[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200615213843.2ojgm3ob7kp3qq6a@wittgenstein>
Date: Mon, 15 Jun 2020 23:38:43 +0200
From: Christian Brauner <christian.brauner@...ntu.com>
To: Jan Stancek <jstancek@...hat.com>
Cc: Cyril Hrubis <chrubis@...e.cz>,
kernel test robot <rong.a.chen@...el.com>,
"Eric W. Biederman" <ebiederm@...ssion.com>,
Jann Horn <jannh@...gle.com>,
LKML <linux-kernel@...r.kernel.org>, lkp@...ts.01.org,
Aleksa Sarai <cyphar@...har.com>,
Michael Kerrisk <mtk.manpages@...il.com>, ltp@...ts.linux.it,
Serge Hallyn <serge@...lyn.com>
Subject: Re: [LTP] 303cc571d1: ltp.setns01.fail
On Mon, Jun 15, 2020 at 10:03:01AM -0400, Jan Stancek wrote:
>
>
> ----- Original Message -----
> > Hi!
> > > setns01 6 TFAIL : setns01.c:176: regular file fd exp_errno=22:
> > > errno=EBADF(9): Bad file descriptor
> > > setns01 0 TINFO : setns(12, 0x20000)
> > > setns01 7 TFAIL : setns01.c:176: regular file fd exp_errno=22:
> > > errno=EBADF(9): Bad file descriptor
> > > setns01 0 TINFO : setns(12, 0x40000000)
> > > setns01 8 TFAIL : setns01.c:176: regular file fd exp_errno=22:
> > > errno=EBADF(9): Bad file descriptor
> > > setns01 0 TINFO : setns(12, 0x20000000)
> > > setns01 9 TFAIL : setns01.c:176: regular file fd exp_errno=22:
> > > errno=EBADF(9): Bad file descriptor
> > > setns01 0 TINFO : setns(12, 0x4000000)
> > > setns01 10 TFAIL : setns01.c:176: regular file fd exp_errno=22:
> > > errno=EBADF(9): Bad file descriptor
> >
> > The messages here are a bit cryptic, I will fix that later on, but what
> > it means here is that the errno has changed from EINVAL to EBADF in a
> > case we pass file descriptor to a regular file to setns().
>
> I posted a series that accepts both errnos about week ago:
> https://lists.linux.it/pipermail/ltp/2020-June/017467.html
When you used to pass an fd that referred to an open file but the file
was not a nsfd the kernel would report EINVAL. Since the pidfd support
this now reports EBADF. While that is a user visible change it is not a
particular worrying one but something we should still fix. Here's the
patch that fixes this:
diff --git a/kernel/nsproxy.c b/kernel/nsproxy.c
index b03df67621d0..cd356630a311 100644
--- a/kernel/nsproxy.c
+++ b/kernel/nsproxy.c
@@ -531,7 +531,7 @@ SYSCALL_DEFINE2(setns, int, fd, int, flags)
} else if (!IS_ERR(pidfd_pid(file))) {
err = check_setns_flags(flags);
} else {
- err = -EBADF;
+ err = -EINVAL;
}
if (err)
goto out;
I'll send a pr for this to Linus this week (or next since I'm on
vacation this week) and get this fixed. Thanks for spotting this. What's
the Reported-by: line format that LTP uses?
Christian
Powered by blists - more mailing lists