os_trace turns out to be a macro that creates static object.
Function-static objects use __cxa_atexit and __dso_handle
which are not present in Go runtime.
llvm-svn: 267720
Ifdef out global variables with destructors.
This requires runtime support that is not provided by Go runtime
(in particular _dso_handle symbol).
llvm-svn: 267709
Current interface assumes that Go calls ProcWire/ProcUnwire
to establish the association between thread and proc.
With the wisdom of hindsight, this interface does not work
very well. I had to sprinkle Go scheduler with wire/unwire
calls, and any mistake leads to hard to debug crashes.
This is not something one wants to maintian.
Fortunately, there is a simpler solution. We can ask Go
runtime as to what is the current Processor, and that
question is very easy to answer on Go side.
Switch to such interface.
llvm-svn: 267703
tsan_debugging.cc: In function ‘void* __tsan_get_current_report()’:
tsan_debugging.cc:61:18: warning: cast from type ‘const __tsan::ReportDesc*’
to type ‘void*’ casts away qualifiers [-Wcast-qual]
return (void *)rep;
llvm-svn: 267679
This is reincarnation of http://reviews.llvm.org/D17648 with the bug fix pointed out by Adhemerval (zatrazz).
Currently ThreadState holds both logical state (required for race-detection algorithm, user-visible)
and physical state (various caches, most notably malloc cache). Move physical state in a new
Process entity. Besides just being the right thing from abstraction point of view, this solves several
problems:
Cache everything on P level in Go. Currently we cache on a mix of goroutine and OS thread levels.
This unnecessary increases memory consumption.
Properly handle free operations in Go. Frees are issue by GC which don't have goroutine context.
As the result we could not do anything more than just clearing shadow. For example, we leaked
sync objects and heap block descriptors.
This will allow to get rid of libc malloc in Go (now we have Processor context for internal allocator cache).
This in turn will allow to get rid of dependency on libc entirely.
Potentially we can make Processor per-CPU in C++ mode instead of per-thread, which will
reduce resource consumption.
The distinction between Thread and Processor is currently used only by Go, C++ creates Processor per OS thread,
which is equivalent to the current scheme.
llvm-svn: 267678
This reverts commit r267477.
It broke our bots that enables the AArch64 backends, it seems that
this code is using a Darwin *X86 specific* field.
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267526
This fixes fails in test/msan/dlerror.cc - when real dlerror calls strcmp,
our strcmp interceptor now skips poison checking, since it's called in
interceptor context. Strictly speaking, only the dlerror change is
necessary to fix the fail, but let's also change the other two just in case.
Differential Revision: http://reviews.llvm.org/D19499
llvm-svn: 267486
The interception context is not used by esan, but the compiler complains
about it being uninitialized all the same. We set it to null to avoid the
warning.
llvm-svn: 267376
Summary:
Adds libc interceptors to the runtime library for the new
EfficiencySanitizer ("esan") family of tools. The interceptors cover
the memory operations in most common library calls and will be shared
among all esan tools.
Reviewers: aizatsky
Subscribers: zhaoqin, tberghammer, danalbert, srhines, llvm-commits, vitalybuka, eugenis, kcc
Differential Revision: http://reviews.llvm.org/D19411
llvm-svn: 267293
This reverts commit r267094, because it broke a lot of MSAN tests in AArch64.
Being NFC and all, this needs some deeper investigation before it goes in again.
llvm-svn: 267136
Summary: When using 32-bit python with 64-bit asan the pc array in sancov.py cannot fit in 64-bit pc's because the type-code 'L' for
arrays in python corresponds to the C type long which is only of 4 bytes. Because of this some of the coverage tool tests fail on
mips. To fix these test possible solutions are to use 64-bit python or use struct.unpack with the 'Q' type-code. We have used
struct.unpack with 'Q' type code since it is not appropriate to have a 64-bit python on all hosts.
Reviewed by kcc, aizatsky
Differential: http://reviews.llvm.org/D18817
llvm-svn: 267126
Summary:
Adds the initial version of a runtime library for the new
EfficiencySanitizer ("esan") family of tools. The library includes:
+ Slowpath code via callouts from the compiler instrumentation for
each memory access.
+ Registration of atexit() to call finalization code.
+ Runtime option flags controlled by the environment variable
ESAN_OPTIONS. The common sanitizer flags are supported such as
verbosity and log_path.
+ An initial simple test.
