[PATCH blktests 09/15] nvme: check for controller device in _nvme_connect_subsys()

[Daniel Wagner]

On Mon, Oct 28, 2024 at 06:39:59AM GMT, Shinichiro Kawasaki wrote:
> On Oct 25, 2024 / 16:40, Daniel Wagner wrote:
> > On Thu, Oct 24, 2024 at 10:00:19AM GMT, Shin'ichiro Kawasaki wrote:
> [...]
> > > @@ -408,21 +411,17 @@ _nvme_disconnect_subsys() {
> > >  }
> > >  
> > >  
> > > -_nvme_ns_ready() {
> > > -	local subsysnqn="${1}"
> > > -	local ns_path ns_id dev
> > > -	local cfs_path="${NVMET_CFS}/subsystems/$subsysnqn"
> > > +_nvme_ctrl_ready() {
> > > +	local ctrldev="${1}"
> > > +	local subsysnqn="${2:-$def_subsysnqn}"
> > > +	local ctrlpath="/sys/class/nvme/${ctrldev}"
> > >  
> > > -	dev=$(_find_nvme_dev "$subsysnqn")
> > > -	for ns_path in "${cfs_path}/namespaces/"*; do
> > > -		ns_id=${ns_path##*/}
> > > -		if [[ ! -b /dev/${dev}n${ns_id} ||
> > > -			   ! -e /sys/block/${dev}n${ns_id}/uuid ||
> > > -			   ! -e /sys/block/${dev}n${ns_id}/wwid ]]; then
> > > -			return 1
> > 
> > IIRC, the uuid/wwid test were originally from nvme/004 and copied to the
> > other test. When we refactored we moved the uuid/wwid here. I suggest we
> > add this part back to nvme/004 so that we don't loose this test case.
> 
> I checked the history the uuid/wwid test here, and it does not look originated
> from nvme/004:
> 
>  8251c724 ... Ofar moved it from tests/nvme/rc to common/nvme
>  8fc8eb14 ... Shin'ichiro turned the test into _nvme_ns_ready()
>  c766fccf ... Bart introduced the test
> 
> I guess the uuid/wwid test Daniel mentions is in the _check_uuid() helper
> function, and it has different story.

Ah indeed, I've mixed these two things up. In this case all is good.

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