* [ndctl PATCH] ndctl, test: add a new unit test for monitor
@ 2018-07-05 11:23 QI Fuli
2018-07-05 21:06 ` Masayoshi Mizuma
0 siblings, 1 reply; 4+ messages in thread
From: QI Fuli @ 2018-07-05 11:23 UTC (permalink / raw)
To: linux-nvdimm
Add a new unit test to test the follow options of the monitor command.
--dimm
--bus
--region
--namespace
--logfile
--config-file
Based-on-patch-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: QI Fuli <qi.fuli@jp.fujitsu.com>
---
test/Makefile.am | 3 +-
test/monitor.sh | 115 +++++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 117 insertions(+), 1 deletion(-)
create mode 100755 test/monitor.sh
diff --git a/test/Makefile.am b/test/Makefile.am
index cd451e9..8c76462 100644
--- a/test/Makefile.am
+++ b/test/Makefile.am
@@ -21,7 +21,8 @@ TESTS =\
btt-pad-compat.sh \
firmware-update.sh \
ack-shutdown-count-set \
- rescan-partitions.sh
+ rescan-partitions.sh \
+ monitor.sh
check_PROGRAMS =\
libndctl \
diff --git a/test/monitor.sh b/test/monitor.sh
new file mode 100755
index 0000000..4f113c8
--- /dev/null
+++ b/test/monitor.sh
@@ -0,0 +1,115 @@
+#!/bin/bash -Ex
+
+# SPDX-License-Identifier: GPL-2.0
+# Copyright(c) 2018, FUJITSU LIMITED. All rights reserved.
+
+rc=77
+logfile=""
+conf_file=""
+monitor_pid=65536
+FILTER_DIMM="nmem1"
+FILTER_BUS="ndbus1"
+FILTER_BUS_MAP_DIMM="nmem0:nmem1:nmem2:nmem3"
+FILTER_REGION="region1"
+FILTER_REGION_MAP_DIMM="nmem0:nmem1"
+FILTER_NAMESPACE="namespace1.0"
+FILTER_NAMESPACE_MAP_DIMM="nmem0:nmem1"
+CONF_FILE_SET_DIMM="nmem1:nmem3"
+
+. ./common
+
+trap 'err $LINENO' ERR
+
+check_min_kver "4.15" || do_skip "kernel $KVER may not support monitor service"
+
+start_monitor()
+{
+ logfile=$(mktemp)
+ $NDCTL monitor -l $logfile $1 &
+ monitor_pid=$!
+ truncate --size 0 $logfile #remove startup log
+}
+
+call_notify()
+{
+ ./smart-notify $NFIT_TEST_BUS0
+}
+
+check_result()
+{
+ json=$(cat $logfile)
+ dimms=$(jq ."dimm"."dev" <<<$json | sort | uniq | sed -e ':loop; N; $!b loop; s/\n/:/g' | sed 's/\"//g')
+ [[ $1 == $dimms ]]
+}
+
+stop_monitor()
+{
+ kill $monitor_pid
+ rm $logfile
+}
+
+create_conf_file()
+{
+ conf_file=$(mktemp)
+ echo 'dimm=nmem1 nmem3' > $conf_file
+ cat $conf_file
+}
+
+test_filter_dimm()
+{
+ start_monitor "-d $FILTER_DIMM"
+ call_notify
+ check_result "$FILTER_DIMM"
+ stop_monitor
+}
+
+test_filter_bus()
+{
+ start_monitor "-b $FILTER_BUS"
+ call_notify
+ check_result $FILTER_BUS_MAP_DIMM
+ stop_monitor
+}
+
+test_filter_region()
+{
+ start_monitor "-r $FILTER_REGION"
+ call_notify
+ check_result $FILTER_REGION_MAP_DIMM
+ stop_monitor
+}
+
+test_filter_namespace()
+{
+ $NDCTL create-namespace -r region1 -n $FILTER_NAMESPACE
+ start_monitor "-n $FILTER_NAMESPACE"
+ call_notify
+ check_result $FILTER_NAMESPACE_MAP_DIMM
+ stop_monitor
+ $NDCTL destroy-namespace $FILTER_NAMESPACE -f
+}
+
+test_conf_file()
+{
+ create_conf_file
+ start_monitor "-c $conf_file"
+ call_notify
+ check_result $CONF_FILE_SET_DIMM
+ stop_monitor
+ rm $conf_file
+}
+
+do_tests()
+{
+ test_filter_dimm
+ test_filter_bus
+ test_filter_region
+ test_filter_namespace
+ test_conf_file
+}
+
+modprobe nfit_test
+rc=1
+do_tests
+_cleanup
+exit 0
--
2.18.0
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [ndctl PATCH] ndctl, test: add a new unit test for monitor
2018-07-05 11:23 [ndctl PATCH] ndctl, test: add a new unit test for monitor QI Fuli
@ 2018-07-05 21:06 ` Masayoshi Mizuma
2018-07-05 21:25 ` Dan Williams
0 siblings, 1 reply; 4+ messages in thread
From: Masayoshi Mizuma @ 2018-07-05 21:06 UTC (permalink / raw)
To: qi.fuli, linux-nvdimm
Hi Qi,
I tried this test after I applied your latest monitor patches [1].
