Re: [PATCH blktests] nvme: Add new test case about nvme rescan/reset/remove during IO

[Omar Sandoval <osandov@xxxxxxxxxxx>]

On Tue, Sep 03, 2019 at 04:17:52PM +0800, Yi Zhang wrote:
> Add one test to cover NVMe SSD rescan/reset/remove operation during
> IO, the steps found several issues during my previous testing, check
> them here:
> http://lists.infradead.org/pipermail/linux-nvme/2017-February/008358.html
> http://lists.infradead.org/pipermail/linux-nvme/2017-May/010259.html
> 
> Signed-off-by: Yi Zhang <yi.zhang@xxxxxxxxxx>
> ---
>  tests/nvme/031     | 43 +++++++++++++++++++++++++++++++++++++++++++
>  tests/nvme/031.out |  2 ++
>  2 files changed, 45 insertions(+)
>  create mode 100755 tests/nvme/031
>  create mode 100644 tests/nvme/031.out
> 
> diff --git a/tests/nvme/031 b/tests/nvme/031
> new file mode 100755
> index 0000000..4113d12
> --- /dev/null
> +++ b/tests/nvme/031
> @@ -0,0 +1,43 @@
> +#!/bin/bash
> +# SPDX-License-Identifier: GPL-3.0+
> +# Copyright (C) 2019 Yi Zhang <yi.zhang@xxxxxxxxxx>
> +#
> +# Test nvme pci adapter rescan/reset/remove operation during I/O
> +#
> +# Regression test for bellow two commits:
> +# http://lists.infradead.org/pipermail/linux-nvme/2017-May/010367.html
> +# 986f75c876db nvme: avoid to use blk_mq_abort_requeue_list()
> +# 806f026f9b90 nvme: use blk_mq_start_hw_queues() in nvme_kill_queues()
> +
> +. tests/nvme/rc
> +
> +DESCRIPTION="test nvme pci adapter rescan/reset/remove during I/O"
> +TIMED=1
> +
> +requires() {
> +	_have_fio
> +}
> +
> +device_requires() {
> +	_test_dev_is_nvme
> +}
> +
> +test_device() {
> +	echo "Running ${TEST_NAME}"
> +
> +	pdev="$(_get_pci_dev_from_blkdev)"
> +
> +	# start fio job
> +	_run_fio_rand_io --filename="$TEST_DEV" --size=1g \
> +		--ignore_error=EIO,ENXIO,ENODEV --group_reporting  &> /dev/null &
> +
> +	# do rescan/reset/remove operation
> +	echo 1 > /sys/bus/pci/devices/"${pdev}"/rescan
> +	echo 1 > /sys/bus/pci/devices/"${pdev}"/reset

My QEMU VM doesn't have the "reset" attribute, I'm guessing because of
this code in pci_create_capabilities_sysfs():

	if (dev->reset_fn) {
		retval = device_create_file(&dev->dev, &reset_attr);
		if (retval)
			goto error;
	}

We can skip the reset if the attribute doesn't exist.

> +	echo 1 > /sys/bus/pci/devices/"${pdev}"/remove
> +	sleep .5
> +	echo 1 > /sys/bus/pci/rescan
> +	sleep 5

Instead of sleep, we can kill and wait for fio.

Thanks!