[PATCH blktests] block tests: nvme metadata passthrough

[Shinichiro Kawasaki]

On Jun 09, 2025 / 01:25, Anuj gupta wrote:
> On Sat, Jun 7, 2025 at 6:25 PM Shinichiro Kawasaki
> <shinichiro.kawasaki at wdc.com> wrote:
> >
> > On Jun 05, 2025 / 17:30, Keith Busch wrote:
> > > From: Keith Busch <kbusch at kernel.org>
> > >
> > > Get more coverage on nvme metadata passthrough. Specifically in this
> > > test, read-only metadata is targeted as this had been a gap in previous
> > > test coveraged.
> >
> > Thanks for the patch. I ran the test case on the kernel v6.15, and it passed.
> > Is this pass result expected? I guess this test case intends to extend test
> > coverage, and does not intend to recreate the failure reported in the Link.
> > So, I'm guessing the test case should pass with v6.15 kernel.
> 
> As Keith mentioned in that thread [1], this test can catch the issue
> only in the more unusual or corner-case configurations.
> 
> Additionally, if your NVMe device is not formatted with protection
> information, the test will exit early and report success, as the
> integrity path won't be exercised in that case.
> 
> [1] https://lore.kernel.org/linux-block/aD-J9mzq_bJe26rD@kbusch-mbp/

I see, thanks for the explanation. Then I wish the expected device conditions to
be described in the test case comment. Even if blktests can not prepare devices
with the condition automatically, I think it's worth documenting.

In daily blktests run, I use a normal QEMU nvme device, but running the test
case with this device does not extend the test coverage much, probably.
Can we prepare the best device for the new test case using QEMU? I hope that
some QEMU boot options and/or nvme format commands can prepare it.