[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <eb05bc4e50464579a60b80ddfd596a6a@AcuMS.aculab.com>
Date: Fri, 28 Apr 2023 08:40:31 +0000
From: David Laight <David.Laight@...LAB.COM>
To: 'Linus Torvalds' <torvalds@...ux-foundation.org>,
Christian Brauner <brauner@...nel.org>
CC: Al Viro <viro@...iv.linux.org.uk>,
"linux-fsdevel@...r.kernel.org" <linux-fsdevel@...r.kernel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: RE: [GIT PULL] pidfd updates
From: Linus Torvalds
> Sent: 25 April 2023 17:29
>
> On Tue, Apr 25, 2023 at 5:34 AM Christian Brauner <brauner@...nel.org> wrote:
> >
> > Hell, you could even extend that proposal below to wrap the
> > put_user()...
> >
> > struct fd_file {
> > struct file *file;
> > int fd;
> > int __user *fd_user;
> > };
>
> So I don't like this extended version, but your proposal patch below
> looks good to me.
>
> Why? Simply because the "two-word struct" is actually a good way to
> return two values. But a three-word one would be passed on the stack.
>
> Both gcc and clang return small structs (where "small" is literally
> just two words) in registers, and it's part of most (all?) ABIs and
> we've relied on that before.
It is definitely architecture dependant.
x86-64 and arm-64 will return two 64bit values in registers.
x86-32 and arm-32 return two 32bit values on stack.
Pretty much everything passes short structures directly by value.
(I'm not sure about sparc-32 though, I'm sure it passed all
structures by reference back in the 1980s)
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
Powered by blists - more mailing lists