Re: [PATCH 4/6] x86/irq: merge common code in DEFINE_IDTENTRY_SYSVEC[_SIMPLE]

[Ingo Molnar <mingo@kernel.org>]

* H. Peter Anvin <hpa@zytor.com> wrote:

> 
> 
> On 5/12/21 1:38 AM, Ingo Molnar wrote:
> > 
> > * H. Peter Anvin <hpa@zytor.com> wrote:
> > 
> > > On 5/11/21 7:22 AM, Thomas Gleixner wrote:
> > > > 
> > > > That's not really making the code more readable. Something like the
> > > > below perhaps?
> > > > 
> > > > #define IDTENTRY_INVOKE_SYSVEC(func, regs, raw)				\
> > > > do {									\
> > > > 	irqentry_state_t state = irqentry_enter(regs);			\
> > > > 									\
> > > > 	instrumentation_begin();					\
> > > > 	kvm_set_cpu_l1tf_flush_l1d();					\
> > > >           if (raw) {							\
> > > > 		__irq_enter_raw();					\
> > > > 		func(regs);						\
> > > > 		__irq_exit_raw();			                \
> > > > 	} else {	                                                \
> > > > 		run_sysvec_on_irqstack_cond(func, regs);	        \
> > > >           }                                                               \
> > > > 	instrumentation_end();						\
> > > >           irqentry_exit(regs, state);					\
> > > > } while (0)								\
> > > > 
> > > 
> > > Digging more into it, it looks like a *lot* of the macros in
> > > <asm/irq_stack.h> and <asm/idtentry.h> can be replaced with inlines without
> > > any change in functionality or generated code.
> > 
> > That would be a much preferred outcome ...
> 
> Well, here is an RFC. This is obviously a much bigger change and I don't
> feel it is mature yet, but it would be good to know if you (plural) feel it
> is in the right direction.

Looks much cleaner IMO, and there's also some linecount reduction:

>  7 files changed, 117 insertions(+), 142 deletions(-)

Thanks,

	Ingo