This is patch complements D55117 implementing __hwasan_mem*
functions in runtime
Differential revision: https://reviews.llvm.org/D55554
llvm-svn: 349730
As of r349413 it's now possible for a binary to contain an empty
hwasan frame section. Handle that case simply by doing nothing.
Differential Revision: https://reviews.llvm.org/D55796
llvm-svn: 349428
Summary:
This is a follow up patch to r346956 for the `SizeClassAllocator32`
allocator.
This patch makes `AddressSpaceView` a template parameter both to the
`ByteMap` implementations (but makes `LocalAddressSpaceView` the
default), some `AP32` implementations and is used in `SizeClassAllocator32`.
The actual changes to `ByteMap` implementations and
`SizeClassAllocator32` are very simple. However the patch is large
because it requires changing all the `AP32` definitions, and users of
those definitions.
For ASan and LSan we make `AP32` and `ByteMap` templateds type that take
a single `AddressSpaceView` argument. This has been done because we will
instantiate the allocator with a type that isn't `LocalAddressSpaceView`
in the future patches. For the allocators used in the other sanitizers
(i.e. HWAsan, MSan, Scudo, and TSan) use of `LocalAddressSpaceView` is
hard coded because we do not intend to instantiate the allocators with
any other type.
In the cases where untemplated types have become templated on a single
`AddressSpaceView` parameter (e.g. `PrimaryAllocator`) their name has
been changed to have a `ASVT` suffix (Address Space View Type) to
indicate they are templated. The only exception to this are the `AP32`
types due to the desire to keep the type name as short as possible.
In order to check that template is instantiated in the correct a way a
`static_assert(...)` has been added that checks that the
`AddressSpaceView` type used by `Params::ByteMap::AddressSpaceView` matches
the `Params::AddressSpaceView`. This uses the new `sanitizer_type_traits.h`
header.
rdar://problem/45284065
Reviewers: kcc, dvyukov, vitalybuka, cryptoad, eugenis, kubamracek, george.karpenkov
Subscribers: mgorny, llvm-commits, #sanitizers
Differential Revision: https://reviews.llvm.org/D54904
llvm-svn: 349138
Summary:
Add a check that TLS_SLOT_TSAN / TLS_SLOT_SANITIZER, whichever
android_get_tls_slot is using, is not conflicting with
TLS_SLOT_DLERROR.
Reviewers: rprichard, vitalybuka
Subscribers: srhines, kubamracek, llvm-commits
Differential Revision: https://reviews.llvm.org/D55587
llvm-svn: 348979
Summary:
With free_checks_tail_magic=1 (default) HWASAN
writes magic bytes to the tail of every heap allocation
(last bytes of the last granule, if the last granule is not fully used)
and checks these bytes on free().
This feature will detect buffer overwires within the last granule
at the time of free().
This is an alternative to malloc_align_right=[1289] that should have
fewer compatibility issues. It is also weaker since it doesn't
detect read overflows and reports bugs at free() instead of at access.
Reviewers: eugenis
Subscribers: kubamracek, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D54656
llvm-svn: 347116
Summary:
... so that we can find intra-granule buffer overflows.
The default is still to always align left.
It remains to be seen wether we can enable this mode at scale.
Reviewers: eugenis
Reviewed By: eugenis
Subscribers: jfb, dvyukov, kubamracek, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D53789
llvm-svn: 347082
Summary:
When reporting a fatal error, collect and add the entire report text to
android_set_abort_message so that it can be found in the tombstone.
Reviewers: kcc, vitalybuka
Subscribers: srhines, kubamracek, llvm-commits
Differential Revision: https://reviews.llvm.org/D54284
llvm-svn: 346557
We have seen failing builds due to a race condition between
RTAsan_dynamic and libc++ headers builds, specifically libc++
headers depend on __config and if this header hasn't been copied
into the final location, including other headers will typically
result in failure. To avoid this race, we add an explicit dependency
on libc++ headers which ensures that they've been copied into place
before the sanitizer object library build starts.
Differential Revision: https://reviews.llvm.org/D54198
llvm-svn: 346339
Summary:
At compile-time, create an array of {PC,HumanReadableStackFrameDescription}
for every function that has an instrumented frame, and pass this array
to the run-time at the module-init time.
