[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <dc6b7446-ae7b-748d-25dc-565bf9eea14b@gmail.com>
Date: Mon, 21 Jan 2019 10:40:42 -0700
From: David Ahern <dsahern@...il.com>
To: Matt Ellison <matt@...oyo.io>
Cc: netdev@...r.kernel.org, stephen@...workplumber.org
Subject: Re: [PATCH v2 iproute2] ip: support for xfrm interfaces
On 1/21/19 10:05 AM, Matt Ellison wrote:
> On Mon, 21 Jan 2019 09:14:52 -0700 David Ahern <dsahern@...il.com> wrote:
>
>> You always add IF_ID even if not set by user. The kernel code does not
>> appear to require it so why pass a default value?
>
> 0 (the default) is a valid IF_ID, so setting an interface with a non-zero IF_ID
> back to 0 is possible. I think the better solution would be to check for existing
> values so an "ip link set" doesn't try and automatically use 0 if unsepcified.
I would expect IF_ID to only be added to the request if passed on the
command line. That should handle the set link case.
>
>> The kernel code does appear to require this parameter, so why have
>> this requirement in iproute2?
>
> Looks to be required, without the check I get:
> RTNETLINK answers: No such device
>
>> What about IFLA_XFRM_LINK?
>
> It shows up as the parent interface when you ip link show:
>
> 13: xfrm2@...n0: <NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
> link/none 48:e2:44:f6:77:51 brd ff:ff:ff:ff:ff:ff
>
> But I can print it again if you think I should.
>
ok, so IFLA_XFRM_LINK duplicates IFLA_LINK. no need to print twice.
Powered by blists - more mailing lists