[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20150420233520.GC1409@phlsvsds.ph.intel.com>
Date: Mon, 20 Apr 2015 19:35:21 -0400
From: "ira.weiny" <ira.weiny@...el.com>
To: Jason Gunthorpe <jgunthorpe@...idianresearch.com>
Cc: Haggai Eran <haggaie@...lanox.com>,
Doug Ledford <dledford@...hat.com>,
Roland Dreier <roland@...nel.org>,
Sean Hefty <sean.hefty@...el.com>, linux-rdma@...r.kernel.org,
netdev@...r.kernel.org, Liran Liss <liranl@...lanox.com>,
Guy Shapiro <guysh@...lanox.com>,
Shachar Raindel <raindel@...lanox.com>,
Yotam Kenneth <yotamke@...lanox.com>
Subject: Re: [PATCH v2 07/11] IB/cm: Add network namespace support
On Mon, Apr 20, 2015 at 11:06:59AM -0600, Jason Gunthorpe wrote:
> On Mon, Apr 20, 2015 at 12:03:38PM +0300, Haggai Eran wrote:
> > From: Guy Shapiro <guysh@...lanox.com>
> >
> > Add namespace support to the IB-CM layer.
>
> > - Each CM-ID now has a network namespace it is associated with, assigned at
> > creation. This namespace is used as needed during subsequent action on the
> > CM-ID or related objects.
>
> There is really something weird about this layering. At the CM layer
> there should be no concept of an IP address, it only deals with GIDs.
>
> So how can a CM object have a network namespace associated with it?
>
> > {
> > av->port = port;
> > av->pkey_index = wc->pkey_index;
> > ib_init_ah_from_wc(port->cm_dev->ib_device, port->port_num, wc,
> > - grh, &av->ah_attr, &init_net);
> > + grh, &av->ah_attr, net);
>
> There is something deeply wrong with adding network namespace
> arguments to verbs.
>
> For rocee the gid index clearly specifies the network namespace
> to use, so much of this should go away and have rocee get the
> namespace from the gid index.
>
> Ie in ib_init_ah_from_wc we have the ib_wc which contains the sgid
> index.
>
> I'm really not excited at how many places are gaining a net when those
> layers shouldn't even need to care about IP layer details. Just acting
> as a pass through for rocee doesn't make sense.
>
I had the same feeling when I saw the addition of the network namespace to the
MAD code, especially the RMPP code.
It seems like there should be a better way to deal with this. My gut says that
the namespace should be handled separate from the ib_init_ah_from_wc. Perhaps
as a secondary call used only when the namespace is needed? But I'm not sure
when it is appropriate/needed.
Ira
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists