[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAOssrKcVddL5URQ0Vy79eQOscqTTK115Ro0Eqe8Q9kdkNJspCg@mail.gmail.com>
Date: Mon, 20 Apr 2020 10:58:40 +0200
From: Miklos Szeredi <mszeredi@...hat.com>
To: Stefan Metzmacher <metze@...ba.org>
Cc: Al Viro <viro@...iv.linux.org.uk>,
lkml <linux-kernel@...r.kernel.org>,
linux-fsdevel <linux-fsdevel@...r.kernel.org>,
Linux API <linux-api@...r.kernel.org>,
Eric Sandeen <sandeen@...deen.net>
Subject: Re: [PATCH] vfs: add faccessat2 syscall
On Sat, Apr 18, 2020 at 10:23 PM Stefan Metzmacher <metze@...ba.org> wrote:
>
> Am 18.04.20 um 21:00 schrieb Miklos Szeredi:
> > On Sat, Apr 18, 2020 at 8:36 PM Stefan Metzmacher <metze@...ba.org> wrote:
> >>
> >> Hi Miklos,
> >>
> >>> POSIX defines faccessat() as having a fourth "flags" argument, while the
> >>> linux syscall doesn't have it. Glibc tries to emulate AT_EACCESS and
> >>> AT_SYMLINK_NOFOLLOW, but AT_EACCESS emulation is broken.
> >>>
> >>> Add a new faccessat(2) syscall with the added flags argument and implement
> >>> both flags.
> >>>
> >>> The value of AT_EACCESS is defined in glibc headers to be the same as
> >>> AT_REMOVEDIR. Use this value for the kernel interface as well, together
> >>> with the explanatory comment.
> >>
> >> It would be nice if resolv_flags would also be passed in addition to the
> >> at flags.
> >> See:https://lore.kernel.org/linux-api/CAHk-=wiaL6zznNtCHKg6+MJuCqDxO=yVfms3qR9A0czjKuSSiA@mail.gmail.com/
> >>
> >> We should avoid expecting yet another syscall in near future.
> >
> > What is the objection against
> >
> > openat(... O_PATH)
> > foobarat(fd, AT_EMPTY_PATH, ...)
>
> openat2(), foobarat(), close() are 3 syscalls vs. just one.
That's not a good argument. We could have a million specialized
syscalls that all do very useful things. Except it would be a
nightmare in terms of maintenance...
"do one thing and do it well"
> As we have the new features available, I think it would be
> good to expose them to userspace for all new syscalls, so
> that applications can avoid boiler plate stuff around each syscall
> and get better performance in a world where context switches are not for
> free.
The io-uring guys are working on that problem, AFAIK.
Thanks,
Miklos
Powered by blists - more mailing lists