had to force the longest ones. Worse than
that, sometimes it would generate a replacement JMP which is longer than
the original one, thus overwriting the beginning of the next instruction
at patching time.
So, in order to alleviate all that and make using JMPs more
straight-forward we go and pad the original instruction in an
alternative block with NOPs at build time, should the replacement(s) be
longer. This way, alternatives users shouldn't pay special attention
so that original and replacement instruction sizes are fine but the
assembler would simply add padding where needed and not do anything
otherwise.
As a second aspect, we go and recompute JMPs at patching time so that we
can try to make 5-byte JMPs into two-byte ones if possible. If not, we
still have to recompute the offsets as the replacement JMP gets put far
away in the .altinstr_replacement section leading to a wrong offset if
copied verbatim.
For example, on a locally generated kernel image
old insn VA: 0xffffffff810014bd, CPU feat: X86_FEATURE_ALWAYS, size: 2
__switch_to:
ffffffff810014bd: eb 21 jmp ffffffff810014e0
repl insn: size: 5
ffffffff81d0b23c: e9 b1 62 2f ff jmpq ffffffff810014f2
gets corrected to a 2-byte JMP:
apply_alternatives: feat: 3*32+21, old: (ffffffff810014bd, len: 2), repl: (ffffffff81d0b23c, len: 5)
alt_insn: e9 b1 62 2f ff
recompute_jumps: next_rip: ffffffff81d0b241, tgt_rip: ffffffff810014f2, new_displ: 0x00000033, ret len: 2
converted to: eb 33 90 90 90
and a 5-byte JMP:
old insn VA: 0xffffffff81001516, CPU feat: X86_FEATURE_ALWAYS, size: 2
__switch_to:
ffffffff81001516: eb 30 jmp ffffffff81001548
repl insn: size: 5
ffffffff81d0b241: e9 10 63 2f ff jmpq ffffffff81001556
gets shortened into a two-byte one:
apply_alternatives: feat: 3*32+21, old: (ffffffff81001516, len: 2), repl: (ffffffff81d0b241, len: 5)
alt_insn: e9 10 63 2f ff
recompute_jumps: next_rip: ffffffff81d0b246, tgt_rip: ffffffff81001556, new_displ: 0x0000003e, ret len: 2
converted to: eb 3e 90 90 90
... and so on.
This leads to a net win of around
40ish replacements * 3 bytes savings =~ 120 bytes of I$
on an AMD guest which means some savings of precious instruction cache
bandwidth. The padding to the shorter 2-byte JMPs are single-byte NOPs
which on smart microarchitectures means discarding NOPs at decode time
and thus freeing up execution bandwidth.
Signed-off-by: Borislav Petkov <bp@suse.de>
Linus Torvalds
Pu Wen
Juergen Gross
Jiri Kosina
Pavel Tatashin
Josh Poimboeuf
Ingo Molnar
Borislav Petkov
Zhou Chengming
Peter Zijlstra
Mateusz Jurczyk
Andy Lutomirski
Thomas Gleixner