Still TODO: common code for libc interceptors and shadow memory mapping,
and tool-specific code for shadow state updating.
Reviewers: eugenis, vitalybuka, aizatsky, filcab
Subscribers: filcab, vkalintiris, kubabrecka, llvm-commits, zhaoqin, kcc
Differential Revision: http://reviews.llvm.org/D19168
llvm-svn: 267060
The field "pid" in ReportThread is used to store the OS-provided thread ID (pthread_self or gettid). The name "pid" suggests it's a process ID, which it isn't. Let's rename it.
Differential Revision: http://reviews.llvm.org/D19365
llvm-svn: 266994
Let's use pthread_threadid_np which returns a more reasonable ID than pthread_self (which is actually a stack pointer). The numbers from pthread_threadid_np are already used in other tools, e.g. in LLDB, and often appear in logs, so it's much more useful than pthread_self.
Differential Revision: http://reviews.llvm.org/D18951
llvm-svn: 266991
Not sure what changed, but on my machine this is literally one byte
short. Only happens when malloc_context_size <= 2 due to the special
case in GET_STACK_TRACE definition (see asan_stack.h):
StackTrace::GetCurrentPc() on the right (context size > 2) branch
returns the address that is 200-something bytes from the return
address it is later matched to, while the same call on the left
branch is 321 bytes away from it.
This fixes the double-free test on my machine.
llvm-svn: 266932
Instead of calling a sanitizer_common function, implement GetPageSize in the
test directly. MSan runtime does not export __sanitizer::* symbols, and the
current code breaks when the test and the runtime library are in the separate
link units (ex. when the test is built as a shared library).
llvm-svn: 266910
Windows does not honour the __attribute__((pcs)) on ARM. Although this will
result in ABI mismatches, compiler-rt should largely be unneeded for resolving
dependencies as we generate MS ABI compliant library calls now for the most
part.
llvm-svn: 266891
The real problem is that sanitizer_print_stack_trace obtains current PC and
expects the PC to be in the stack trace after function calls. We don't
prevent tail calls in sanitizer runtimes, so this assumption does not
necessary hold.
We add "always inline" attribute on PrintCurrentStackSlow to address this
issue, however this solution is not reliable enough, but unfortunately, we
don't see any simple, reliable solution.
Reviewers: samsonov hfinkel kbarton tjablin dvyukov kcc
http://reviews.llvm.org/D19148
Thanks Hal, dvyukov, and kcc for invaluable discussion, I have even borrowed
part of dvyukov's summary as my commit message!
llvm-svn: 266869
This patch fixes https://github.com/google/sanitizers/issues/669. On older Darwin systems (in particular, Darwin 10),
dyld doesn't export '_dyldVersionNumber' symbol so we would have 'undefined reference' error in sanitzer library. Although
sanitizers support was added to LLVM on OS X 10.7+ where '_dyldVersionNumber' symbol is already exported, GCC users still
may want use them on older systems.
Differential Revision: http://reviews.llvm.org/D19218
llvm-svn: 266868
Summary: There is no frame validity check in the slow unwinder like there is in the fast unwinder due to which lsan reports a leak even for heap allocated coroutine in the test swapcontext.cc. Since mips/linux uses slow unwindwer instead of fast unwinder, the test fails for mips/linux. Therefore adding the checks before unwinding fixes the test for mips/linux.
Reviewers: samsonov, earthdok, kcc
Subscribers: llvm-commits, mohit.bhakkad, jaydeep
Differential: http://reviews.llvm.org/D18690
llvm-svn: 266716
This breaks the valloc test on PowerPC, which has 64kiB pages. Since
getting page size portably is nontrivial, and there's already a function
for that in __sanitizer, just use it. Unfortunately, sanitizer_common.h
conflicts with the interface headers inclucded by msan_test.cc (and a few
of its own macros), so we have to declare it manually.
Differential Revision: http://reviews.llvm.org/D19227
llvm-svn: 266688
On s390, siginfo reports the faulting address with page granularity -
we need to mask off the low bits of sp before comparison.
Differential Revision: http://reviews.llvm.org/D19112
llvm-svn: 266593
The test is failing on Windows because we do not have a definition for
DemangleSwiftAndCXX nor DemangleCXXABI, which I am replacing, on Windows.
llvm-svn: 266499
Add support for Swift names when symbolicating sanitizer traces. This is
now relevant since TSan and ASan support have been added to Swift on OS X.
