[PATCH 2/5] node_device: detect CSS devices

[Erik Skultety]

On Fri, Sep 11, 2020 at 06:11:49PM +0200, Boris Fiuczynski wrote:
> On 9/9/20 9:03 AM, Cornelia Huck wrote:
> > > > [snip]
> > > > >
> > > > > +static int
> > > > > +udevProcessCSS(struct udev_device *device,
> > > > > +               virNodeDeviceDefPtr def)
> > > > > +{
> > > > > +    /* do not process EADM and CHSC devices to keep the list sane */
> > > > > +    if (STREQ(def->driver, "eadm_subchannel") ||
> > > > > +        STREQ(def->driver, "chsc_subchannel"))
> > > > [2] Also mentions message subchannel, although apparently there are no Linux
> > > > kernel drivers for that yet, IIUC that one would have to be added here as well
> > > > as some point, is that correct?
> > > We have never seen those, but we would want to add them here if they
> > > have no relevance for the user.
> > I think you want to filter message subchannels as well, my assumption
> > is that libvirt will only care about I/O subchannels.
> >
> In
> https://www.kernel.org/doc/html/latest/driver-api/s390-drivers.html
> it is stated that for
> Message subchannels. No Linux driver currently exists.
>
> Therefore I do not know how I could currently filter message subchannels
> out. Due you want me to reverse the logic and just filter out everything but
> "io_subchannel" instead (which I would regard as fencing the unknown)?

Well, I don't see a problem filtering out everything but the single thing you
actually care about, rather than enumerating the things you don't care about
and won't care about in the future, so yeah, I'd go with that.

Erik