Re: [PATCH] hppa: Wire up cacheflush syscall

["Andreas K. Huettel" <dilfridge@gentoo.org>]

[-- Attachment #1: Type: text/plain, Size: 1691 bytes --]

Am Montag, 8. Juli 2024, 10:58:35 CEST schrieb Helge Deller:
> On 7/8/24 10:13, Florian Weimer wrote:
> > * Helge Deller:
> >
> >> diff --git a/sysdeps/unix/sysv/linux/hppa/sys/cachectl.h b/sysdeps/unix/sysv/linux/hppa/sys/cachectl.h
> >> new file mode 100644
> >> index 0000000000..16e47d1329
> >> --- /dev/null
> >> +++ b/sysdeps/unix/sysv/linux/hppa/sys/cachectl.h
> >> @@ -0,0 +1,36 @@
> >
> >> +#ifndef _SYS_CACHECTL_H
> >> +#define _SYS_CACHECTL_H 1
> >> +
> >> +#include <features.h>
> >> +
> >> +/* Get the kernel definition for the op bits.  */
> >> +#include <asm/cachectl.h>
> >
> > This makes this header (<sys/cachectl.h>) unusable with older kernel
> > headers.  I think it also results in a test failure with older headers.
> > Is this a problem?
> 
> hppa lives in debian unstable, so basically you should always use
> the latest kernel & kernel headers when upgrading glibc.

Ahem.
https://www.gentoo.org/downloads/#hppa

> So, personally I think it's ok to ask people to upgrade.
> Dave, any opinion on this?
> 
> > In similar cases, we use LINUX_VERSION_CODE checks (for older compilers)
> > and __has_include to support backports.
> 
> I'd prefer to not add additional checks.
> Another (maybe easier) possibility is to simply backport the kernel asm/cachectl.h
> file (just the header, not the implementation, so the syscall will return -ENOSYS on
> older kernels):
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=c6d96328fecdda16e12f3b3c33f3677f4bcef89f
> 
> Helge
> 


-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]