From mboxrd@z Thu Jan 1 00:00:00 1970 From: hch@lst.de (Christoph Hellwig) Date: Mon, 22 Jun 2015 17:39:00 +0200 Subject: [PATCH] NVMe: Fix IO for extended metadata formats In-Reply-To: References: <1434733650-16440-1-git-send-email-keith.busch@intel.com> <20150622053417.GA29813@lst.de> Message-ID: <20150622153900.GA7952@lst.de> On Mon, Jun 22, 2015@03:17:07PM +0000, Busch, Keith wrote: > Similar to SG_IO, we can submit arbitrary length commands using the nvme > passthrough. It seems to work. Here's a test I did using the "nvme-cli" > program. Yeah, I was mostly worried about the request mapping code tripping up on these unaligned sized. I don't have a device that supports it so I couldn't test it myself. Btw, given your test below: I had thought for a while to add unit tests to the nvme-cli package for cases like the one you had below. The use case would be mostly to verify the driver, less the hardware unless the two are tangled up or the hardware specific cases are simple enough. I've had good succeed with the cunit based SCSI testsuite in libiscsi, so I could come up with a skeleton for it.