Re: [PATCH] x86: clean up after: move entry_64.S register saving out of the macros

[Alexander van Heukelum]

On Thu, Nov 20, 2008 at 04:39:54PM +0100, Ingo Molnar wrote:
> 
> * Alexander van Heukelum <heukelum@xxxxxxxxxxxxx> wrote:
> 
> > On Thu, Nov 20, 2008 at 04:04:12PM +0100, Ingo Molnar wrote:
> > > 
> > > * Alexander van Heukelum <heukelum@xxxxxxxxxxxxx> wrote:
> > > 
> > > > This add-on patch to x86: move entry_64.S register saving out of the 
> > > > macros visually cleans up the appearance of the code by introducing 
> > > > some basic helper macro's. It also adds some cfi annotations which 
> > > > were missing.
> > > > 
> > > > Signed-off-by: Alexander van Heukelum <heukelum@xxxxxxxxxxx>
> > > > ---
> > > >  arch/x86/kernel/entry_64.S |  220 ++++++++++++++++++++++----------------------
> > > >  1 files changed, 112 insertions(+), 108 deletions(-)
> > > > 
> > > > Hello Ingo,
> > > > 
> > > > This patch improves the CFI-situation in entry_64.S, but restricted 
> > > > mostly to the areas touched by "x86: move entry_64.S register saving 
> > > > out of the macros". I'm sure there will be some small errors 
> > > > somewhere, but it compiles and runs fine.
> > > 
> > > very nice cleanup! This is exactly what should be done. Applied to 
> > > tip/x86/irq.
> > > 
> > > Note, i did a small rename:
> > > 
> > >     cfi_pushq  =>  pushq_cfi
> > >     cfi_popq   =>  popq_cfi
> > >     cfi_store  =>  movq_cfi

Does not work... But if you are attached to the underscores, I
think we can force it to work by using CPP to convert it to
something the assembler does parse right:

#define pushq_cfi pushq.cfi

etc?

Or is that just too ugly?

Alexander

> > > as the goal is to have the actual source code read mostly as regular 
> > > assembly code. The fact that the macro is equivalent to a 
> > > default-annotated pushq/popq/movq instruction is much more important 
> > > than the fact that it also does CFI annotations.
> > > 
> > > Also, while cfi_store is correct as well, the usual x86 assembly term 
> > > (and instruction used here) is movq.
> > 
> > Now I have a little problem with my next patch... I wanted to 
> > introduce cfi_load. Guess what assembly instruction that maps to ;).
> 
> heh ;-)
> 
> the restore direction could be named movq_cfi_restore, and have the same 
> order of arguments as the regular movq that it replaces. I.e.:
> 
>         movq 8(%rsp),%r11
>         CFI_RESTORE r11
> 
> would map to:
> 
> 	movq_cfi_restore 8, r11
> 
> or so.
> 
> cfi_store has really a bad name: it's confusing whether it's the CFI 
> info we are storing/registering (which we are), or a 'store' instruction 
> (which this is too).
> 
> If then we should name it movq_cfi_store or movq_cfi_register - but 
> that's too long.
> 
> movq_cfi for the frame construction direction and movq_cfi_restore for 
> the frame deconstruction phase sounds like a good naming compromise, hm?
> 
> 	Ingo

-- 
Alexander van Heukelum
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/