Similar to how we handle pc-table in SanitizerCoverage.
The run-time is dummy, will add the actual logic in later commits.
Reviewers: morehouse, eugenis
Reviewed By: eugenis
Subscribers: srhines, llvm-commits, kubamracek
Differential Revision: https://reviews.llvm.org/D53227
llvm-svn: 344985
Summary:
GetStackTrace treats top PC as a return address from an error reporting
function, and adjusts it down by 1 instruction. This is not necessary in
a signal handler, so adjust PC up to compensate.
Reviewers: kcc, vitalybuka, jfb
Subscribers: kubamracek, llvm-commits
Differential Revision: https://reviews.llvm.org/D52802
llvm-svn: 343638
Summary:
This essentially reverts r337010 since it breaks UBSan, which is used
for a few platform libraries. The "-z global" flag is now added for
Scudo as well. The only other sanitizer shared libraries are for asan
and hwasan, which have also been reinstated to use the global flag.
Reviewers: cryptoad, eugenis
Reviewed By: cryptoad
Subscribers: kubamracek, mgorny, delcypher, #sanitizers, nickdesaulniers, chh, kongyi, pirama, llvm-commits
Differential Revision: https://reviews.llvm.org/D52770
llvm-svn: 343599
Summary:
Display a list of recent stack frames (not a stack trace!) when
tag-mismatch is detected on a stack address.
The implementation uses alignment tricks to get both the address of
the history buffer, and the base address of the shadow with a single
8-byte load. See the comment in hwasan_thread_list.h for more
details.
Developed in collaboration with Kostya Serebryany.
Reviewers: kcc
Subscribers: srhines, kubamracek, mgorny, hiraditya, jfb, llvm-commits
Differential Revision: https://reviews.llvm.org/D52249
llvm-svn: 342923
Summary:
Display a list of recent stack frames (not a stack trace!) when
tag-mismatch is detected on a stack address.
The implementation uses alignment tricks to get both the address of
the history buffer, and the base address of the shadow with a single
8-byte load. See the comment in hwasan_thread_list.h for more
details.
Developed in collaboration with Kostya Serebryany.
Reviewers: kcc
Subscribers: srhines, kubamracek, mgorny, hiraditya, jfb, llvm-commits
Differential Revision: https://reviews.llvm.org/D52249
llvm-svn: 342921
Summary:
When building without COMPILER_RT_HWASAN_WITH_INTERCEPTORS, skip
interceptors for malloc/free/etc and only export their versions with
__sanitizer_ prefix.
Also remove a hack in mallinfo() interceptor that does not apply to
hwasan.
Reviewers: kcc
Subscribers: kubamracek, krytarowski, llvm-commits
Differential Revision: https://reviews.llvm.org/D51711
llvm-svn: 341598
Sigtrap is used for error reporting, but all other signals are better
left for the platform.
In particular, sanitizer signal handlers do not dump registers or
memory which makes debugging harder for no good reason.
llvm-svn: 341500
Summary:
We need this in order to properly report heap-use-after-free,
since we don't have a quarantine.
This is a first part of the code, more like a proof of concept.
But I'd like to commit at as is and proceed with refactoring,
adding a ThreadRegistry, and extending the functionality.
Reviewers: eugenis
Subscribers: kubamracek, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D51394
llvm-svn: 340971
Summary:
The previous version of the patch makes some code unable to distinguish
failure to map address 0 and error.
Revert to turn the bots back to green while figuring out a new approach.
Reviewers: eugenis
Reviewed By: eugenis
Subscribers: kubamracek, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D51451
llvm-svn: 340957
Summary:
`MmapNoAccess` & `MmapFixedNoAccess` return directly the result of
`internal_mmap`, as opposed to other Mmap functions that return nullptr.
This inconsistency leads to some confusion for the callers, as some check for
`~(uptr)0` (`MAP_FAILED`) for failure (while it can fail with `-ENOMEM` for
example).
Two potential solutions: change the callers, or make the functions return
`nullptr` on failure to follow the precedent set by the other functions.
The second option looked more appropriate to me.
