https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83302
--- Comment #3 from Michael Matz <matz at gcc dot gnu.org> --- In any case, this is not something that GCC could do anything about. A probe necessarily has to be a write, and writing to something not belonging to own stack (or guard page) will always have this problem of seemingly destroying other data. The non-atomicness of this particular read-mod-write instruction doesn't enter the picture.