Differential Revision: http://reviews.llvm.org/D19135
llvm-svn: 266494
sanitizer_common is now in good enough shape on s390x to support UBSan
- all tests passing. Let's enable it.
Differential Revision: http://reviews.llvm.org/D19157
llvm-svn: 266483
This file will contain s390-specific code. For now, let's move the s390
version of internal_mmap here.
Differential Revision: http://reviews.llvm.org/D19174
llvm-svn: 266482
Clang's StaticAnalyzer seems to (correctly) complain about code like:
T *p = calloc(sizeof(U), N);
...Where T and U are different types.
This patch removes some instances of this pattern from compiler-rt.
Patch by Apelete Seketeli.
Differential Revision: http://reviews.llvm.org/D19085
llvm-svn: 266388
On s390, the return address is in %r14, which is saved 14 words from
the frame pointer.
Unfortunately, there's no way to do a proper fast backtrace on SystemZ
with current LLVM - the saved %r15 in fixed-layout register save
area points to the containing frame itself, and not to the next one.
Likewise for %r11 - it's identical to %r15, unless alloca is used
(and even if it is, it's still useless). There's just no way to
determine frame size / next frame pointer. -mbackchain would fix that
(and make the current code just work), but that's not yet supported
in LLVM. We will thus need to XFAIL some asan tests
(Linux/stack-trace-dlclose.cc, deep_stack_uaf.cc).
Differential Revision: http://reviews.llvm.org/D18895
llvm-svn: 266371
This is the first part of upcoming asan support for s390 and s390x.
Note that there are bits for 31-bit support in this and subsequent
patches - while LLVM itself doesn't support it, gcc should be able
to make use of it just fine.
Differential Revision: http://reviews.llvm.org/D18888
llvm-svn: 266370
The PS_STRINGS constant can easily be incorrect with mismatched
kernel/userland - e.g. when building i386 sanitizers on FreeBSD/amd64
with -m32. The kern.ps_strings sysctl was introduced over 20 years ago
as the supported way to fetch the environment and argument string
addresses from the kernel, so the fallback is never used.
Differential Revision: http://reviews.llvm.org/D19027
llvm-svn: 266305
In short, CVE-2016-2143 will crash the machine if a process uses both >4TB
virtual addresses and fork(). ASan, TSan, and MSan will, by necessity, map
a sizable chunk of virtual address space, which is much larger than 4TB.
Even worse, sanitizers will always use fork() for llvm-symbolizer when a bug
is detected. Disable all three by aborting on process initialization if
the running kernel version is not known to contain a fix.
Unfortunately, there's no reliable way to detect the fix without crashing
the kernel. So, we rely on whitelisting - I've included a list of upstream
kernel versions that will work. In case someone uses a distribution kernel
or applied the fix themselves, an override switch is also included.
Differential Revision: http://reviews.llvm.org/D18915
llvm-svn: 266297
This teaches sanitizer_common about s390 and s390x virtual space size.
s390 is unusual in that it has 31-bit virtual space.
Differential Revision: http://reviews.llvm.org/D18896
llvm-svn: 266296
mmap on s390 is quite a special snowflake: since it has too many
parameters to pass them in registers, it passes a pointer to a struct
with all the parameters instead.
Differential Revision: http://reviews.llvm.org/D18889
llvm-svn: 266295
The custom zone implementation for OS X must not return 0 (even for 0-sized allocations). Returning 0 indicates that the pointer doesn't belong to the zone. This can break existing applications. The underlaying allocator allocates 1 byte for 0-sized allocations anyway, so returning 1 in this case is okay.
Differential Revision: http://reviews.llvm.org/D19100
llvm-svn: 266283
With -fsized-deallocation, new[] vs delete mismatch is reported as
new-delete-type-mismatch. This is technically true, but
alloc-dealloc-mismatch describes it better.
llvm-svn: 266246
We need to handle the case when handler is NULL in dispatch_source_set_cancel_handler and similar interceptors.
Differential Revision: http://reviews.llvm.org/D18968
llvm-svn: 266080
In `AtosSymbolizer`, we're using `forkpty()` to create a new pseudo-terminal to communicate with the `atos` tool (we need that to avoid output buffering in interactive mode). This however redirects both stdout and stderr into a single stream, so when we read the output, we can't distinguish between errors and standard replies. Let's save&restore stderr to avoid that.
