Re: WARNING: can't access registers at asm_common_interrupt

[Shinichiro Kawasaki]

On Nov 18, 2020 / 09:22, Peter Zijlstra wrote:
> On Wed, Nov 18, 2020 at 07:47:36AM +0100, Jürgen Groß wrote:
> > On 16.11.20 14:04, Peter Zijlstra wrote:
> > > On Mon, Nov 16, 2020 at 12:56:32PM +0100, Jürgen Groß wrote:
> > > > > > > > > > static inline notrace unsigned long arch_local_save_flags(void)
> > > > > > > > > > {
> > > > > > > > > >      PVOP_CALL_ARGS;
> > > > > > > > > >      PVOP_TEST_NULL(irq.save_fl);
> > > > > > > > > >      asm_inline volatile(ALTERNATIVE(paravirt_alt(PARAVIRT_CALL),
> > > > > > > > > >                                      "PUSHF; POP _ASM_AX",
> > > > > > > > > >                                      X86_FEATURE_NATIVE)
> > > > 
> > > > I am wondering whether we really want a new feature (basically "not
> > > > XENPV). We could use ~X86_FEATURE_XENPV and teach apply_alternatives()
> > > > to understand negated features (yes, this limits the number of features
> > > > to 32767, but I don't think this is a real problem for quite some time).
> > > > 
> > > > Thoughts?
> > > 
> > > I went with the simple thing for now... If we ever want/need another
> > > negative alternative I suppose we can do as you suggest...
> > > 
> > > I was still poking at objtool to actually dtrt though..
> > 
> > I'd like to include this part in my series (with a different solution
> > for the restore_fl part, as suggested by Andy before).
> > 
> > Peter, are you fine with me taking your patch and adding your SoB?
> 
> Sure, note that I only compile tested it, as it was my test-bed for
> playing with objtool (which I still haven't managed to get back to and
> finish :/).
> 
> So if it actually _works_ feel free to take it, otherwise you can have
> whatever bits and pieces remain after you've butchered it to do as you
> want.

All, thank you for the actions.

I tried Peter's patch in my test system and result looks good. The warning is
still observed with 5.10-rc4. With the patch on top of 5.10-rc4, the warning
disappeared.

Of note is that when I built kernel with the patch, objtool reported many
warnings, as follows:

arch/x86/events/intel/core.o: warning: objtool: .altinstr_replacement+0x0: alternative modifies stack
arch/x86/events/core.o: warning: objtool: .altinstr_replacement+0x0: alternative modifies stack

-- 
Best Regards,
Shin'ichiro Kawasaki