[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190813060758.GD2428@nanopsycho>
Date: Tue, 13 Aug 2019 08:07:58 +0200
From: Jiri Pirko <jiri@...nulli.us>
To: Jakub Kicinski <jakub.kicinski@...ronome.com>
Cc: David Ahern <dsahern@...il.com>, netdev@...r.kernel.org,
davem@...emloft.net, stephen@...workplumber.org, mlxsw@...lanox.com
Subject: Re: [patch net-next v3 0/3] net: devlink: Finish network namespace
support
Tue, Aug 13, 2019 at 03:11:00AM CEST, jakub.kicinski@...ronome.com wrote:
>On Mon, 12 Aug 2019 18:24:41 -0600, David Ahern wrote:
>> On 8/12/19 7:47 AM, Jiri Pirko wrote:
>> > From: Jiri Pirko <jiri@...lanox.com>
>> >
>> > Devlink from the beginning counts with network namespaces, but the
>> > instances has been fixed to init_net. The first patch allows user
>> > to move existing devlink instances into namespaces:
>> >
>> > $ devlink dev
>> > netdevsim/netdevsim1
>> > $ ip netns add ns1
>> > $ devlink dev set netdevsim/netdevsim1 netns ns1
>> > $ devlink -N ns1 dev
>> > netdevsim/netdevsim1
>> >
>> > The last patch allows user to create new netdevsim instance directly
>> > inside network namespace of a caller.
>>
>> The namespace behavior seems odd to me. If devlink instance is created
>> in a namespace and never moved, it should die with the namespace. With
>> this patch set, devlink instance and its ports are moved to init_net on
>> namespace delete.
>
>If the devlink instance just disappeared - that'd be a very very strange
>thing. Only software objects disappear with the namespace.
>Netdevices without ->rtnl_link_ops go back to init_net.
Agreed. It makes sense to be moved to init_net.
Powered by blists - more mailing lists