On Fri, Dec 23, 2022 at 06:41:37PM +0530, Kanchan Joshi wrote:
> On Wed, Dec 21, 2022 at 02:34:37AM -0800, Luis Chamberlain wrote:
> > This does basic rand-read testing of the character device of a
> > conventional NVMe drive.
> >
> > Signed-off-by: Luis Chamberlain <mcgrof@xxxxxxxxxx>
> > ---
> > tests/nvme/046 | 42 ++++++++++++++++++++++++++++++++++++++++++
> > tests/nvme/046.out | 2 ++
> > 2 files changed, 44 insertions(+)
> > create mode 100755 tests/nvme/046
> > create mode 100644 tests/nvme/046.out
> >
> > diff --git a/tests/nvme/046 b/tests/nvme/046
> > new file mode 100755
> > index 000000000000..3526ab9eedab
> > --- /dev/null
> > +++ b/tests/nvme/046
> > @@ -0,0 +1,42 @@
> > +#!/bin/bash
> > +# SPDX-License-Identifier: GPL-3.0+
> > +# Copyright (C) 2022 Luis Chamberlain <mcgrof@xxxxxxxxxx>
> > +#
> > +# This does basic sanity test for the nvme character device. This is a basic
> > +# test and if it fails it is probably very likely other nvme character device
> > +# tests would fail.
> > +#
> > +. tests/nvme/rc
> > +
> > +DESCRIPTION="basic rand-read io_uring_cmd engine for nvme-ns character device"
> > +QUICK=1
> > +
> > +requires() {
> > + _nvme_requires
> > + _have_fio
> > +}
> > +
> > +device_requires() {
> > + _require_test_dev_is_nvme
> > +}
> > +
> > +test_device() {
> > + echo "Running ${TEST_NAME}"
> > + local ngdev=${TEST_DEV/nvme/ng}
> > + local fio_args=(
> > + --size=1M
> > + --cmd_type=nvme
> > + --filename="$ngdev"
> > + --time_based
> > + --runtime=10
> > + ) &&
>
> Is this && needed?
This form was inspired by commit 238c7e0b by Bart, but yeah you're
right, I can't see any reason for it, so we can clean zbd/010 from it too.
>
> > + _run_fio_rand_iouring_cmd "${fio_args[@]}" >>"${FULL}" 2>&1 ||
>
> Something to change here (and therefore in other patches too).
> If we change "cmd_type = something_random", test continues to show the
> success while it should show failure.
Definitely no bueno.
> How about changing above line to:
> _run_fio_rand_iouring_cmd "${fio_args[@]}" || fail=true
We'd loose the 046.full log then.
If we just return $? at the end of _run_fio_rand_iouring_cmd() that
seems to fix the undetected error. Whatyda think?
I noticed an odd thing in the last two patches which work for zone
storage, if I change the runtime it doesn't take longer, so I think
something is still off there too... can you take a look?
Luis
