Express the strto* interceptors though macros. This removes a lot of
duplicate code and fixes a couple of copypasto bugs (where "res" was declared of
a different type than the actual return type). Also, add a few more interceptors
for strto*_l.
llvm-svn: 200316
This change is a part of refactoring intended to have common signal handling behavior in all tools.
Note that this particular change doesn't enable use_sigaltstack support in every tool.
llvm-svn: 200310
allow SIGABRT to spoil errno, because some real programs
reset SIGABRT handler in the handler, re-raise SIGABRT and return from the handler
llvm-svn: 200304
It was redundant (since ARM mode is the default) and misleading since
(e.g.) Cortex-A15 would not satisfy the #ifdef but would be in ARM
mode regardless.
llvm-svn: 200043
remains ARM mode only, supporting thumb requires explicit it prefixes
for the predicted adds/subs and adjusting the offset computation for the
different block sizes.
llvm-svn: 200035
We left ignore_interceptors>0 when calling signal handlers
from blocking interceptors, this leads to missing synchronization in such signal handler.
llvm-svn: 200003
test-and-subtract loop and compute the initial block as address,
shaving off between 5% and 10% on Cortex A9 and 30%+ a Raspberry Pi.
Code written by Matt Thomas and Joerg Sonnenberger.
Differential Revision: http://llvm-reviews.chandlerc.com/D2595
llvm-svn: 200001
Currently correct programs can deadlock after fork, because atomic operations and async-signal-safe calls are not async-signal-safe under tsan.
With this change:
- if a single-threaded program forks, the child continues running with verification enabled (the tsan background thread is recreated as well)
- if a multi-threaded program forks, then the child runs with verification disabled (memory accesses, atomic operations and interceptors are disabled); it's expected that it will exec soon anyway
- if the child tries to create more threads after multi-threaded fork, the program aborts with error message
- die_after_fork flag is added that allows to continue running, but all bets are off
http://llvm-reviews.chandlerc.com/D2614
llvm-svn: 199993
warning: ISO C99 requires rest arguments to be used [enabled by default]
INTERCEPTOR(char *, dlerror) {
warning: invoking macro INTERCEPTOR argument 3: empty macro arguments are undefined in ISO C90 and ISO C++98 [enabled by default]
llvm-svn: 199873
New/delete implementations in system libraries almost always are built without
frame pointers. As we switched to frame pointer based unwinder on ARM, they no
longer work for us, resulting in broken allocation/deallocation stacks.
Note that this does not work with statically linked
libstdc++/libc++/libstlport.
llvm-svn: 199872
Clang now uses the *-apple-macho triples for embedded targets using MachO. The
environment makes less pretences to be running an actual Mach kernel with all
the support that would imply, so a few changes are needed to the build scripts:
+ The Apple way of detecting endianness applies in all cases so drop the
__MACH__ test.
+ A quick rename since "darwin_embedded" no longer makes any sense. If there's
no Mach going on, there's certainly no Darwin.
+ Change how we test for support for a platform. The functions we build don't
need any SDK support (stdio.h, ...) so we don't need to check for it.
llvm-svn: 199731
Intercept and sanitize arguments passed to printf functions in ASan and TSan
(don't do this in MSan for now). The checks are controlled by runtime flag
(off by default for now).
Patch http://llvm-reviews.chandlerc.com/D2480 by Yuri Gribov!
llvm-svn: 199729
This code works with AAPCS-style {fp, lr} frames. Unlike linux perf, it does
not unwind through code compiled with -mapcs-frame (which Clang does not
support anyway).
llvm-svn: 199725
In optimized hybrid execution we do not use DynamoRIO private loader, which
mangles TLS access, so we can access the application's TLS directly.
Patch by Qin Zhao.
llvm-svn: 199655
This change allows to compile sanitizer sources so that *san runtime
will attempt to use libbacktrace and/or libiberty for symbolization
(instead of communicating with llvm-symbolizer).
I've tested this patch by manually defining SANITIZER_LIBBACKTRACE and/or
SANITIZER_CP_DEMANGLE, linking with necessary libraries and verifying that
all tests from ASan test suite work.
