On Tue, 2018-01-30 at 00:23 +0000, Bart Van Assche wrote:
> On Wed, 2018-01-17 at 18:14 -0500, Doug Ledford wrote:
> > On Tue, 2018-01-16 at 16:14 -0800, Bart Van Assche wrote:
> > > Hello Jason and Doug,
> > >
> > > This patch series not only adds RDMA/CM support to the SRP target driver but
> > > also fixes a number of race conditions in that driver.
> > >
> > > The RDMA/CM listener port number has to be specified as an ib_srpt kernel
> > > module parameter. The default value for that parameter is zero which means
> > > that RDMA/CM support is disabled.
> >
> > Since srpt is already configured via the lIO framework, wouldn't that be
> > a better place for the listen port? In fact, shouldn't it be part of a
> > portal like you have for iSERt?
>
> Wouldn't that be overkill to have one listen port per RDMA port? I think
> it will be easier for users if they have to configure the RDMA/CM port once
> instead of one time per RDMA port. How about using the following location in
> configfs for the RDMA/CM port:
>
> /sys/kernel/config/target/srpt/discovery_auth/rdma_cm_port
Hmmm...maybe the real answer here is to start considering how serious we
are about the RDMA_CM support in SRP. If we're serious, should someone
contact IANNA about a reserving port number and just use whatever they
give us? If we want to do the simple thing, a module option is fine,
while we decide on this issue, then I can see that being OK. I'm more
OK with using configfs if there's a chance we won't get a well known
reserved port and the config option will stick around long term. And I
tend to agree, per interface port configuration is probably not that
interesting. But the ability to specify something other than the
wildcard IP address to listen on, and the ability to specify more than
one IP address to listen on, are.
--
Doug Ledford <dledford@xxxxxxxxxx>
GPG KeyID: B826A3330E572FDD
Key fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDDAttachment:
signature.asc
Description: This is a digitally signed message part
