Re: [PATCH v3 0/6] locking/rwsem/x86: Add stack frame dependency for some inline asm

From: Miguel Bernal Marin
Date: Mon Sep 25 2017 - 17:34:28 EST


On Mon, Sep 25, 2017 at 02:03:21PM -0500, Josh Poimboeuf wrote:
> On Mon, Sep 25, 2017 at 02:00:43PM -0500, Josh Poimboeuf wrote:
> > On Mon, Sep 25, 2017 at 12:34:19PM -0500, Miguel Bernal Marin wrote:
> > > On Wed, Sep 20, 2017 at 04:24:18PM -0500, Josh Poimboeuf wrote:
> > > > On Tue, Sep 19, 2017 at 06:37:39PM -0500, Miguel Bernal Marin wrote:
> > > > > Some warning were showed by objtool using gcc 7.2.0
> > > > >
> > > > > kernel/locking/rwsem.o: warning: objtool: up_read()+0x11: call without frame pointer save/setup
> > > > > kernel/locking/rwsem.o: warning: objtool: up_write()+0x17: call without frame pointer save/setup
> > > > > kernel/locking/rwsem.o: warning: objtool: downgrade_write()+0x22: call without frame pointer save/setup
> > > > >
> > > > > which means gcc placed an inline asm function and its call instruction before
> > > > > the frame pointer setup.
> > > > >
> > > > > This series forces a stack frame to be created before the call instruction
> > > > > by listing the stack pointer as an output operand in the inline asm statement.
> > > > >
> > > > > Also to be easy to maintain and understand the operands from the extended
> > > > > assembler instructions were converted to named operands.
> > > >
> > > > I've got a patch going around which will change the way we do this, so
> > > > you'll probably need to do a v3 after my patch gets merged. I'll add
> > > > you to cc for the next revision.
> > > >
> > >
> > > With your new patches (at v4.14.-rc2) the warning is not seen any more,
> > > so I will send only the named operand patches (in separate thread), as
> > > this fix is not more needed.
> >
> > Any chance you tested with GCC 7? With GCC 6 and older you might still
> > see the warnings.
>
> Sorry, reading again it looks like your warnings only started showing up
> in GCC 7.2.0? If so, then it does make sense that your fix isn't needed
> any more, because my patch fixed this issue for *all* inline asm for GCC
> 7+.
>

Yes it was with GCC 7.2.0


> --
> Josh

--
Regards,

Miguel Bernal Marin Open Source Technology Center
https://clearlinux.org Intel Corporation