Based on patches by Jakub Jelinek!
llvm-svn: 199384
This change adds ASAN_OPTIONS=start_deactivated=1 flag. When present, ASan will
start in "deactivated" mode, with no heap poisoning, no quarantine, no stack
trace gathering, and minimal redzones. All this features come back when
__asan_init is called for the constructor of an instrumented library.
The primary use case for this feature is Android. Code itself is not
Android-specific, and this patch includes a Linux test for it.
llvm-svn: 199377
struct stat defined in <asm/stat.h> is incorrect for x32. <asm/stat.h>
is included to get struct __old_kernel_stat. But struct __old_kernel_stat
isn't used for x86-64 nor x32. This patch includes <sys/stat.h> instead
of <asm/stat.h> and comments out size check of struct __old_kernel_stat
for x86-64.
Patch by H.J. Lu
llvm-svn: 199192
This unbreaks the configure build on darwin, where we build with the
stub SDK in -isysroot. Oddly, the cmake build doesn't have this
restriction, which probably means the cmake configs are out of date.
llvm-svn: 198676
Otherwise on (some) 64-bit systems, -Wformat will trigger a warning
because uint64_t is an 'unsigned long' not an 'unsigned long long'.
Consequently, PGOProfiling.c would fail to build if -Werror and
-Wformat are both enabled.
llvm-svn: 198644
This is fairly minimal support for instrumentation based PGO. The data
format is inefficient, and the output file name is hardcoded to
pgo-data.
llvm-svn: 198638
Currently fails with:
gotsan.cc:7686:40: error: the address of 'int __sanitizer_pthread_attr_getstack(void*, void**, size_t*)' will always evaluate as 'true' [-Werror=address]
llvm-svn: 198261
A helper function is a C++ function, and so even though one of the two
definitions is weak, it still technically triggers the ODR. Perhaps
these two definitions are ODR equivalent, but I'm not even confident in
that.
Instead, just define the function once, declare it as weak, and use
a wrapper that is clearly file-local. This avoids two definitions. Also
make the function extern "C" so that we can't even mess up the type
signature somehow or otherwise fail to match up the weak declaration
here with the interceptor defined elsewhere.
llvm-svn: 198253
Per Nick Kledzik (http://llvm.org/bugs/show_bug.cgi?id=17976):
"""
For MacOSX, the labels in __cstring section should be 'L' labels, and the labels
in the __const section should be non-L or 'l' labels. The later ('l') will cause
the linker to remove the label in the final executable.
"""
, so we need to ensure that the globals emitted by ASan have appropriate labels.
llvm-svn: 198022
This should fix http://llvm.org/bugs/show_bug.cgi?id=17976
Another test checking for the global variables' locations and prefixes on Darwin will be committed separately.
llvm-svn: 198017
This is intended to address the following problem.
Episodically we see CHECK-failures when recursive interceptors call back into user code. Effectively we are not "in_rtl" at this point, but it's very complicated and fragile to properly maintain in_rtl property. Instead get rid of it. It was used mostly for sanity CHECKs, which basically never uncover real problems.
Instead introduce ignore_interceptors flag, which is used in very few narrow places to disable recursive interceptors (e.g. during runtime initialization).
llvm-svn: 197979
Print the list of leaked objects after each leak report. Previously we
printed only a joint list of all leaked objects. As a bonus, suppressed objects
are no longer reported.
llvm-svn: 197977
Added the following custom wrappers:
strstr
strrchr
memchr
nanosleep
socketpair
Tweaked a couple of existing wrappers: if a wrapper returns a pointers derived
from an input pointer, then return the label of the input pointer
Sorted tests invocation
Differential Revision: D2354
llvm-svn: 197601
This code is not robust enough and triggers when simply linking with
libdynamorio.so, without any code translation at all. Disabling it is safe
(i.e. we may unpoison too much memory and see false negatives, but never false
positives).
llvm-svn: 197568
It conflicted with the verbosity flag we had in common flags. We don't need an
LSan-specific flag anyway.
Also, shift some logging levels and remove some unnecessary code.
llvm-svn: 197512
Introduce a flag to either always or never print matched suppressions.
Previously, matched suppressions were printed unconditionally if there were
unsuppressed leaks. Also, verbosity=1 no longer has the semantics of "always
print suppressions and summary".
llvm-svn: 197510
Add an interface for telling LSan that a region of memory is to be treated as a
source of live pointers. Useful for code which stores pointers in mapped memory.
llvm-svn: 197489
Joerg Sonnenberger