Correct the callers that were wrongly checking for `~(uptr)0` or
`MAP_FAILED`.
TODO for follow up CLs:
- There are a couple of `internal_mmap` calls in XRay that check for
MMAP_FAILED as a result as well (cc: @dberris); they should use
`internal_iserror`;
Reviewers: eugenis, alekseyshl, dberris, kubamracek
Reviewed By: alekseyshl
Subscribers: kristina, kubamracek, delcypher, #sanitizers, dberris, llvm-commits
Differential Revision: https://reviews.llvm.org/D50940
llvm-svn: 340576
Summary:
The idea behind this change is to allow sanitization of libc. We are prototyping on Bionic,
but the tool interface will be general enough (or at least generalizable) to support any other libc.
When libc depends on libclang_rt.hwasan, the latter can not interpose libc functions.
In fact, majority of interceptors become unnecessary when libc code is instrumented.
This change gets rid of most hwasan interceptors and provides interface for libc to notify
hwasan about thread creation and destruction events. Some interceptors (pthread_create)
are kept under #ifdef to enable testing with uninstrumented libc. They are expressed in
terms of the new libc interface.
The new cmake switch, COMPILER_RT_HWASAN_WITH_INTERCEPTORS, ON by default, builds testing
version of the library with the aforementioned pthread_create interceptor.
With the OFF setting, the library becomes more of a libc plugin.
Reviewers: vitalybuka, kcc, jfb
Subscribers: srhines, kubamracek, mgorny, jfb, llvm-commits
Differential Revision: https://reviews.llvm.org/D50922
llvm-svn: 340216
This reapplies commit r339935 with the following changes:
* make longjmp test C, not C++, to avoid dependency on libc++/libstdc++
* untag pointer in memset interceptor
x86_64 does not have TBI, so hwasan barely works there. Tests must be carefully
written in a way that does not leak tagged pointer to system libraries.
llvm-svn: 339963
Summary:
A callback to annotate longjmp-like code.
Unlike __asan_handle_no_return, in hwasan we can not conservatively
"unpoison" the entire thread stack, because there is no such thing as
unpoisoned memory. Pointer and memory tags must always match.
Reviewers: vitalybuka, kcc
Subscribers: kubamracek, llvm-commits
Differential Revision: https://reviews.llvm.org/D50752
llvm-svn: 339935
Summary:
Export __sanitizer_malloc, etc as aliases to malloc, etc.
This way users can wrap sanitizer malloc, even in fully static binaries.
Both jemalloc and tcmalloc provide similar aliases (je_* and tc_*).
Reviewers: vitalybuka, kcc
Subscribers: llvm-commits, kubamracek
Differential Revision: https://reviews.llvm.org/D50570
llvm-svn: 339614
Summary:
Don't crash when /proc/self/maps is inaccessible from main thread.
It's not a big deal, really.
Reviewers: vitalybuka, kcc
Subscribers: kubamracek, llvm-commits
Differential Revision: https://reviews.llvm.org/D50574
llvm-svn: 339607
Summary:
Provide __hwasan_shadow_init that can be used to initialize shadow w/o touching libc.
It can be used to bootstrap an unusual case of fully-static executable with
hwasan-instrumented libc, which needs to run hwasan code before it is ready to serve
user calls like madvise().
Reviewers: vitalybuka, kcc
Subscribers: kubamracek, llvm-commits
Differential Revision: https://reviews.llvm.org/D50581
llvm-svn: 339606
HWASan will not run on older Android releases where we use
__android_log_write for logging.
This dependency is also harmful in the case when libc itself depends
on hwasan, because it creates a loop of
libc -> hwasan -> liblog -> libc
which makes liblog vs libc initialization order undetermined.
Without liblog the loop is just
libc -> hwasan -> libc
and any init order issues can be solved in hwasan.
llvm-svn: 339449
MmapFixedNoReserve does not terminate process on failure.
Failure to check its result and die will always lead to harder
to debug crashes later in execution. This was observed in Go
processes due to some address space conflicts.
Consistently check result of MmapFixedNoReserve.
While we are here also add warn_unused_result attribute
to prevent such bugs in future and change return type to bool
as that's what all callers want.