I found some points. Could you check the following?
[1] https://lists.01.org/pipermail/linux-nvdimm/2018-June/016562.html
On 07/05/2018 07:23 AM, QI Fuli wrote:
> Add a new unit test to test the follow options of the monitor command.
> --dimm
> --bus
> --region
> --namespace
> --logfile
> --config-file
>
> Based-on-patch-by: Yasunori Goto <y-goto@jp.fujitsu.com>
> Signed-off-by: QI Fuli <qi.fuli@jp.fujitsu.com>
> ---
> test/Makefile.am | 3 +-
> test/monitor.sh | 115 +++++++++++++++++++++++++++++++++++++++++++++++
> 2 files changed, 117 insertions(+), 1 deletion(-)
> create mode 100755 test/monitor.sh
>
> diff --git a/test/Makefile.am b/test/Makefile.am
> index cd451e9..8c76462 100644
> --- a/test/Makefile.am
> +++ b/test/Makefile.am
> @@ -21,7 +21,8 @@ TESTS =\
> btt-pad-compat.sh \
> firmware-update.sh \
> ack-shutdown-count-set \
> - rescan-partitions.sh
> + rescan-partitions.sh \
> + monitor.sh
>
> check_PROGRAMS =\
> libndctl \
> diff --git a/test/monitor.sh b/test/monitor.sh
> new file mode 100755
> index 0000000..4f113c8
> --- /dev/null
> +++ b/test/monitor.sh
> @@ -0,0 +1,115 @@
> +#!/bin/bash -Ex
> +
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright(c) 2018, FUJITSU LIMITED. All rights reserved.
> +
> +rc=77
> +logfile=""
> +conf_file=""
> +monitor_pid=65536
> +FILTER_DIMM="nmem1"
> +FILTER_BUS="ndbus1"
Is this ndbus0?
because the dimms under nfit_test.1 are not supported smart.
# ./test/smart-notify nfit_test.1
nmem5: no smart command support
nmem4: no smart command support
libdaxctl: daxctl_unref: context 0x20d41e0 released
libndctl: ndctl_unref: context 0x20d4310 released
#
> +FILTER_BUS_MAP_DIMM="nmem0:nmem1:nmem2:nmem3"
How about getting the dimms by ndctl list command instead of hard-cording?
> +FILTER_REGION="region1"
> +FILTER_REGION_MAP_DIMM="nmem0:nmem1"
Is this "nmem0:nmem1:nmem2:nmem3"?
Or why don't you getting the dimms by ndctl list command instead of hard-cording?
# ./ndctl/ndctl list -r region1 -D
[
{
"dev":"nmem1",
"id":"cdab-0a-07e0-feffffff",
"handle":1,
"phys_id":1
},
{
"dev":"nmem3",
"id":"cdab-0a-07e0-fefeffff",
"handle":257,
"phys_id":3
},
{
"dev":"nmem0",
"id":"cdab-0a-07e0-ffffffff",
"handle":0,
"phys_id":0
},
{
"dev":"nmem2",
"id":"cdab-0a-07e0-fffeffff",
"handle":256,
"phys_id":2
}
]
#
> +FILTER_NAMESPACE="namespace1.0"
> +FILTER_NAMESPACE_MAP_DIMM="nmem0:nmem1"
Same as above.
> +CONF_FILE_SET_DIMM="nmem1:nmem3"
> +
> +. ./common
> +
> +trap 'err $LINENO' ERR
> +
> +check_min_kver "4.15" || do_skip "kernel $KVER may not support monitor service"
> +
> +start_monitor()
> +{
> + logfile=$(mktemp)
> + $NDCTL monitor -l $logfile $1 &
> + monitor_pid=$!
> + truncate --size 0 $logfile #remove startup log
> +}
> +
> +call_notify()
> +{
> + ./smart-notify $NFIT_TEST_BUS0
I think monitor daemon needs some time for writing out the log.