Differential Revision: http://reviews.llvm.org/D15073
llvm-svn: 265923
Summary:
The strlen interceptor is sometimes invoked too early for REAL(strlen) to
be initialized. A special check is added to use internal_strlen for this
situation.
Reviewers: dim
Subscribers: llvm-commits, samsonov
Differential Revision: http://reviews.llvm.org/D18851
Change-Id: I3acc58f4abbae1904f25324abd84efea67aad0a2
llvm-svn: 265705
OS X provides atomic functions in libkern/OSAtomic.h. These provide atomic guarantees and they have alternatives which have barrier semantics. This patch adds proper TSan support for the functions from libkern/OSAtomic.h.
Differential Revision: http://reviews.llvm.org/D18500
llvm-svn: 265665
To avoid using the public header (tsan_interface_atomic.h), which has different data types, let's add all the __tsan_atomic* functions to tsan_interface.h.
Differential Revision: http://reviews.llvm.org/D18543
llvm-svn: 265663
Adding an interceptor with two more release+acquire pairs to avoid false positives with dispatch_apply.
Differential Revision: http://reviews.llvm.org/D18722
llvm-svn: 265662
XPC APIs have async callbacks, and we need some more happen-before edges to avoid false positives. This patch add them, plus a test case (sorry for the long boilerplate code, but XPC just needs all that).
Differential Revision: http://reviews.llvm.org/D18493
llvm-svn: 265661
GCD has APIs for event sources, we need some more release-acquire pairs to avoid false positives in TSan.
Differential Revision: http://reviews.llvm.org/D18515
llvm-svn: 265660
In the interceptor for dispatch_sync, we're currently missing synchronization between the callback and the code *after* the call to dispatch_sync. This patch fixes this by adding an extra release+acquire pair to dispatch_sync() and similar APIs. Added a testcase.
Differential Revision: http://reviews.llvm.org/D18502
llvm-svn: 265659
Summary:
After patch https://lkml.org/lkml/2015/12/21/340 is introduced in
linux kernel, the random gap between stack and heap is increased
from 128M to 36G on 39-bit aarch64. And it is almost impossible
to cover this big range. So we need to disable randomized virtual
space on aarch64 linux.
Reviewers: llvm-commits, zatrazz, dvyukov, rengolin
Subscribers: aemerson, rengolin, tberghammer, danalbert, srhines
Differential Revision: http://reviews.llvm.org/D18526
llvm-svn: 265366
We've reset thr->ignore_reads_and_writes, but forget to do
thr->fast_state.ClearIgnoreBit(). So ignores were not effective
reset and fast_state.ignore_bit was corrupted if signal handler
itself uses ignores.
Properly reset/restore fast_state.ignore_bit around signal handlers.
llvm-svn: 265288
This addresses PR27077. For some historical reason Darwin wasn't shipping multi3 in the compiler builtin library or in the OS builtin library. This caused building ffmpeg to fail because Polly was generating calls to multi3. It is easy enough to just add the builtin.
llvm-svn: 264750
to function names
Summary:
Hopefully this will make it easier for the next person to figure all
this out...
Reviewers: bogner, davidxl
Subscribers: davidxl, cfe-commits
Differential Revision: http://reviews.llvm.org/D18489
llvm-svn: 264680
This is implicitly needed at least by gcc-flag-compatibility.test
The thing that needs it is the `\` preceding the "default.profraw"
appended internally by clang when doing `-fprofile-use=`.
Clang uses `\` because is uses sys::path::append which will use `\` on a
Windows host. This is wrong, but I don't think there's an easy way to
solve it (maybe just always using `/` since places that accept `\` also
tend to accept `/`, but not the other way around).
llvm-svn: 264665
This change introduces routines that register and unregister all
instrumented globals in a loaded executable image.
These routines are only implemented on Darwin, where globals metadata
is expected to be placed in the __DATA,__asan_globals section.
Review: http://reviews.llvm.org/D16841
llvm-svn: 264644
This patch fixes the custom ThreadState destruction on OS X to avoid crashing when dispatch_main calls pthread_exit which quits the main thread.
Differential Revision: http://reviews.llvm.org/D18496
llvm-svn: 264627
Summary:
Hopefully this will make it easier for the next person to figure all
this out...
