[ovs-dev] [PATCH] datapath: internal-dev: Disable rtnl_link_ops register under linux < 3.17

[张东亚]

I am almost forgeting this patch :(

I do not think it's easy to check upstream commit because the change does
not like add a field to a struct.

And I think if we add the dellink support, we must detach the vport from
the datapath because that's the reason which make the vport hash bucket
list corrupt, while the reason rtnl_link_ops is introduced  for a display
issue,  I do not think that's worthy and have not try that yet.

2017-01-25 1:30 GMT+08:00 Pravin Shelar <pshelar at ovn.org>:

> On Mon, Jan 23, 2017 at 10:04 AM, Jarno Rajahalme <jarno at ovn.org> wrote:
> > Thomas & Pravin,
> >
> > Apparently the commit 5282e284ac57 (“datapath: introduce rtnl ops stub”)
> > applied to OVS two years ago is not compatible with Linux 3.16 or
> earlier,
> > as the corresponding upstream change was only applied in Linux 3.17. For
> > Linux 3.16 and earlier, in your opinion, is the right solution to
> > effectively revert this commit, or would there be a way to make the ops
> stub
> > compatible with these Linux versions by somehow amending the stub
> (.setup,
> > .dellink, etc.)?
> >
> Sorry for the delay. the patch fell off my radar. I have couple of
> comments below.
>
>
> >   Jarno
> >
> > Begin forwarded message:
> >
> > From: Markos Chandras <mchandras at suse.de>
> > Subject: Re: [ovs-dev] [PATCH] datapath: internal-dev: Disable
> rtnl_link_ops
> > register under linux < 3.17
> > Date: January 23, 2017 at 7:13:03 AM PST
> > To: ovs-dev at openvswitch.org
> >
> > Hi,
> >
> > On 12/14/2016 03:44 AM, Zhang Dongya wrote:
> >
> > From: "fortitude.zhang" <fortitude.zhang at gmail.com>
> >
> > Under linux < 3.17, b0ab2fabb5b91da99c189db02e91ae10bc8355c5 is not
> > introduced,
> > which allow internal-dev created by openvswitch datapath deleted by using
> > rtnl
> > interface, this causes data related to internal-dev not freed and stops
> > datapath
> > working correctly.
> >
>
> What about backports? can we detect if the upstream commit is
> backported by distribution kernel?
> if it is not possible to detect it, have you tried setting
> unregister_netdevice_queue() to vport rtnl dellink operation or
> something similar?
>
> Thanks,
> Pravin.
>
> > Signed-off-by: fortitude.zhang <fortitude.zhang at gmail.com>
> >
> >
> > I recently encountered this bug and this commit fixed it for me. So...
> >
> > Tested-by: Markos Chandras <mchandras at suse.de>
> >
> > It might also worth backporting it to stable branches as well.
> >
> > --
> > markos
> >
> > SUSE LINUX GmbH | GF: Felix Imendörffer, Jane Smithard, Graham Norton
> > HRB 21284 (AG Nürnberg) Maxfeldstr. 5, D-90409, Nürnberg
> > _______________________________________________
> > dev mailing list
> > dev at openvswitch.org
> > https://mail.openvswitch.org/mailman/listinfo/ovs-dev
> >
> >
>