https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119597
--- Comment #4 from Peter Bergner <bergner at gcc dot gnu.org> --- Jeevitha is looking into this, but the last thing I saw, it looked like an issue on powerpc where the callee mistakenly thinks the caller has allocated a param save area (which is located in the caller's stack frame) when it hasn't. When we spill a pseudo of a formal argument, instead of spilling into a stack slot in the callee's frame, LRA uses the argument's param save area slot to spill into. Since the caller didn't actually allocate the param save area, we scribble on the caller's frame info, and in this specific case, its return address which leads to the SEGV.
