Re: git: c29cba408d19 - main - stand: Narrow the 'signal emulation' we provide

From: Jessica Clarke <jrtc27_at_freebsd.org>
Date: Wed, 20 Nov 2024 03:55:57 UTC
On 20 Nov 2024, at 03:35, Warner Losh <imp@FreeBSD.org> wrote:
> 
> The branch main has been updated by imp:
> 
> URL: https://cgit.FreeBSD.org/src/commit/?id=c29cba408d197ed2c267d0605e4225bb54153e01
> 
> commit c29cba408d197ed2c267d0605e4225bb54153e01
> Author:     Warner Losh <imp@FreeBSD.org>
> AuthorDate: 2024-11-20 03:05:11 +0000
> Commit:     Warner Losh <imp@FreeBSD.org>
> CommitDate: 2024-11-20 03:24:41 +0000
> 
>    stand: Narrow the 'signal emulation' we provide
> 
>    We only need to provide sig_atomic_t in emulation. However, including
>    machine/signal.h brings in too much namespace pollution related to
>    signals. Instead, define sig_atomic_t as long. Setting long is async
>    atomic on all platforms (though powerpc64 defines it to an int), though
>    that doesn't matter since the boot loader doesn't use signals.
> 
>    Sponsored by:           Netflix
> ---
> stand/libsa/stand.h | 9 +++++++--
> 1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/stand/libsa/stand.h b/stand/libsa/stand.h
> index 260defa3a33d..e1188fb73a26 100644
> --- a/stand/libsa/stand.h
> +++ b/stand/libsa/stand.h
> @@ -82,8 +82,13 @@
> #define EOFFSET (ELAST+8) /* relative seek not supported */
> #define ESALAST (ELAST+8) /* */
> 
> -/* Partial signal emulation for sig_atomic_t */
> -#include <machine/signal.h>
> +/*
> + * LUA needs sig_atomic_t. This is defined to be long or int on all our
> + * platforms. On all but powerpc, these are all the same thing as long. 64-bit
> + * powerpc defines this as int, but long can also be accessed atomically. It's
> + * also OK because we don't have signal handlers in the boot loader.
> + */

Presumably an oversight when adding the powerpc64* ports :( Probably we
should fix this by taking advantage of being able to break ABI for tier
2 architectures between major versions. Shame it wasn’t done the same
time as ELFv1 -> ELFv2. We fixed riscv and arm64’s SIG_ATOMIC_MIN/MAX
recently, but the underlying types were still long in that case,
thankfully.

Jess

> +typedef long sig_atomic_t;
> 
> __BEGIN_DECLS
>