Reviewers: bogner, davidxl
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D18490
llvm-svn: 264612
Summary:
Currently, sanitizer_common_interceptors.inc has an implicit, undocumented
assumption that the sanitizer including it has previously declared
interceptors for memset and memmove. Since the memset, memmove, and memcpy
routines require interception by many sanitizers, we add them to the
set of common interceptions, both to address the undocumented assumption
and to speed future tool development. They are intercepted under a new
flag intercept_intrin.
The tsan interceptors are removed in favor of the new common versions. The
asan and msan interceptors for these are more complex (they incur extra
interception steps and their function bodies are exposed to the compiler)
so they opt out of the common versions and keep their own.
Reviewers: vitalybuka
Subscribers: zhaoqin, llvm-commits, kcc
Differential Revision: http://reviews.llvm.org/D18465
llvm-svn: 264451
ucrtbase.dll appears to be built with some kind of cross-module
inlining, because there are calls to imported Heap* routines sprinkled
throughout the code. This inlining defeats our attempts to hotpatch
malloc, _malloc_base, and related functions. Failing to intercept an
allocation or deallocation results in a crash when the program attempts
to deallocate or reallocate memory with the wrong allocator.
This change patches the IAT of ucrtbase.dll to replace the addresses of
the imported Heap* functions with implementations provided by ASan. We
don't globally intercept the win32 Heap* functions because they are
typically used by system DLLs that run before ASan initializes.
Eventually, we may want to intercept them, but for now I think this is
the minimal change that will keep ASan stable.
Reviewers: samsonov
Differential Revision: http://reviews.llvm.org/D18413
llvm-svn: 264327
On OS X, fork() under TSan asserts (in debug builds only) because REAL(fork) calls some intercepted functions, which check that no internal locks are held via CheckNoLocks(). But the wrapper of fork intentionally holds some locks. This patch fixes that by using ScopedIgnoreInterceptors during the call to REAL(fork). After that, all the fork-based tests seem to pass on OS X, so let's just remove all the UNSUPPORTED: darwin annotations we have.
Differential Revision: http://reviews.llvm.org/D18409
llvm-svn: 264261
On OS X, internal_mmap just uses mmap, which can invoke callbacks into libmalloc (e.g. when MallocStackLogging is enabled). This can subsequently call other intercepted functions, and this breaks our Darwin-specific ThreadState initialization. Let's use direct syscalls in internal_mmap and internal_munmap. Added a testcase.
Differential Revision: http://reviews.llvm.org/D18431
llvm-svn: 264259
Summary:
Adds strnlen to the common interceptors, under the existing flag
intercept_strlen.
Removes the now-duplicate strnlen interceptor from asan and msan.
This adds strnlen to tsan, which previously did not intercept it.
Adds a new test of strnlen to the sanitizer_common test cases.
Reviewers: samsonov
Subscribers: zhaoqin, llvm-commits, kcc
Differential Revision: http://reviews.llvm.org/D18397
llvm-svn: 264195
This is necessary to support the dynamic CRT (/MD) with VS2015. In
VS2015, these symbols are no longer imported from a DLL, they provided
statically by msvcrt.lib. This means our approach of hotpatching the DLL
no longer works.
By exporting the symbols, we end up relying on the same mechanism that
we use to intercept symbols in the static CRT (/MT) case. The ASan
runtime always needs to appear first on the link line, and the linker
searches for symbol definitions from left to right. This means we can
stop hotpatching operator new and delete in the CRT, which is nice.
I think that the only reason we weren't exporting the symbols already is
because MSVC doesn't allow you to do it directly with
__declspec(dllexport). Instead, we can use
`#pragma comment(linker, "/export:foo")`, which is most of what the
attribute does under the hood. It does mean we have to write down the
mangled names of the operators, but that's not too bad.
llvm-svn: 264190
This reverts commits r264068 and r264079, and they were breaking the build and
weren't reverted in time, nor they exhibited expected behaviour from the
reviewers. There is more to discuss than just a test fix.
llvm-svn: 264150
Summary:
After patch https://lkml.org/lkml/2015/12/21/340 is introduced in
linux kernel, the random gap between stack and heap is increased
from 128M to 36G on 39-bit aarch64. And it is almost impossible
to cover this big range. So I think we need to disable randomized
virtual space on aarch64 linux.
Reviewers: kcc, llvm-commits, eugenis, zatrazz, dvyukov, rengolin
Subscribers: rengolin, aemerson, tberghammer, danalbert, srhines, enh
Differential Revision: http://reviews.llvm.org/D18003
llvm-svn: 264068
Dmitry Vyukov