Reviewed in https://reviews.llvm.org/D49367
llvm-svn: 337531
Summary:
Use `-Wl,-z,global` for all Sanitizer shared libraries on
Android. We want them to be in the global group
(https://android.googlesource.com/platform/bionic/+/master/android-changes-for-ndk-developers.md#changes-to-library-search-order)
to avoid any alloc/dealloc mismatch between the libc allocator & said library.
`audioserver` was one of the binary that exhibited the problem with Scudo,
this seems to fix it.
[edited for accuracy]
Reviewers: eugenis, alekseyshl
Reviewed By: eugenis
Subscribers: kubamracek, srhines, mgorny, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D49198
llvm-svn: 337010
when building with an IDE so that header files show up in the UI.
This massively improves the development workflow in IDEs.
To implement this a new function `compiler_rt_process_sources(...)` has
been added that adds header files to the list of sources when the
generator is an IDE. For non-IDE generators (e.g. Ninja/Makefile) no
changes are made to the list of source files.
The function can be passed a list of headers via the
`ADDITIONAL_HEADERS` argument. For each runtime library a list of
explicit header files has been added and passed via
`ADDITIONAL_HEADERS`. For `tsan` and `sanitizer_common` a list of
headers was already present but it was stale and has been updated
to reflect the current state of the source tree.
The original version of this patch used file globbing (`*.{h,inc,def}`)
to find the headers but the approach was changed due to this being a
CMake anti-pattern (if the list of headers changes CMake won't
automatically re-generate if globbing is used).
The LLVM repo contains a similar function named `llvm_process_sources()`
but we don't use it here for several reasons:
* It depends on the `LLVM_ENABLE_OPTION` cache variable which is
not set in standalone compiler-rt builds.
* We would have to `include(LLVMProcessSources)` which I'd like to
avoid because it would include a bunch of stuff we don't need.
Differential Revision: https://reviews.llvm.org/D48422
llvm-svn: 336663
Summary:
Currently many allocator specific errors (OOM, for example) are reported as
a text message and CHECK(0) termination, not stack, no details, not too
helpful nor informative. To improve the situation, detailed and
structured errors were defined and reported under the appropriate conditions.
Reviewers: eugenis
Subscribers: kubamracek, delcypher, llvm-commits, #sanitizers
Differential Revision: https://reviews.llvm.org/D47798
llvm-svn: 334248
Fuchsia is no longer treated as UNIX which means we need to explicitly
enable building of shared versions of runtimes.
Differential Revision: https://reviews.llvm.org/D46609
llvm-svn: 331922
Retire the fixed shadow memory mapping to avoid conflicts with default
process memory mapping (currently manifests on Android).
Tests on AArch64 show <1% performance loss and code size increase,
making it possible to use dynamic shadow memory by default.
Keep the fixed shadow memory mapping around to be able to run
performance comparison tests later.
Re-commiting D45847 with fixed shadow for x86-64.
llvm-svn: 330624
This commit causes internal errors with ld.bfd 2.24. My guess is that
the ifunc usage in this commit is causing problems. This is the default
system linker on Trusty Tahr, which is from 2014. I claim it's still in
our support window. Maybe we will decide to drop support for it, but
let's get the bots green while we do the investigation and have that
discussion.
Discovered here: https://crbug.com/835864
llvm-svn: 330619
Summary:
Retire the fixed shadow memory mapping to avoid conflicts with default
process memory mapping (currently manifests on Android).
Tests on AArch64 show <1% performance loss and code size increase,
making it possible to use dynamic shadow memory by default.
For the simplicity and unifirmity sake, use dynamic shadow memory mapping
with base address accessed via ifunc resolver on all supported platforms.
Keep the fixed shadow memory mapping around to be able to run
performance comparison tests later.
Complementing D45840.
Reviewers: eugenis
Subscribers: srhines, kubamracek, dberris, mgorny, kristof.beyls, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D45847
llvm-svn: 330474
Summary:
"N" suffix is added by the instrumentation and interface functions
are expected to be exported from the library as __hwasan_loadN* and
__hwasan_storeN*.
Reviewers: eugenis
Subscribers: kubamracek, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D45739
llvm-svn: 330297
Summary:
Host symbolizer & stacktraces related code in their own RT:
`RTSanitizerCommonSymbolizer`, which is "libcdep" by nature. Symbolizer &
stacktraces specific code that used to live in common files is moved to a new
file `sanitizer_symbolizer_report.cc` as is.
