From mboxrd@z Thu Jan 1 00:00:00 1970 From: hch@infradead.org (Christoph Hellwig) Date: Mon, 19 Jun 2017 23:48:00 -0700 Subject: remove long broken SCSI to NVMe translations In-Reply-To: <5943E3BC.6050807@hpe.com> References: <20170614063537.1044-1-hch@lst.de> <20170614150025.GD26073@localhost.localdomain> <20170614150138.GA9428@lst.de> <20170616083151.GB14698@infradead.org> <5943E3BC.6050807@hpe.com> Message-ID: <20170620064800.GA11019@infradead.org> On Fri, Jun 16, 2017@09:57:16AM -0400, Linda Knippers wrote: > > Everything that could be done using the scsi emulation can be trivially > > done using nvme-cli, as they should have done from day 0. > > I don't quite know the history of our use of the scsi translation but I > suspect it's because we adapted some existing utility to also support nvme. > > David Milburn's patch is very simple. Can we please fix the bug for now > so we have a little time to fix our tools? I don't think I want to expand usage of it. The code only "worked" for about two years, during which not too many NVMe products even existed. And I say "worked" because I found countless of sever bugs (use after free, memory overruns, etc) issues when I first ran a SCS I testsuite over it, which obviously didn't even cover the commands you care about. You might have a slightly better leaverage with distros already shipping this code, but enabling it upstream does not seem in any way helpful.