[PATCH blktests 0/2] nvme: check that nvme-cli supports json output format

[PATCH blktests 0/2] nvme: check that nvme-cli supports json output format

[Shin'ichiro Kawasaki]

The commit 444f0df9f0c5 ("nvme: check for controller device in
_nvme_connect_subsys()") assumed that the nvme command always supports
json output format. However, when nvme-cli is built on a system without
the json-c library, the nvme command does not support the format. This
leads to silent failures in many test cases in the nvme test group [1].

This series address the problem. The first patch introduces a helper
function to prepare for the fix. The second patch fixes the problem by
checking that the nvme command supports the json format.

[1] https://lore.kernel.org/linux-nvme/20250519054905.GA6156@lst.de/

Shin'ichiro Kawasaki (2):
  common/nvme: introduce _have_nvme_cli_with_json_support()
  nvme/rc, md/001: check that nvme-cli supports json output format

 common/nvme   | 12 +++++++++++-
 tests/md/001  |  1 +
 tests/nvme/rc |  2 +-
 3 files changed, 13 insertions(+), 2 deletions(-)

-- 
2.49.0

[PATCH blktests 1/2] common/nvme: introduce _have_nvme_cli_with_json_support()

[Shin'ichiro Kawasaki]

To check if the nvme command supports json output format, introduce the
helper function _have_nvme_cli_with_json_support(), which verifies the
option "--output-format=" works. While at it, replace the short option
"-o" in _nvme_connect_subsys() with the long option "--output-format=".

Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
---
 common/nvme | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/common/nvme b/common/nvme
index ac1bbe7..04d167b 100644
--- a/common/nvme
+++ b/common/nvme
@@ -62,6 +62,16 @@ _require_nvme_trtype_is_fabrics() {
 	return 0
 }
 
+_have_nvme_cli_with_json_support() {
+	_have_program nvme || return $?
+
+	if ! nvme list --output-format=json &> /dev/null; then
+		SKIP_REASONS+=("nvme-cli does not support json output format")
+		return 1
+	fi
+	return 0
+}
+
 _nvme_fcloop_add_rport() {
 	local local_wwnn="$1"
 	local local_wwpn="$2"
@@ -415,7 +425,7 @@ _nvme_connect_subsys() {
 	if [[ ${concat} = true ]]; then
 		ARGS+=(--concat)
 	fi
-	ARGS+=(-o json)
+	ARGS+=(--output-format=json)
 	connect=$(nvme connect "${ARGS[@]}" 2> /dev/null)
 
 	# Wait until device file and sysfs attributes get ready
-- 
2.49.0

Re: [PATCH blktests 1/2] common/nvme: introduce _have_nvme_cli_with_json_support()

[Chaitanya Kulkarni]

On 5/19/25 23:18, Shin'ichiro Kawasaki wrote:
> To check if the nvme command supports json output format, introduce the
> helper function _have_nvme_cli_with_json_support(), which verifies the
> option "--output-format=" works. While at it, replace the short option
> "-o" in _nvme_connect_subsys() with the long option "--output-format=".
>
> Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
> ---
>

Looks good.

Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>

-ck

[PATCH blktests 2/2] nvme/rc, md/001: check that nvme-cli supports json output format

[Shin'ichiro Kawasaki]

When nvme-cli is built on a system without the json-c library, the nvme
command does not support the json output format. Consequently, the
"-o json" option, or in the long form "--output-format=json", of the
nvme command fails. It causes the helper function _nvme_connect_subsys()
to break. This leads to silent failures in many test cases in the nvme
test group as well as the test case md/001.

To inform blktests users this prerequisite, call the helper function
_have_nvme_cli_with_json_support(). As to the nvme test group, do the
check in group_requires() so that the check is done only once for all of
the test cases in the group. If the format is unsupported, the nvme test
group and the test case md/001 are skipped, and the reason to skip is
printed. This change makes the nvme command existence check in
_nvme_requires() redundant, then remove it.

