[llvm-dev] Query regarding usage of optimizations on Sanitizers in Clang

[Reid Kleckner via llvm-dev]

This looks like ASan FAQ item #2:
https://github.com/google/sanitizers/wiki/AddressSanitizer#faq

Q: Why didn't ASan report an obviously invalid memory access in my code?
A1: If your errors is too obvious, compiler might have already optimized it
out by the time Asan runs.

I would have liked to find clear documentation on the tradeoffs of using
optimizations in the main documentation here, but I only see a
recommendation to use -O1:
https://clang.llvm.org/docs/AddressSanitizer.html

To answer your question, the behavior is expected. Also, Clang's default
optimization level is -O0.

I think the consensus among most ASan users is that optimizing before
instrumentation is the right tradeoff between accuracy and performance. You
are welcome to test with -O0 to find these kinds of trivial issues, but
many users find that running tests under -O0+ASan is too slow to be
feasible.

Another way of looking at it is that the compiler has optimized away the
OOB store here, and if you ship an optimized build, ASan is correctly
telling you that there is no OOB write in the shipping build. Godbolt shows
your example optimizes main to just return 0, so the program doesn't do an
OOB write at runtime:
https://gcc.godbolt.org/z/cTsdb4Thz

Hope that helps.

On Wed, Dec 1, 2021 at 7:10 AM Bharadwaj, Ritanya B via llvm-dev <
llvm-dev at lists.llvm.org> wrote:

> Hi,
>
>
>
> I have been using Sanitizers/Archer for testing purposes on C,C++ and
> Fortran test suites. I notice some change in sanitizer behaviour when I add
> any kind of optimization to it ( O1, O2, Ofast etc). For example, in the
> following test case:
>
> #include <alloca.h>
>
> void foo(int index, int len) {
>
>                              volatile char *str = (volatile
> char*)alloca(len);
>
>                              str[index] = '1';  // Boom!
>
> }
>
> int main(int argc, char **argv) {
>
>                             foo(-1, 10);
>
> }
>
> Compiling it with : clang -g -fsanitize=address *Test_case.c*
>
>
>
> Gives me ASan report :
>
> ==80838==ERROR: AddressSanitizer: dynamic-stack-buffer-overflow on address
> 0x7ffdd959ac3f at pc 0x00000050ea21 bp 0x7ffdd959ac10 sp 0x7ffdd959ac08
>
> WRITE of size 1 at 0x7ffdd959ac3f thread T0
>
> ………..
>
> Address 0x7ffdd959ac3f is located in stack of thread T0
>
> SUMMARY: AddressSanitizer: dynamic-stack-buffer-overflow
> (/ptmp1/bhararit/new_petest/petest/
> cc_address_sanitizer.ar/diffs/a.out+0x50ea20)
>              ……….
>
>
>
> But when I compile it with optimization: clang -g -fsanitize=address -O1
> *Test_case.c*
>
> I do not see any Sanitizer report.
>
> Is this an expected behaviour? If yes, should we be disabling
> optimizations before using sanitizers although the clang document states
> that O2 or higher level of optimizations can be used to enhance the
> performance. Lastly, what is the default optimization in clang ( is it
> -O0?)?
>
> It would be very helpful if you could give me more insights on this. Hope
> to hear from you. Thanks in advance.
>
> Best Regards,
>
> Ritanya.
>
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20211202/92e69584/attachment.html>