[PATCH blktests 0/2] nvme/060: avoid false-positive failures

[PATCH blktests 0/2] nvme/060: avoid false-positive failures

[Shin'ichiro Kawasaki]

The test case nvme/060 fails often due to a couple of expected errors.
One of them is observed for all transports, and the other is unique to
the rdma transport. The two patches in this series suppress these two
expected errors respectively.

Shin'ichiro Kawasaki (2):
  nvme/060: suppress state file write error
  nvme/060: suppress "nvme connect" failure for rdma transport

 tests/nvme/060 | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

-- 
2.54.0

[PATCH blktests 1/2] nvme/060: suppress state file write error

[Shin'ichiro Kawasaki]

The test case nvme/060 sometimes fails with the message below:

nvme/060 (tr=rdma) (test nvme fabrics target reset)          [failed]
    runtime    ...  27.967s
    --- tests/nvme/060.out	2026-01-19 04:30:04.820982813 +0000
    +++ /home/fedora/blktests/results/nodev_tr_rdma/nvme/060.out.bad	2026-01-19 06:44:49.223643735 +0000
    @@ -1,2 +1,3 @@
     Running nvme/060
    +_: line 1: /sys/kernel/debug/nvmet//blktests-subsystem-1/ctrl1/state: No such file or directory
     Test complete

To reset the test target controller, the test case locates the "state"
file and writes to it. This works in most cases, but the write can
fail: the test case connects and disconnects the target controller's
subsystem in parallel, so the "state" file may disappear after being
located but before the write happens.

Suppress the write error message to avoid this sporadic failure.

Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
---
 tests/nvme/060 | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/tests/nvme/060 b/tests/nvme/060
index 997512d..9e694af 100755
--- a/tests/nvme/060
+++ b/tests/nvme/060
@@ -19,11 +19,16 @@ set_conditions() {
 	_set_nvme_trtype "$@"
 }
 
+filter_state_write_msg() {
+	grep --invert-match "state: No such file or directory"
+}
+
 nvmet_debug_trigger_reset() {
 	local nvmet_subsystem="$1"
 	local dfs_path="${NVMET_DFS}/${nvmet_subsystem}"
 
-	find "${dfs_path}" -maxdepth 1 -type d -name 'ctrl*' -exec sh -c 'echo "fatal" > "$1/state"' _ {} \;
+	find "${dfs_path}" -maxdepth 1 -type d -name 'ctrl*' -exec sh -c \
+	     'echo "fatal" > "$1/state" ' _ {} \; |& filter_state_write_msg
 }
 
 nvmet_reset_loop() {
-- 
2.54.0

Re: [PATCH blktests 1/2] nvme/060: suppress state file write error

[Daniel Wagner]

On Fri, Jun 19, 2026 at 10:33:28AM +0900, Shin'ichiro Kawasaki wrote:
> The test case nvme/060 sometimes fails with the message below:
> 
> nvme/060 (tr=rdma) (test nvme fabrics target reset)          [failed]
>     runtime    ...  27.967s
>     --- tests/nvme/060.out	2026-01-19 04:30:04.820982813 +0000
>     +++ /home/fedora/blktests/results/nodev_tr_rdma/nvme/060.out.bad	2026-01-19 06:44:49.223643735 +0000
>     @@ -1,2 +1,3 @@
>      Running nvme/060
>     +_: line 1: /sys/kernel/debug/nvmet//blktests-subsystem-1/ctrl1/state: No such file or directory
>      Test complete
> 
> To reset the test target controller, the test case locates the "state"
> file and writes to it. This works in most cases, but the write can
> fail: the test case connects and disconnects the target controller's
> subsystem in parallel, so the "state" file may disappear after being
> located but before the write happens.
> 
> Suppress the write error message to avoid this sporadic failure.
> 
> Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>

Thanks for fixing this up!

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

[PATCH blktests 2/2] nvme/060: suppress "nvme connect" failure for rdma transport

[Shin'ichiro Kawasaki]

The test case nvme/060 often fails for the rdma transport with the
message "FAIL: nvme connect return error code":

nvme/060 (tr=rdma) (test nvme fabrics target reset)          [failed]
    runtime  86.258s  ...  23.604s
    --- tests/nvme/060.out	2025-08-26 21:28:52.798847739 +0900
    +++ /home/shin/Blktests/blktests/results/nodev_tr_rdma/nvme/060.out.bad	2026-06-18 15:36:18.239424796 +0900
    @@ -1,2 +1,3 @@
     Running nvme/060
    +FAIL: nvme connect return error code
     Test complete

The test case repeats nvme connect and disconnect while controller
resets are repeated in parallel. In this situation, connect failures are
expected and should not alarm users. Suppress the failure message for
the rdma transport.

Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
---
 tests/nvme/060 | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/tests/nvme/060 b/tests/nvme/060
index 9e694af..49786af 100755
--- a/tests/nvme/060
+++ b/tests/nvme/060
@@ -19,6 +19,12 @@ set_conditions() {
 	_set_nvme_trtype "$@"
 }
 
+filter_connect_msg() {
+	if [[ ${nvme_trtype} == rdma ]]; then
+		grep --invert-match "nvme connect return error code"
+	fi
+}
+
 filter_state_write_msg() {
 	grep --invert-match "state: No such file or directory"
 }
@@ -56,7 +62,8 @@ test() {
 	# The target reset is triggered with an even number timeout, while the
 	# host reconnects with an odd number timeout.
 	for ((i = 0; i <= 5; i++)); do
-		_nvme_connect_subsys --keep-alive-tmo 1 --reconnect-delay 1 --no-wait-ns
+		_nvme_connect_subsys --keep-alive-tmo 1 --reconnect-delay 1 \
+				     --no-wait-ns | filter_connect_msg
 		sleep 3
 		_nvme_disconnect_subsys >> "$FULL" 2>&1
 	done
-- 
2.54.0

Re: [PATCH blktests 2/2] nvme/060: suppress "nvme connect" failure for rdma transport

[Daniel Wagner]

On Fri, Jun 19, 2026 at 10:33:29AM +0900, Shin'ichiro Kawasaki wrote:
> The test case nvme/060 often fails for the rdma transport with the
> message "FAIL: nvme connect return error code":
> 
> nvme/060 (tr=rdma) (test nvme fabrics target reset)          [failed]
>     runtime  86.258s  ...  23.604s
>     --- tests/nvme/060.out	2025-08-26 21:28:52.798847739 +0900
>     +++ /home/shin/Blktests/blktests/results/nodev_tr_rdma/nvme/060.out.bad	2026-06-18 15:36:18.239424796 +0900
>     @@ -1,2 +1,3 @@
>      Running nvme/060
>     +FAIL: nvme connect return error code
>      Test complete
> 
> The test case repeats nvme connect and disconnect while controller
> resets are repeated in parallel. In this situation, connect failures are
> expected and should not alarm users. Suppress the failure message for
> the rdma transport.
> 
> Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>

Looks good to me!

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

Re: [PATCH blktests 0/2] nvme/060: avoid false-positive failures

[Shin'ichiro Kawasaki]

On Jun 19, 2026 / 10:33, Shin'ichiro Kawasaki wrote:
> The test case nvme/060 fails often due to a couple of expected errors.
> One of them is observed for all transports, and the other is unique to
> the rdma transport. The two patches in this series suppress these two
> expected errors respectively.

FYI, I applied the patches.