[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190830144934.GQ29594@unicorn.suse.cz>
Date: Fri, 30 Aug 2019 16:49:34 +0200
From: Michal Kubecek <mkubecek@...e.cz>
To: netdev@...r.kernel.org
Cc: Roopa Prabhu <roopa@...ulusnetworks.com>,
Jiri Pirko <jiri@...nulli.us>,
David Miller <davem@...emloft.net>,
Jakub Kicinski <jakub.kicinski@...ronome.com>,
David Ahern <dsahern@...il.com>,
Stephen Hemminger <sthemmin@...rosoft.com>, dcbw@...hat.com,
Andrew Lunn <andrew@...n.ch>, parav@...lanox.com,
Saeed Mahameed <saeedm@...lanox.com>,
mlxsw <mlxsw@...lanox.com>
Subject: Re: [patch net-next rfc 3/7] net: rtnetlink: add commands to add and
delete alternative ifnames
On Fri, Aug 30, 2019 at 07:35:23AM -0700, Roopa Prabhu wrote:
> On Wed, Aug 28, 2019 at 10:26 PM Michal Kubecek <mkubecek@...e.cz> wrote:
> >
> > On Wed, Aug 28, 2019 at 09:36:41PM -0700, Roopa Prabhu wrote:
> > >
> > > yes, correct. I mentioned that because I was wondering if we can
> > > think along the same lines for this API.
> > > eg
> > > (a) RTM_NEWLINK always replaces the list attribute
> > > (b) RTM_SETLINK with NLM_F_APPEND always appends to the list attribute
> > > (c) RTM_DELLINK with NLM_F_APPEND updates the list attribute
> > >
> > > (It could be NLM_F_UPDATE if NLM_F_APPEND sounds weird in the del
> > > case. I have not looked at the full dellink path if it will work
> > > neatly..its been a busy day )
> >
> > AFAICS rtnl_dellink() calls nlmsg_parse_deprecated() so that even
> > current code would ignore any future attribute in RTM_DELLINK message
> > (any kernel before the strict validation was introduced definitely will)
> > and it does not seem to check NLM_F_APPEND or NLM_F_UPDATE either. So
> > unless I missed something, such message would result in deleting the
> > network device (if possible) with any kernel not implementing the
> > feature.
>
> ok, ack. yes today it does. I was hinting if that can be changed to
> support list update with a flag like the RTM_DELLINK AF_BRIDGE does
> for vlan list del.
What I wanted to say is that it would have to be done in a way that
would make current kernel (or even older, e.g. one with no strict
attribute checking at all) reject or at least ignore such request,
not delete the device.
Michal
Powered by blists - more mailing lists