So why don't we have some sleep time? Like as:
sync; sleep 3
> +}
> +
> +check_result()
> +{
> + json=$(cat $logfile)
> + dimms=$(jq ."dimm"."dev" <<<$json | sort | uniq | sed -e ':loop; N; $!b loop; s/\n/:/g' | sed 's/\"//g')
> + [[ $1 == $dimms ]]
> +}
> +
> +stop_monitor()
> +{
> + kill $monitor_pid
> + rm $logfile
> +}
> +
> +create_conf_file()
> +{
> + conf_file=$(mktemp)
> + echo 'dimm=nmem1 nmem3' > $conf_file
> + cat $conf_file
> +}
> +
> +test_filter_dimm()
> +{
> + start_monitor "-d $FILTER_DIMM"
> + call_notify
> + check_result "$FILTER_DIMM"
> + stop_monitor
> +}
> +
> +test_filter_bus()
> +{
> + start_monitor "-b $FILTER_BUS"
> + call_notify
> + check_result $FILTER_BUS_MAP_DIMM
> + stop_monitor
> +}
> +
> +test_filter_region()
> +{
> + start_monitor "-r $FILTER_REGION"
> + call_notify
> + check_result $FILTER_REGION_MAP_DIMM
> + stop_monitor
> +}
> +
> +test_filter_namespace()
> +{
> + $NDCTL create-namespace -r region1 -n $FILTER_NAMESPACE
> + start_monitor "-n $FILTER_NAMESPACE"
> + call_notify
> + check_result $FILTER_NAMESPACE_MAP_DIMM
> + stop_monitor
> + $NDCTL destroy-namespace $FILTER_NAMESPACE -f
> +}
> +
> +test_conf_file()
> +{
> + create_conf_file
> + start_monitor "-c $conf_file"
> + call_notify
> + check_result $CONF_FILE_SET_DIMM
> + stop_monitor
> + rm $conf_file
> +}
> +
> +do_tests()
> +{
> + test_filter_dimm
> + test_filter_bus
> + test_filter_region
> + test_filter_namespace
> + test_conf_file
> +}
> +
> +modprobe nfit_test
> +rc=1
Why don't you add initialize the region before testing?
like as follows.
init()
{
$NDCTL disable-region -b $NFIT_TEST_BUS0 all
$NDCTL zero-labels -b $NFIT_TEST_BUS0 all
$NDCTL enable-region -b $NFIT_TEST_BUS0 all
}
init
Thanks,
Masa
> +do_tests
> +_cleanup
> +exit 0
>
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [ndctl PATCH] ndctl, test: add a new unit test for monitor
2018-07-05 21:06 ` Masayoshi Mizuma
@ 2018-07-05 21:25 ` Dan Williams
2018-07-06 4:10 ` Qi, Fuli
0 siblings, 1 reply; 4+ messages in thread
From: Dan Williams @ 2018-07-05 21:25 UTC (permalink / raw)
To: Masayoshi Mizuma; +Cc: linux-nvdimm
On Thu, Jul 5, 2018 at 2:06 PM, Masayoshi Mizuma <msys.mizuma@gmail.com> wrote:
> Hi Qi,
>
> I tried this test after I applied your latest monitor patches [1].
> I found some points. Could you check the following?
>
> [1] https://lists.01.org/pipermail/linux-nvdimm/2018-June/016562.html
>
> On 07/05/2018 07:23 AM, QI Fuli wrote:
[..]
> Is this "nmem0:nmem1:nmem2:nmem3"?
> Or why don't you getting the dimms by ndctl list command instead of hard-cording?
Yes, this is needed. The test fails for me because I have an e820
defined bus and a typical ACPI NFIT bus along with the nfit_test buses
on my test system.
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm
^ permalink raw reply [flat|nested] 4+ messages in thread
* RE: [ndctl PATCH] ndctl, test: add a new unit test for monitor
2018-07-05 21:25 ` Dan Williams
@ 2018-07-06 4:10 ` Qi, Fuli
0 siblings, 0 replies; 4+ messages in thread
From: Qi, Fuli @ 2018-07-06 4:10 UTC (permalink / raw)
To: 'Dan Williams', Masayoshi Mizuma; +Cc: linux-nvdimm
> -----Original Message-----
> From: Dan Williams [mailto:dan.j.williams@intel.com]
> Sent: Friday, July 6, 2018 6:26 AM
> To: Masayoshi Mizuma <msys.mizuma@gmail.com>
> Cc: Qi, Fuli/斉 福利 <qi.fuli@jp.fujitsu.com>; linux-nvdimm
> <linux-nvdimm@lists.01.org>
> Subject: Re: [ndctl PATCH] ndctl, test: add a new unit test for monitor
>
> On Thu, Jul 5, 2018 at 2:06 PM, Masayoshi Mizuma <msys.mizuma@gmail.com> wrote:
> > Hi Qi,
> >
> > I tried this test after I applied your latest monitor patches [1].
> > I found some points. Could you check the following?
> >
> > [1] https://lists.01.org/pipermail/linux-nvdimm/2018-June/016562.html
> >
> > On 07/05/2018 07:23 AM, QI Fuli wrote:
> [..]
> > Is this "nmem0:nmem1:nmem2:nmem3"?
> > Or why don't you getting the dimms by ndctl list command instead of hard-cording?
>
> Yes, this is needed. The test fails for me because I have an e820 defined bus and
> a typical ACPI NFIT bus along with the nfit_test buses on my test system.
Masa, Dan,
Thank you very much for your comments.
I will fix it in next version.
QI
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2018-07-06 4:10 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2018-07-05 11:23 [ndctl PATCH] ndctl, test: add a new unit test for monitor QI Fuli
2018-07-05 21:06 ` Masayoshi Mizuma
2018-07-05 21:25 ` Dan Williams
2018-07-06 4:10 ` Qi, Fuli
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox