[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAHk-=wgCttiyp+3BBzhqKv+uXuUr-fzw2QbmH8kXwO+sB+FAaQ@mail.gmail.com>
Date: Wed, 3 Jul 2024 12:00:33 -0700
From: Linus Torvalds <torvalds@...ux-foundation.org>
To: Christian Brauner <brauner@...nel.org>
Cc: Xi Ruoyao <xry111@...111.site>, libc-alpha@...rceware.org,
"Andreas K. Huettel" <dilfridge@...too.org>, Arnd Bergmann <arnd@...db.de>,
Huacai Chen <chenhuacai@...nel.org>, Mateusz Guzik <mjguzik@...il.com>,
Alexander Viro <viro@...iv.linux.org.uk>, Jan Kara <jack@...e.cz>, linux-kernel@...r.kernel.org,
linux-fsdevel@...r.kernel.org, io-uring@...r.kernel.org,
Jens Axboe <axboe@...nel.dk>, loongarch@...ts.linux.dev
Subject: Re: [PATCH 2/2] vfs: support statx(..., NULL, AT_EMPTY_PATH, ...)
On Wed, 3 Jul 2024 at 11:14, Christian Brauner <brauner@...nel.org> wrote:
>
> In any case, which one of these does a new architecture have to add for
> reasonable backward compatibility:
>
> fstat()
> fstat64()
> fstatat64()
>
> lstat()
> lstat64()
>
> stat()
> stat64()
> statx()
>
> newstat()
> newlstat()
> newfstat()
> newfstatat()
Well, I do think that a *new* architecture should indeed add all of
those, but make the 'struct stat' for all of them be the same.
So then if people do the system call rewriting thing - or just do the
manual system call thing with
syscall(__NR_xyz, ...)
it is all available, but it ends up being all the same code.
Wouldn't that be lovely?
Of course, I also happen to think that "new architecture" and "32-bit"
is just crazy to begin with, so honestly, I don't even care. 32-bit
architectures aren't really relevant for any new development, and I
think we should just codify that.
And on 64-bit architectures, the standard 'stat' works fine.
Linus
Powered by blists - more mailing lists