FATAL ThreadSanitizer Exception on Startup
I'm trying to get ThreadSanitizer to work with the gcc release I have (4.8.2), so I took a simple example of them:
#include <pthread.h>
#include <stdio.h>
#include <string>
#include <map>
typedef std::map<std::string, std::string> map_t;
void *threadfunc(void *p) {
map_t& m = *(map_t*)p;
m["foo"] = "bar";
return 0;
}
int main() {
map_t m;
pthread_t t;
pthread_create(&t, 0, threadfunc, &m);
printf("foo=%s\n", m["foo"].c_str());
pthread_join(t, 0);
}
And compiled it without -fsanitize=thread
like below:
g ++ -o testtsan testtsan.cpp -lpthread
This is good, then I added a disinfectant to flow
g ++ -o testtsan testtsan.cpp -lpthread -fsanitize = thread
But, of course, this cannot be done without -pie -fPIC
g ++ -o testtsan testtsan.cpp -lpthread -fsanitize = thread -pie -fPIC
Which is then compiled, however when run I get:
FATAL: ThreadSanitizer CHECK failed: ../../../../libsanitizer/sanitizer_common/sanitizer_allocator.h:310 "((kSpaceBeg)) == ((reinterpret_cast<uptr>(Mprotect(kSpaceBeg, kSpaceSize))))" (0x7d0000000000, 0xffffffffffffffff)
FATAL: ThreadSanitizer: failed to intercept pthread_mutex_lock
Checking through strace
it looks like it is trying mmap
1TB of memory, so it fails ENOMEM
. I turned on ASLR and now I'm at a loss as to what it might be - so the question is, did anyone get it successfully?
Before I dived into the library code, I was hoping someone else had run into this already ...
Environment: GCC 4.8.2 Tried Kernel: 3.0.10 and 2.6.32 (all Suse), no luck ...
source to share
When I ran into this error on SLES11SP3 using clang 3.4.2
TSan, I was able to fix it by first modifying it ulimit
for my shell so that I could create a large mapping and then I had to execute it as superuser.
$ ./a.out
FATAL: ThreadSanitizer CHECK failed: bri/llvm-3.4.2.src/projects/compiler-rt/lib/sanitizer_common/sanitizer_allocator.h:316 "((kSpaceBeg)) == (( reinterpret_cast<uptr>(Mprotect(kSpaceBeg, kSpaceSize))))" (0x7d0000000000, 0xfffffffffffffff4)
FATAL: ThreadSanitizer: failed to intercept pthread_mutex_lock
$ ulimit -v
10588960
$ ulimit -v $((10588960*1024))
$ ulimit -v
10843095040
$ ./a.out
==11348==WARNING: Program is run with limited virtual address space, which wouldn't work with ThreadSanitizer.
==11348==Re-execing with unlimited virtual address space.
==11348==WARNING: Program is run with limited virtual address space, which wouldn't work with ThreadSanitizer.
==11348==Re-execing with unlimited virtual address space.
==11348==WARNING: Program is run with limited virtual address space, which wouldn't work with ThreadSanitizer.
==11348==Re-execing with unlimited virtual address space.
...
# only able to recover w/Ctrl-C...
$ sudo ./a.out
root password:
==11351==WARNING: Program is run with limited virtual address space, which wouldn't work with ThreadSanitizer.
==11351==Re-execing with unlimited virtual address space.
llvm-symbolizer: Unknown command line argument '--default-arch=x86_64'. Try: '/usr/bin/llvm-symbolizer -help'
llvm-symbolizer: Did you mean '-demangle=x86_64'?
==11351==WARNING: Can't read from symbolizer at fd 3
llvm-symbolizer: Unknown command line argument '--default-arch=x86_64'. Try: '/usr/bin/llvm-symbolizer -help'
llvm-symbolizer: Did you mean '-demangle=x86_64'?
==11351==WARNING: external symbolizer didn't start up correctly!
==11351==WARNING: Failed to use and restart external symbolizer!
==================
WARNING: ThreadSanitizer: data race (pid=11351)
Write of size 4 at 0x7fbca5148c48 by thread T1:
#0 Thread1 /home/bri/tmp/tsan/tiny_race.c:4 (exe+0x0000000ad64f)
#1 <null> <null>:0 (a.out+0x000000052af4)
Previous write of size 4 at 0x7fbca5148c48 by main thread:
#0 main /home/bri/tmp/tsan/tiny_race.c:11 (exe+0x0000000ad6a3)
Thread T1 (tid=11354, running) created by main thread at:
#0 pthread_create bri/tsan/rtl/tsan_interceptors.cc:877 (exe+0x000000052c2b)
#1 main /home/bri/tmp/tsan/tiny_race.c:10 (exe+0x0000000ad694)
SUMMARY: ThreadSanitizer: data race /home/bri/tmp/tsan/tiny_race.c:4 Thread1
==================
ThreadSanitizer: reported 1 warnings
The warnings symbolizer
are probably assembly-specific failures clang
, just ignore them. It is probably only used to denote C ++ symbol names.
source to share