Re: [PATCH rdma-next V2 5/5] IB/core: Integrate IB address resolution module into core

On Mon, May 16, 2016 at 02:39:10PM -0400, Doug Ledford wrote:

On 05/16/2016 02:27 PM, Jason Gunthorpe wrote:

quoted

On Mon, May 16, 2016 at 01:42:34PM -0400, Doug Ledford wrote:

quoted

Can you build netlink in and then init the ib_addr module after the
netlink init is complete?  Wouldn't that resolve the dependency ordering
issue without changing the module names?

Why are you so worried about ib_addr's module name? Is that actually
hand loaded instead of relying in the implicit dependencies???

It's hand unloaded.  Depending on the release of software you are
talking about, the init script has the ability to unload/reload the IB
stack.  In that case, it has to know about ib_addr so that it can unload
after ib_core.  I should know, when the order changed from ib_core first
and ib_addr second to the opposite it broke the Red Hat init script for
a release, including all the bug reports about the rdma stack failing to
shut down properly because it was unloading modules in the wrong order.
This falls in the same category of a lot of other things: you don't
break user space if you don't have to.

I think the reason that this thread fails to die is our trouble to
understand if code refactoring is enough justification to get rid of all
these modules.

All guides/scripts which I saw, instructs to load all these ib_...
modules together and it looks redundant to have such large number of
them [1].

[1]
http://www.rdmamojo.com/2014/11/08/working-rdma-ubuntu/#Starting_the_RDMA_services

quoted

Is failure to load a module going to break any existing scripts?

Worse comes to worse, just make dummy empty ib_addr.ko, but
I've never seen that in-kernel before.

Jason

-- 
Doug Ledford [off-list ref]
              GPG KeyID: 0E572FDD