On Mär 29 2021, Andreas Schwab wrote: > On Mär 29 2021, Christophe Leroy wrote: > >> Le 29/03/2021 à 10:33, Benjamin Herrenschmidt a écrit : >>> On Fri, 2021-03-12 at 11:20 +1000, Nicholas Piggin wrote: >>>> >>>> +static inline void nap_adjust_return(struct pt_regs *regs) >>>> >>>> +{ >>>> >>>> +#ifdef CONFIG_PPC_970_NAP >>>> >>>> + if (unlikely(test_thread_local_flags(_TLF_NAPPING))) { >>>> + /* Can avoid a test-and-clear because NMIs do not call >>>> this */ >>>> + clear_thread_local_flags(_TLF_NAPPING); >>>> + regs->nip = (unsigned long)power4_idle_nap_return; >>>> + } >>> Is this a pointer to a function descriptor or the actual code ? >>> >> >> --- a/arch/powerpc/kernel/idle_book3s.S >> +++ b/arch/powerpc/kernel/idle_book3s.S >> @@ -209,4 +209,8 @@ _GLOBAL(power4_idle_nap) >> mtmsrd r7 >> isync >> b 1b >> + >> + .globl power4_idle_nap_return >> +power4_idle_nap_return: >> + blr >> #endif > > The problem is not the definition, it is the reference. In C, a > function symbol always resolves to the address of the descriptor.
Sorry, this is wrong, I have misremembered how function descriptors work on ppc64. The address is really pointing to the actual code. Andreas. -- Andreas Schwab, sch...@linux-m68k.org GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510 2552 DF73 E780 A9DA AEC1 "And now for something completely different."