Re: [musl] SA_RESTORER for rv64?

[Rich Felker <dalias@libc.org>]

On Fri, Feb 03, 2023 at 10:44:56AM -0800, enh wrote:
> oops, never actually sent the patch. attached...
> 
> On Thu, Nov 10, 2022 at 9:31 AM Khem Raj <raj.khem@gmail.com> wrote:
> >
> > On Thu, Nov 10, 2022 at 9:19 AM Rich Felker <dalias@libc.org> wrote:
> > >
> > > On Thu, Nov 10, 2022 at 07:44:23AM -0800, enh wrote:
> > > > arch/riscv64/bits/signal.h has contained a definition for SA_RESTORER since
> > > > the initial commit, but i think that's just copy & paste from whichever
> > > > architecture the rv64 headers were based on? the linux kernel itself
> > > > doesn't have SA_RESTORER for rv64, unless i'm missing something?
> > >
> > > I suspect this is just a mistake. Have you seen any ill effects from
> > > it? If riscv folks can confirm it's wrong, I'll remove it.
> >
> > Yeah I think it should be removed. Perhaps mips is in same boat.
> >
> > >
> > > Rich

> From 6413de6d9f785c98e5bc0cf40be947f1169d2fd7 Mon Sep 17 00:00:00 2001
> From: Elliott Hughes <enh@google.com>
> Date: Fri, 3 Feb 2023 10:42:55 -0800
> Subject: [PATCH] risc-v does not have SA_RESTORER.
> 
> The kernel's include/uapi/asm-generic/signal-defs.h explicitly calls
> this out as obsolete. New architectures like risc-v do not define it.
> ---
>  arch/riscv64/bits/signal.h | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/arch/riscv64/bits/signal.h b/arch/riscv64/bits/signal.h
> index 287367db..fd6157a3 100644
> --- a/arch/riscv64/bits/signal.h
> +++ b/arch/riscv64/bits/signal.h
> @@ -76,7 +76,6 @@ typedef struct __ucontext
>  #define SA_RESTART   0x10000000
>  #define SA_NODEFER   0x40000000
>  #define SA_RESETHAND 0x80000000
> -#define SA_RESTORER  0x04000000
>  
>  #endif
>  
> -- 
> 2.39.1.519.gcb327c4b5f-goog
> 

I don't think this patch works as-is, since musl unconditionally uses
SA_RESTORER. We probably need to make that conditional on its
presence, and it looks like there's also a wrong-struct-layout issue
on archs where it's absent...

Rich