On Tue, Jul 07, 2015 at 09:54:09AM -0500, Josh Poimboeuf wrote: > I did some more looking and it turns out that inline assembly doesn't > play nicely with frame pointers at all. If the inline asm is at the > beginning of the function, gcc sometimes emits the inline asm code > before setting up the frame pointer. That can break stack traces > when the inline asm has a call instruction. > > That turns out to be a very common problem. Stackvalidate found 37 C > object files which break frame pointer rules, thanks to inline asm. > > I don't know of a solution to this problem yet. Basically I think we > need a way to ensure that gcc emits the frame pointer setup before > inserting any inline asm (particularly when the inline asm has a call > instruction). A solution to this problem was posted by Segher Boessenkool in a related thread on the gcc mailing list: https://gcc.gnu.org/ml/gcc/2015-07/msg00080.html The suggestion is to use something like: register void *sp asm("%sp"); asm volatile("call func" : "+r"(sp)); I can confirm that it seems to fix the issue. (I had tried something like this before, but I guess I wasn't able to get the incantation just right.) Thanks to Jiri for the pointer to the thread, and Martin for raising the issue on the gcc list. -- Josh -- To unsubscribe from this list: send the line "unsubscribe live-patching" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