The purpose of this is the enforce a separation between code that relies on
symbolization and code that doesn't. This saves the inclusion of spurious code
due to the interface functions with default visibility, and the extra data
associated.
The following sanitizers makefiles were modified & tested locally:
- dfsan: doesn't require the new symbolizer RT
- esan: requires it
- hwasan: requires it
- lsan: requires it
- msan: requires it
- safestack: doesn't require it
- xray: doesn't require it
- tsan: requires it
- ubsan: requires it
- ubsan_minimal: doesn't require it
- scudo: requires it (but not for Fuchsia that has a minimal runtime)
This was tested locally on Linux, Android, Fuchsia.
Reviewers: alekseyshl, eugenis, dberris, kubamracek, vitalybuka, dvyukov, mcgrathr
Reviewed By: alekseyshl, vitalybuka
Subscribers: srhines, kubamracek, mgorny, krytarowski, delcypher, llvm-commits, #sanitizers
Differential Revision: https://reviews.llvm.org/D45457
llvm-svn: 330131
Summary:
Porting HWASan to Linux x86-64, first of the three patches, compiler-rt part.
The approach is similar to ARM case, trap signal is used to communicate
memory tag check failure. int3 instruction is used to generate a signal,
access parameters are stored in nop [eax + offset] instruction immediately
following the int3 one
Had to add HWASan init on malloc because, due to much less interceptors
defined (most other sanitizers intercept much more and get initalized
via one of those interceptors or don't care about malloc), HWASan was not
initialized yet when libstdc++ was trying to allocate memory for its own
fixed-size heap, which led to CHECK-fail in AllocateFromLocalPool.
Also added the CHECK() failure handler with more detailed message and
stack reporting.
Reviewers: eugenis
Subscribers: kubamracek, dberris, mgorny, kristof.beyls, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D44705
llvm-svn: 328385
Summary:
`sanitizer_common`'s coverage support is fairly well separated, and libcdep by
default. Several sanitizers don't make use of coverage, and as far as I can
tell do no benefit from the extra dependencies pulled in by the coverage public
interface functions.
The following sanitizers call `InitializeCoverage` explicitely: MSan, ASan,
LSan, HWAsan, UBSan. On top of this, any sanitizer bundling RTUBSan should
add the coverage RT as well: ASan, Scudo, UBSan, CFI (diag), TSan, MSan, HWAsan.
So in the end the following have no need: DFSan, ESan, CFI, SafeStack (nolibc
anyway), XRay, and the upcoming Scudo minimal runtime.
I tested this with all the sanitizers check-* with gcc & clang, and in
standalone on Linux & Android, and there was no issue. I couldn't test this on
Mac, Fuchsia, BSDs, & Windows for lack of an environment, so adding a bunch of
people for additional scrunity. I couldn't test HWAsan either.
Reviewers: eugenis, vitalybuka, alekseyshl, flowerhack, kubamracek, dberris, rnk, krytarowski
Reviewed By: vitalybuka, alekseyshl, flowerhack, dberris
Subscribers: mgorny, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D44701
llvm-svn: 328204
I can't reproduce this build error locally, but it appears
straightforward enough to fix. r326851 renamed two of the params of this
interceptor, but apparently to update their use here.
Failure:
http://lab.llvm.org:8011/builders/clang-cmake-aarch64-full/builds/4569
llvm-svn: 326876
This patch changes hwasan inline instrumentation:
Fixes address untagging for shadow address calculation (use 0xFF instead of 0x00 for the top byte).
Emits brk instruction instead of hlt for the kernel and user space.
Use 0x900 instead of 0x100 for brk immediate (0x100 - 0x800 are unavailable in the kernel).
Fixes and adds appropriate tests.
Patch by Andrey Konovalov.
Differential Revision: https://reviews.llvm.org/D43135
llvm-svn: 325711
Summary:
Implement the skeleton of NetBSD syscall hooks for use with sanitizers.
