Re: [PATCH 05/24] x86/entry: Introduce __entry_text for entry code written in C

[Lai Jiangshan]

On 2021/9/1 03:34, Peter Zijlstra wrote:
> On Wed, Sep 01, 2021 at 01:50:06AM +0800, Lai Jiangshan wrote:
> > From: Lai Jiangshan <laijs@xxxxxxxxxxxxxxxxx>
> >
> > Some entry code will be implemented in traps.c.  We need __entry_text
> > to set them in .entry.text section.
> >
> > Signed-off-by: Lai Jiangshan <laijs@xxxxxxxxxxxxxxxxx>
> > ---
> >   arch/x86/entry/traps.c | 5 +++++
> >   1 file changed, 5 insertions(+)
> >
> > diff --git a/arch/x86/entry/traps.c b/arch/x86/entry/traps.c
> > index f71db7934f90..ef07ae5fb3a0 100644
> > --- a/arch/x86/entry/traps.c
> > +++ b/arch/x86/entry/traps.c
> > @@ -69,6 +69,11 @@
> >   extern unsigned char native_irq_return_iret[];
> >   extern unsigned char asm_load_gs_index_gs_change[];
> >   
> > +/* Entry code written in C. Must be only used in traps.c */
> > +#define __entry_text							\
> > +	noinline notrace __attribute((__section__(".entry.text")))	\
> > +	__no_kcsan __no_sanitize_address __no_profile __no_sanitize_coverage
>
> Urgh, that's bound to get out of sync with noinstr.. How about you make
> noinstr something like:
>
> #define __noinstr_section(section) \
> 	noinline notrace __attribute((__section__(section)))	\
> 	__no_kcsan __no_sanitize_address __no_profile __no_sanitize_coverage
>
> #define noinstr __noinstr_section(".noinstr.text")
>
> and then write the above like:
>
> #define __entry_text __noinstr_section(".entry.text")

It is a good idea.  There was no "__no_profile __no_sanitize_coverage"
when I first made the patch several month ago.  I added it after I recheck
the definition of "noinstr" yesterday which means we really need __noinstr_section()
to keep them sync.

>