On Sat, Jul 21, 2012 at 5:16 PM, Konstantin Belousov <kostikbel_at_gmail.com> wrote: > On Sat, Jul 21, 2012 at 04:00:45PM -0400, Kim Culhan wrote: >> On Fri, Jul 20, 2012 at 11:40 AM, Dimitry Andric <dim_at_freebsd.org> wrote: >> > On 2012-07-20 16:49, Kim Culhan wrote: >> >> Seeing this for r:238655 >> > ... >> >> In file included from /usr/src/sys/modules/dtrace/dtrace/../../../sys/pcpu.h:44: >> >> ./machine/pcpu.h:226:13: error: indirection of non-volatile null >> >> pointer will be deleted, not trap >> >> [-Werror,-Wnull-dereference] >> >> : "m" (*(char *)OFFSETOF_CURTHREAD)); >> >> ^~~~~~~~~~~~~~~~~~~~~~~~~~~ >> >> ./machine/pcpu.h:226:13: note: consider using __builtin_trap() or >> >> qualifying pointer with 'volatile' >> > >> > That's indeed a valid warning from clang, since OFFSETOF_CURTHREAD is >> > usually zero. It's probably due to recent work on dtrace. I'm not in >> > the neighborhood of a FreeBSD box right now to verify, but can you >> > please try to change the cast to "(volatile char *)"? That should fix >> > the warning. >> >> Yes it did, I know there are many considerations wrt to this warning. > > This should be equivalent to what you tried. Can you test build and > boot resulting kernel with this patch ? > > diff --git a/sys/amd64/include/pcpu.h b/sys/amd64/include/pcpu.h > index 5d1fd4d..7b3c934 100644 > --- a/sys/amd64/include/pcpu.h > +++ b/sys/amd64/include/pcpu.h > _at__at_ -217,16 +217,22 _at__at_ extern struct pcpu *pcpup; > #define PCPU_SET(member, val) __PCPU_SET(pc_ ## member, val) > > #define OFFSETOF_CURTHREAD 0 > +#ifdef __clang__ > +#define VOLATILE volatile > +#else > +#define VOLATILE > +#endif > static __inline __pure2 struct thread * > __curthread(void) > { > struct thread *td; > > __asm("movq %%gs:%1,%0" : "=r" (td) > - : "m" (*(char *)OFFSETOF_CURTHREAD)); > + : "m" (*(VOLATILE char *)OFFSETOF_CURTHREAD)); > return (td); > } > #define curthread (__curthread()) > +#undef VOLATILE > > #define OFFSETOF_CURPCB 32 > static __inline __pure2 struct pcb * The resulting kernel appears to run the same as the kernel tried earlier. Hope this helps. thanks -kimReceived on Sat Jul 21 2012 - 22:37:22 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:29 UTC