Link: https://lore.kernel.org/linux-nvme/20250519054905.GA6156@lst.de/
Fixes: 444f0df9f0c5 ("nvme: check for controller device in _nvme_connect_subsys()")
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
---
 tests/md/001  | 1 +
 tests/nvme/rc | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/tests/md/001 b/tests/md/001
index f1d92c8..ca7c56a 100755
--- a/tests/md/001
+++ b/tests/md/001
@@ -26,6 +26,7 @@ requires() {
 	_have_driver dm-mod
 	_have_driver raid1
 
+	_have_nvme_cli_with_json_support
 	_require_nvme_trtype tcp
 	_have_brd
 }
diff --git a/tests/nvme/rc b/tests/nvme/rc
index d86dc13..9dbd1ce 100644
--- a/tests/nvme/rc
+++ b/tests/nvme/rc
@@ -55,7 +55,6 @@ _set_nvmet_blkdev_type() {
 }
 
 _nvme_requires() {
-	_have_program nvme
 	_require_nvme_test_img_size 4m
 	case ${nvme_trtype} in
 	loop)
@@ -123,6 +122,7 @@ group_setup() {
 group_requires() {
 	_have_root
 	_NVMET_TRTYPES_is_valid
+	_have_nvme_cli_with_json_support
 }
 
 group_device_requires() {
-- 
2.49.0

Re: [PATCH blktests 2/2] nvme/rc, md/001: check that nvme-cli supports json output format

[Chaitanya Kulkarni]

On 5/19/25 23:18, Shin'ichiro Kawasaki wrote:
> When nvme-cli is built on a system without the json-c library, the nvme
> command does not support the json output format. Consequently, the
> "-o json" option, or in the long form "--output-format=json", of the
> nvme command fails. It causes the helper function _nvme_connect_subsys()
> to break. This leads to silent failures in many test cases in the nvme
> test group as well as the test case md/001.
>
> To inform blktests users this prerequisite, call the helper function
> _have_nvme_cli_with_json_support(). As to the nvme test group, do the
> check in group_requires() so that the check is done only once for all of
> the test cases in the group. If the format is unsupported, the nvme test
> group and the test case md/001 are skipped, and the reason to skip is
> printed. This change makes the nvme command existence check in
> _nvme_requires() redundant, then remove it.
>
> Link:https://lore.kernel.org/linux-nvme/20250519054905.GA6156@lst.de/
> Fixes: 444f0df9f0c5 ("nvme: check for controller device in _nvme_connect_subsys()")
> Reported-by: Christoph Hellwig<hch@lst.de>
> Signed-off-by: Shin'ichiro Kawasaki<shinichiro.kawasaki@wdc.com>

Looks good.

Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>

-ck

Re: [PATCH blktests 0/2] nvme: check that nvme-cli supports json output format

[Daniel Wagner]

Looks good to me.

For both patches:

Reviewed-by: Daniel Wagner <dwagner@suse.de>

Re: [PATCH blktests 0/2] nvme: check that nvme-cli supports json output format

[Christoph Hellwig]

The changes looks great:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Although to me they would be easier to follow in one patch instead
of two.

Re: [PATCH blktests 0/2] nvme: check that nvme-cli supports json output format

[Shinichiro Kawasaki]

On May 20, 2025 / 15:18, Shin'ichiro Kawasaki wrote:
> The commit 444f0df9f0c5 ("nvme: check for controller device in
> _nvme_connect_subsys()") assumed that the nvme command always supports
> json output format. However, when nvme-cli is built on a system without
> the json-c library, the nvme command does not support the format. This
> leads to silent failures in many test cases in the nvme test group [1].
> 
> This series address the problem. The first patch introduces a helper
> function to prepare for the fix. The second patch fixes the problem by
> checking that the nvme command supports the json format.
> 
> [1] https://lore.kernel.org/linux-nvme/20250519054905.GA6156@lst.de/

FYI, I applied this series.