On 6/27/24 11:10, Daniel Wagner wrote:
Most of the NVMEeoF tests are exercising the host code of the nvme subsystem. There is no real reason not to run these against a real target. We just have to skip the soft target setup and make it possible to setup a remote target. Because all tests use now the common setup/cleanup helpers we just need to intercept this call and forward it to an external component. As we already have various nvme variables to setup the target which we should allow to overwrite. Also introduce a NVME_TARGET_CONTROL variable which points to a script which gets executed whenever a targets needs to be created/destroyed. Signed-off-by: Daniel Wagner <dwagner@xxxxxxx> --- Documentation/running-tests.md | 33 ++++++++++++++++++++ check | 4 +++ tests/nvme/rc | 57 ++++++++++++++++++++++++++++++++-- 3 files changed, 92 insertions(+), 2 deletions(-) diff --git a/Documentation/running-tests.md b/Documentation/running-tests.md index 968702e76bb5..fe4f729bd037 100644 --- a/Documentation/running-tests.md +++ b/Documentation/running-tests.md @@ -120,6 +120,9 @@ The NVMe tests can be additionally parameterized via environment variables. - NVME_NUM_ITER: 1000 (default) The number of iterations a test should do. This parameter had an old name 'nvme_num_iter'. The old name is still usable, but not recommended. +- NVME_TARGET_CONTROL: When defined, the generic target setup/cleanup code will + be skipped and this script gets called. This makes it possible to run + the fabric nvme tests against a real target.### Running nvme-rdma and SRP tests @@ -167,3 +170,33 @@ if ! findmnt -t configfs /sys/kernel/config > /dev/null; thenmount -t configfs configfs /sys/kernel/config fi ``` +### NVME_TARGET_CONTROL + +When NVME_TARGET_CONTROL is set, blktests will call the script which the +environment variable points to, to fetch the configuration values to be used for +the runs, e.g subsysnqn or hostnqn. This allows the blktest to be run against +external configured/setup targets. + +The blktests expects that the script interface implements following +commands: + +config: + --show-blkdev-type + --show-trtype + --show-hostnqn + --show-hostid + --show-host-traddr + --show-traddr + --show-trsvid + --show-subsys-uuid + --show-subsysnqn + +setup: + --subsysnqn SUBSYSNQN + --subsys-uuid SUBSYS_UUID + --hostnqn HOSTNQN + --ctrlkey CTRLKEY + --hostkey HOSTKEY + +cleanup: + --subsysnqn SUBSYSNQN diff --git a/check b/check index 3ed4510f3f40..d0475629773d 100755 --- a/check +++ b/check @@ -603,6 +603,10 @@ _run_group() { # shellcheck disable=SC1090 . "tests/${group}/rc"+ if declare -fF group_setup >/dev/null; then+ group_setup + fi + if declare -fF group_requires >/dev/null; then group_requires if [[ -v SKIP_REASONS ]]; then diff --git a/tests/nvme/rc b/tests/nvme/rc index c1ddf412033b..4465dea0370b 100644 --- a/tests/nvme/rc +++ b/tests/nvme/rc @@ -23,6 +23,7 @@ _check_conflict_and_set_default NVME_IMG_SIZE nvme_img_size 1G _check_conflict_and_set_default NVME_NUM_ITER nvme_num_iter 1000 nvmet_blkdev_type=${nvmet_blkdev_type:-"device"} NVMET_BLKDEV_TYPES=${NVMET_BLKDEV_TYPES:-"device file"} +nvme_target_control="${NVME_TARGET_CONTROL:-}"_NVMET_TRTYPES_is_valid() {local type @@ -135,6 +136,13 @@ _nvme_requires() { return 0 }+group_setup() {+ if [[ -n "${nvme_target_control}" ]]; then + NVMET_TRTYPES="$(${nvme_target_control} config --show-trtype)" + NVMET_BLKDEV_TYPES="$(${nvme_target_control} config --show-blkdev-type)" + fi +} + group_requires() { _have_root _NVMET_TRTYPES_is_valid @@ -359,6 +367,10 @@ _cleanup_nvmet() { fi done+ if [[ -n "${nvme_target_control}" ]]; then+ return + fi + for port in "${NVMET_CFS}"/ports/*; do name=$(basename "${port}") echo "WARNING: Test did not clean up port: ${name}" @@ -403,11 +415,26 @@ _cleanup_nvmet() {_setup_nvmet() {_register_test_cleanup _cleanup_nvmet + + if [[ -n "${nvme_target_control}" ]]; then + def_hostnqn="$(${nvme_target_control} config --show-hostnqn)" + def_hostid="$(${nvme_target_control} config --show-hostid)" + def_host_traddr="$(${nvme_target_control} config --show-host-traddr)" + def_traddr="$(${nvme_target_control} config --show-traddr)" + def_trsvcid="$(${nvme_target_control} config --show-trsvid)" + def_subsys_uuid="$(${nvme_target_control} config --show-subsys-uuid)" + def_subsysnqn="$(${nvme_target_control} config --show-subsysnqn)" + return + fi + modprobe -q nvmet + if [[ "${nvme_trtype}" != "loop" ]]; then modprobe -q nvmet-"${nvme_trtype}" fi + modprobe -q nvme-"${nvme_trtype}" + if [[ "${nvme_trtype}" == "rdma" ]]; then start_soft_rdma for i in $(rdma_network_interfaces) @@ -425,6 +452,7 @@ _setup_nvmet() { fi done fi + if [[ "${nvme_trtype}" = "fc" ]]; then modprobe -q nvme-fcloop _setup_fcloop "${def_local_wwnn}" "${def_local_wwpn}" \ @@ -873,11 +901,13 @@ _find_nvme_passthru_loop_dev() {_nvmet_target_setup() {local blkdev_type="${nvmet_blkdev_type}" + local subsys_uuid="${def_subsys_uuid}" + local subsysnqn="${def_subsysnqn}" local blkdev + local ARGS=() local ctrlkey="" local hostkey="" - local subsysnqn="${def_subsysnqn}" - local subsys_uuid="${def_subsys_uuid}" + local blkdev local portwhile [[ $# -gt 0 ]]; do@@ -909,6 +939,22 @@ _nvmet_target_setup() { esac done+ if [[ -n "${hostkey}" ]]; then+ ARGS+=(--hostkey "${hostkey}") + fi + if [[ -n "${ctrlkey}" ]]; then + ARGS+=(--ctrkey "${ctrlkey}") + fi + + if [[ -n "${nvme_target_control}" ]]; then + eval "${nvme_target_control}" setup \ + --subsysnqn "${subsysnqn}" \ + --subsys-uuid "${subsys_uuid}" \ + --hostnqn "${def_hostnqn}" \ + "${ARGS[@]}" &> /dev/null + return + fi + truncate -s "${NVME_IMG_SIZE}" "$(_nvme_def_file_path)" if [[ "${blkdev_type}" == "device" ]]; then blkdev="$(losetup -f --show "$(_nvme_def_file_path)")" @@ -948,6 +994,13 @@ _nvmet_target_cleanup() { esac done+ if [[ -n "${nvme_target_control}" ]]; then+ eval "${nvme_target_control}" cleanup \ + --subsysnqn "${subsysnqn}" \ + > /dev/null + return + fi + _get_nvmet_ports "${subsysnqn}" portsfor port in "${ports[@]}"; do
Hmm. This wasn't quite what I had in mind; I think it'd be simpler if we could just check if the requested controller is visible to the host already (ie checking sysfs here), and then skip all the setup steps at they are obviously not required anymore. That would save quite a lot of issues, and we wouldn't need to specify a target setup script (or whatever). Quite a bit of churn, I agree, as that would mean we have to roll _setup_nvmet _nvmet_target_setup _nvme_connect_subsys all into one function. But it might be easier in the long run. Hmm? Cheers, Hannes -- Dr. Hannes Reinecke Kernel Storage Architect hare@xxxxxxx +49 911 74053 688 SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich
