On 24/03/2021 23.34, Sami Tolvanen wrote: > On Wed, Mar 24, 2021 at 2:51 PM Rasmus Villemoes > <li...@rasmusvillemoes.dk> wrote: >> >> On 24/03/2021 18.33, Peter Zijlstra wrote: >>> On Wed, Mar 24, 2021 at 05:45:52PM +0100, Rasmus Villemoes wrote: >>>> Sorry, I think I misread the code. The static calls are indeed >>>> initialized with a function with the right prototype. Try adding >>>> "preempt=full" on the command line so that we exercise these lines >>>> >>>> static_call_update(cond_resched, >>>> (typeof(&__cond_resched)) __static_call_return0); >>>> static_call_update(might_resched, >>>> (typeof(&__cond_resched)) __static_call_return0); >>>> >>>> I would expect that to blow up, since we end up calling a long (*)(void) >>>> function using a function pointer of type int (*)(void). >>> >>> Note that on x86 there won't actually be any calling of function >>> pointers. See what arch/x86/kernel/static_call.c does :-) >> >> I know, but so far x86 is the only one with HAVE_STATIC_CALL, so for >> arm64 which is where CFI seems to be targeted initially, static_calls >> are function pointers. And unless CFI ignores the return type, I'd >> really expect the above to fail. > > I think you're correct, this would trip CFI without HAVE_STATIC_CALL. > However, arm64 also doesn't support PREEMPT_DYNAMIC at the moment, so > this isn't currently a problem there.
Well, there's PREEMPT_DYNAMIC and HAVE_PREEMPT_DYNAMIC. The former doesn't depend on the latter (and the latter does depend on HAVE_STATIC_CALL, so effectively not for anything but x86). You should be able to select both PREEMPT_DYNAMIC and CFI_CLANG, and test if booting with preempt=full does give the fireworks one expects. Rasmus