On Wed, May 14, 2025 at 03:14:07PM -0700, Marc Herbert wrote:
> On 2025-05-13 18:41, alison.schofi...@intel.com wrote:
> > From: Alison Schofield <alison.schofi...@intel.com>
> >
> > monitor.sh runs for 50 seconds and spends 48 of those seconds sleeping.
> > Removing the sleeps entirely has no effect on the test in this users
> > environment. It passes and produces the same test log.
> >
> > Experiments replacing sleeps with polling for monitor ready and log file
> > updates proved that both are always available following the sync so there
> > is no need to replace the sleeps with a more precise or reliable polling
> > method. Simply remove the sleeps. Run time is now < 3s.
> >
> > I'd especially like to get Tested-by tags on this one to confirm that my
> > environment isn't special and that this succeeds elsewhere.
>
> In my configuration, this patch makes the test fail 100% of the time.
> It passes 100% of the time without this patch.
Thanks for testing!
Let me resurrect the polling version and I'll pass it by you to try out.
--Alison
>
> Also, it leaves a "ndctl monitor" behind which makes meson hang
> forever, believing the test is never done.
>
> Tested on top of current "origin/pending" 1850ddcbcbf9
>
> Marc
>
> nmem6: supported alarms: 0x7
> nmem6: set spare threshold: 99
> libndctl: do_cmd: bus: 2 dimm: 0x100 cmd: cmd_call:smart_set_thresh status: 0
> fw: 0 (success)
> nmem6: set mtemp threshold: 11.50
> libndctl: do_cmd: bus: 2 dimm: 0x100 cmd: cmd_call:smart_set_thresh status: 0
> fw: 0 (success)
> nmem6: set ctemp threshold: 12.50
> libndctl: do_cmd: bus: 2 dimm: 0x100 cmd: cmd_call:smart_set_thresh status: 0
> fw: 0 (success)
> nmem6: set thresholds back to defaults
> libndctl: do_cmd: bus: 2 dimm: 0x100 cmd: cmd_call:smart_set_thresh status: 0
> fw: 0 (success)
> libdaxctl: daxctl_unref: context 0x56469566a9d0 released
> libndctl: ndctl_unref: context 0x56469566b050 released
> + sync
> + check_result 'nmem4 nmem5 nmem6 nmem7'
> ++ cat /tmp/tmp.mYY263Hk3f
> + jlog=
> ++ jq .dimm.dev
> ++ sort
> ++ uniq
> ++ xargs
> + notify_dimms=
> + [[ nmem4 nmem5 nmem6 nmem7 == '' ]]
> ++ err 65
> +++ basename /root/CXL/ndctl/test/monitor.sh
> ++ echo test/monitor.sh: failed at line 65
> test/monitor.sh: failed at line 65
> ++ '[' -n '' ']'
> ++ exit 1
>
>
>
> > Signed-off-by: Alison Schofield <alison.schofi...@intel.com>
> > ---
> > test/monitor.sh | 6 +++---
> > 1 file changed, 3 insertions(+), 3 deletions(-)
> >
> > diff --git a/test/monitor.sh b/test/monitor.sh
> > index be8e24d6f3aa..88e253e5df00 100755
> > --- a/test/monitor.sh
> > +++ b/test/monitor.sh
> > @@ -26,7 +26,7 @@ start_monitor()
> > logfile=$(mktemp)
> > $NDCTL monitor -c "$monitor_conf" -l "$logfile" $1 &
> > monitor_pid=$!
> > - sync; sleep 3
> > + sync
> > truncate --size 0 "$logfile" #remove startup log
> > }
> >
> > @@ -49,13 +49,13 @@ get_monitor_dimm()
> > call_notify()
> > {
> > "$TEST_PATH"/smart-notify "$smart_supported_bus"
> > - sync; sleep 3
> > + sync
> > }
> >
> > inject_smart()
> > {
> > $NDCTL inject-smart "$monitor_dimms" $1
> > - sync; sleep 3
> > + sync
> > }
> >
> > check_result()
>
