Re: ping (iputils) review (call for help)

From: Petr Vorel <pvorel@suse.cz>
To: David Ahern <dsahern@gmail.com>
Cc: netdev@vger.kernel.org, "David S. Miller" <davem@davemloft.net>,
	"Michal Kubecek" <mkubecek@suse.cz>,
	"Toke Høiland-Jørgensen" <toke@redhat.com>,
	"Paolo Abeni" <pabeni@redhat.com>,
	"Eric Dumazet" <edumazet@google.com>,
	"Jakub Kicinski" <kuba@kernel.org>,
	"Xin Long" <lucien.xin@gmail.com>,
	"Vasiliy Kulikov" <segoon@openwall.com>,
	"Hideaki YOSHIFUJI" <yoshfuji@linux-ipv6.org>,
	"Sami Kerola" <kerolasa@iki.fi>
Subject: Re: ping (iputils) review (call for help)
Date: Wed, 9 Nov 2022 23:01:47 +0100	[thread overview]
Message-ID: <Y2wjS/xkCtRrKXhs@pevik> (raw)
In-Reply-To: <d47c3f41-2977-3ffb-5c99-953088727a4b@gmail.com>

Hi David,

first, thanks a lot for having a look!

> On 11/3/22 5:30 AM, Petr Vorel wrote:
> > Hi,

> > I'm sorry to bother you about userspace. I'm preparing new iputils release and
> > I'm not sure about these two patches.  As there has been many regressions,
> > review from experts is more than welcome.

> > If you have time to review them, it does not matter if you post your
> > comments/RBT in github or here (as long as you keep Cc me so that I don't
> > overlook it).

> > BTW I wonder if it make sense to list Hideaki YOSHIFUJI as NETWORKING
> > IPv4/IPv6 maintainer. If I'm not mistaken, it has been a decade since he was active.

> > * ping: Call connect() before sending/receiving
> > https://github.com/iputils/iputils/pull/391
> > => I did not even knew it's possible to connect to ping socket, but looks like
> > it works on both raw socket and on ICMP datagram socket.

> no strong opinion on this one. A command line option to use connect
> might be better than always doing the connect.
I was thinking about it, as it'd be safer in case of some regression.
If there is no other opinion I'll probably go this way, although I generally
prefer not adding more command line options.

> > * ping: revert "ping: do not bind to device when destination IP is on device
> > https://github.com/iputils/iputils/pull/396
> > => the problem has been fixed in mainline and stable/LTS kernels therefore I
> > suppose we can revert cc44f4c as done in this PR. It's just a question if we
> > should care about people who run new iputils on older (unfixed) kernels.

> I agree with this change. If a user opts for device binding, the command
> should not try to guess if it is really needed.

I guess Sami Kerola (the patch author) preferred ping functionality on wrongly
used ping's -I option. I've seen it the same even it causes kselftest
regressions and wait a bit longer. But ok, anybody who would care should
update kernel. I'll include this in upcoming iputils release.

Kind regards,
Petr