Thank you Conny for looking in to this. On Tue, 2021-11-02 at 16:31 +0100, Cornelia Huck wrote: ...snip... > device has bound to a driver, which was an information that was > completely missing up to now for subchannels. The main issue is that the > subchannel will get deregistered again (or has that been changed? Sorry, > I've been out of the loop for too long...) You are right. There is a change and we do not unregister the subchannel if we find the corresponding device is not operational or empty. ...snip... > > I think the potentially problematic case is "lots of I/O subchannels > with no valid device", and I think you can't get that under z/VM (which > will not give you those subchannels in the first place), but only on LPAR. Yes. But, this is in case of actual device. I just defined around 5k virtual dasd devices on zVM and did not enable them. i.e those subchannels are not blacklisted anymore, but they does not have an operational device. other than zVM, other than testing this patch on different available LPARs, we tried to mimic the device availability with a custom test-kernel. Basically, modified the subchannel initialization code and inform the subchannel about the presence of a device, which then later fails during SNSID. It is a horrid way to test it, but i think it could simulate some LPARs with more than 1000 non-operational devices connected on it. ...snip... > - non-I/O subchannels (IIRC you can't have CHSC subchannels under z/VM, > don't know about EADM, so that would need to be done on an LPAR as > well) Thanks. Most of my tests were on IO-subchannel. I shall try few tests on CHSC and EADM. > - interaction with driverctl (and maybe mdevctl) for vfio-ccw I have done couple of tests with driverctl. Also, i was wondering if vfio-ccw can make use of 'dev_busid' sysfs-interface under each subchannels while writing the rules. This is totally different topic, which i do not want to mix up in this thread. ...snip...