Add a script that generates the rules to handle syscalls
on NetBSD: generate_netbsd_syscalls.awk. It has been written
in NetBSD awk(1) (patched nawk) and is compatible with gawk.
Generate lib/sanitizer_common/sanitizer_platform_limits_netbsd.h
that is a public header for applications, and included as:
<sanitizer_common/sanitizer_platform_limits_netbsd.h>.
Generate sanitizer_syscalls_netbsd.inc that defines all the
syscall rules for NetBSD. This file is modeled after the Linux
specific file: sanitizer_common_syscalls.inc.
Start recognizing NetBSD syscalls with existing sanitizers:
ASan, ESan, HWASan, TSan, MSan.
Sponsored by <The NetBSD Foundation>
Reviewers: joerg, vitalybuka, kcc, dvyukov, eugenis
Reviewed By: vitalybuka
Subscribers: hintonda, kubamracek, mgorny, llvm-commits, #sanitizers
Tags: #sanitizers
Differential Revision: https://reviews.llvm.org/D42048
llvm-svn: 325206
Summary:
Make common allocator agnostic to failure handling modes and move the
decision up to the particular sanitizer's allocator, where the context
is available (call stack, parameters, return nullptr/crash mode etc.)
It simplifies the common allocator and allows the particular sanitizer's
allocator to generate more specific and detailed error reports (which
will be implemented later).
The behavior is largely the same, except one case, the violation of the
common allocator's check for "size + alignment" overflow is now reportied
as OOM instead of "bad request". It feels like a worthy tradeoff and
"size + alignment" is huge in this case anyway (thus, can be interpreted
as not enough memory to satisfy the request). There's also a Report()
statement added there.
Reviewers: eugenis
Subscribers: kubamracek, llvm-commits, #sanitizers
Differential Revision: https://reviews.llvm.org/D42198
llvm-svn: 322784
Summary: -fPIE can not be used when building a shared library.
Reviewers: alekseyshl, peter.smith
Subscribers: kubamracek, llvm-commits, mgorny
Differential Revision: https://reviews.llvm.org/D42121
llvm-svn: 322588
Summary:
Very basic stack instrumentation using tagged pointers.
Tag for N'th alloca in a function is built as XOR of:
* base tag for the function, which is just some bits of SP (poor
man's random)
* small constant which is a function of N.
Allocas are aligned to 16 bytes. On every ReturnInst allocas are
re-tagged to catch use-after-return.
This implementation has a bunch of issues that will be taken care of
later:
1. lifetime intrinsics referring to tagged pointers are not
recognized in SDAG. This effectively disables stack coloring.
2. Generated code is quite inefficient. There is one extra
instruction at each memory access that adds the base tag to the
untagged alloca address. It would be better to keep tagged SP in a
callee-saved register and address allocas as an offset of that XOR
retag, but that needs better coordination between hwasan
instrumentation pass and prologue/epilogue insertion.
3. Lifetime instrinsics are ignored and use-after-scope is not
implemented. This would be harder to do than in ASan, because we
need to use a differently tagged pointer depending on which
lifetime.start / lifetime.end the current instruction is dominated
/ post-dominated.
Reviewers: kcc, alekseyshl
Subscribers: srhines, kubamracek, javed.absar, hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D41602
llvm-svn: 322324
Summary:
Avoid flaky test failures by by using a monotonic number sequence of
heap tags.
Does not affect stack tags: the way we generate those guarantees
uniqueness for at least 30-something first allocas in any function,
as well as the UAR tag.
Reviewers: alekseyshl, kcc
Subscribers: llvm-commits, kubamracek
Differential Revision: https://reviews.llvm.org/D41882
llvm-svn: 322214
Summary: Very similar to AddressSanitizer, with the exception of the error type encoding.
Reviewers: kcc, alekseyshl
Subscribers: cfe-commits, kubamracek, llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D41417
llvm-svn: 321203
Summary: This brings CPU overhead on bzip2 down from 5.5x to 2x.
Reviewers: kcc, alekseyshl
Subscribers: kubamracek, hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D41137
llvm-svn: 320538
Eugene Leviant