On Thu, 2015-05-21 at 19:12 +0200, Christoph Hellwig wrote: > On Thu, May 21, 2015 at 02:52:57AM -0700, Nicholas A. Bellinger wrote: > > The series is wrong on many levels, and I'm more than happy to comment > > on them individually. > > In general it's more helpful to explain those than just saying things > are wrong. Like I said, I'll comment on them individually later today. > > You'll need to focus on iscsi-target and tcm_qla2xxx specifically with > > real workloads and backend configurations to prove the changes are > > functionally correct beyond simple VM tests, otherwise I'm not even > > going to consider these type of changes. > > One thing that would really help is to have a pre defined set of test > available. The libiscsi test suite is pretty useful for CDB level > testing, but not that useful for error handling and doesn't touch > the vast control plane and shutdown case at all. If a maintainer > (rightfully!) demands testing coverage he also needs to point to tests. The point is the cases this code touches aren't exactly easy to verify, and starting with a fabric that nobody cares about is not the right way to go about it. Different vendors like Qlogic and Mellanox are doing their own internal verification, as well as vendors and users who are running it against VMware Workbench for ESX certification purposes. For people who are making major changes to this code, they will need to verify on at least one fabric that is widely used on real hardware with real backends before I'll consider merging it. So that said, I'd recommend Bart use whatever verification suite that Sandisk is using for SCST production usage, or use VMware workbench instead. --nab -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
