From mboxrd@z Thu Jan 1 00:00:00 1970 From: logang@deltatee.com (Logan Gunthorpe) Date: Fri, 13 Apr 2018 11:37:16 -0600 Subject: [RFC PATCH 8/8] nvmet: add configfs interface for target passthru In-Reply-To: <20180413173004.GC23674@lst.de> References: <20180330065747.20962-1-chaitanya.kulkarni@wdc.com> <20180330065747.20962-9-chaitanya.kulkarni@wdc.com> <20180413173004.GC23674@lst.de> Message-ID: <12238d4c-c01d-9590-2357-dbeabff23dfc@deltatee.com> On 13/04/18 11:30 AM, Christoph Hellwig wrote: > Maybe. But how would you design it? Have a passthrough filed under > the subsystem that you can echo a controller name to? And then make > the exclusive vs creating namespaces? Yes. It only really requires one field (the path to the nvme char device) and making it exclusive to namespaces doesn't seem that unintuitive. An error and appropriate dmesg if one is set before the other would be pretty clear and quickly understood (compared to a whole new top level directory named 'pt'). When I'm playing around with sysfs/configfs/debugfs interfaces I rarely read the documentation and just watch dmesg if I see an error returned. When I first played around with this patchset the current interface required me to look at the code to understand what was going on (there was no documentation). > Ugg. Hell no! Way too much magic. Fair point. Logan