Skip to contentSkip to conversation

memcg_stress newlib porting

PATCH New Cristian Marussi 3 participants 11 messages Jan 4, 2019 – Feb 15, 2019

lore
Share & tools
lore
#0 lore raw

v3 cover memcg_stress newlib porting

This converts memcg stress tests and its helper to LTP newlib and performs a general cleanup.

This patch is dependent on [1] since it factors out a common helper to cgroup_lib.sh (introduced in [1])

Changes
-------
v2->v3:
- cleaned up cgroup_lib.sh helper

v1->v2:
- Restored the original usage of SIGKILL to terminate memcp_process_stress helpers after having a look at what said in previous commit:

	9b9a6bb10258bc975bd1707b418699264d3bdcef

Cristian Marussi (1):
[LTP] memcg_stress_test.sh: ported to newlib

[1]: http://lists.linux.it/pipermail/ltp/2018-December/010385.html

#1 lore raw

PATCH v3 1/1 memcg_stress_test.sh: ported to newlib

Ported to newlib framework and general cleanup.
Moved an helper function out into cgroup_lib.sh.

---
 testcases/kernel/controllers/cgroup_lib.sh    |  16 +++
 .../kernel/controllers/memcg/stress/Makefile  |  21 +--
 .../memcg/stress/memcg_process_stress.c       |  34 ++---
 .../memcg/stress/memcg_stress_test.sh         | 135 ++++++++----------
 4 files changed, 92 insertions(+), 114 deletions(-)
testcases/kernel/controllers/cgroup_lib.sh +16 -0
index 8c19c6c3e..dbf5199e6 100644
@@ -21,3 +21,19 @@ get_cgroup_mountpoint()
 	echo $mntpoint
 	return 0
 }
+
+# Check if given subsystem is supported and enabled
+# is_cgroup_subsystem_available_and_enabled SUBSYSTEM
+# RETURN: 0 if subsystem supported and enabled, otherwise 1
+is_cgroup_subsystem_available_and_enabled()
+{
+	local val
+	local subsystem=$1
+
+	[ $# -eq 0 ] && tst_brk TBROK "is_cgroup_subsystem_available_and_enabled: subsystem not defined"
+
+	val=$(grep -w $subsystem /proc/cgroups | awk '{ print $4 }')
+	[ "x$val" = "x1" ] && return 0
+
+	return 1
+}
testcases/kernel/controllers/memcg/stress/Makefile +3 -18
index a0456ac5d..773363cfc 100644
@@ -1,24 +1,9 @@
+# SPDX-License-Identifier: GPL-2.0-or-later
+# Copyright (C) 2009, Cisco Systems Inc.
+# Author: Ngie Cooper, September 2009
 #
 #    kernel/controllers/memcg/stress testcase suite Makefile.
 #
-#    Copyright (C) 2009, Cisco Systems Inc.
-#
-#    This program is free software; you can redistribute it and/or modify
-#    it under the terms of the GNU General Public License as published by
-#    the Free Software Foundation; either version 2 of the License, or
-#    (at your option) any later version.
-#
-#    This program is distributed in the hope that it will be useful,
-#    but WITHOUT ANY WARRANTY; without even the implied warranty of
-#    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-#    GNU General Public License for more details.
-#
-#    You should have received a copy of the GNU General Public License along
-#    with this program; if not, write to the Free Software Foundation, Inc.,
-#    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-#
-# Ngie Cooper, September 2009
-#
 
 top_srcdir		?= ../../../../..
 
testcases/kernel/controllers/memcg/stress/memcg_process_stress.c +10 -24
index 870575c26..6af550012 100644
@@ -1,24 +1,9 @@
-/******************************************************************************/
-/*                                                                            */
-/* Copyright (c) 2009 FUJITSU LIMITED                                         */
-/*                                                                            */
-/* This program is free software;  you can redistribute it and/or modify      */
-/* it under the terms of the GNU General Public License as published by       */
-/* the Free Software Foundation; either version 2 of the License, or          */
-/* (at your option) any later version.                                        */
-/*                                                                            */
-/* This program is distributed in the hope that it will be useful,            */
-/* but WITHOUT ANY WARRANTY;  without even the implied warranty of            */
-/* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See                  */
-/* the GNU General Public License for more details.                           */
-/*                                                                            */
-/* You should have received a copy of the GNU General Public License          */
-/* along with this program;  if not, write to the Free Software               */
-/* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA    */
-/*                                                                            */
-/* Author: Li Zefan <[email protected]>                                     */
-/*                                                                            */
-/******************************************************************************/
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (c) 2009 FUJITSU LIMITED
+ *
+ * Author: Li Zefan <[email protected]>
+ */
 
 #include <sys/mman.h>
 #include <err.h>
@@ -91,14 +76,15 @@ int main(int argc, char *argv[])
 	if (interval <= 0)
 		interval = 1;
 
-	/* TODO (garrcoop): add error handling. */
 	memset(&sigint_action, 0, sizeof(sigint_action));
 	sigint_action.sa_handler = &sigint_handler;
-	sigaction(SIGINT, &sigint_action, NULL);
+	if (sigaction(SIGINT, &sigint_action, NULL))
+		err(1, "sigaction(%s) failed", "SIGINT");
 
 	memset(&sigusr_action, 0, sizeof(sigusr_action));
 	sigusr_action.sa_handler = &sigusr_handler;
-	sigaction(SIGUSR1, &sigusr_action, NULL);
+	if (sigaction(SIGUSR1, &sigusr_action, NULL))
+		err(1, "sigaction(%s) failed", "SIGUSR1");
 
 	while (!flag_exit) {
 		sleep(interval);
testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +63 -72

181 lines not on this page · raw message

Signature · 1 line
2.17.1
#2 lore raw

PATCH v3 1/1 memcg_stress_test.sh: ported to newlib

Hi Cristian,

thanks for your patch. I was thinking just to fix minor formatting issues, but eval in run_stress() deserve to be rewritten.

Quoted 2 lines

Ported to newlib framework and general cleanup.
Moved an helper function out into cgroup_lib.sh.

..

testcases/kernel/controllers/cgroup_lib.sh get_cgroup_mountpoint() · line 36

Quoted 1 line
testcases/kernel/controllers/cgroup_lib.sh +1 -0
+	[ "x$val" = "x1" ] && return 0

drop x, it's portable enough like this::
[ "$val" = "1" ] && return 0

...

testcases/kernel/controllers/memcg/stress/Makefile line 5

Quoted 7 lines

+++ b/testcases/kernel/controllers/memcg/stress/Makefile

testcases/kernel/controllers/memcg/stress/Makefile +3 -0
@@ -1,24 +1,9 @@
+# SPDX-License-Identifier: GPL-2.0-or-later
+# Copyright (C) 2009, Cisco Systems Inc.
+# Author: Ngie Cooper, September 2009
 
 #    kernel/controllers/memcg/stress testcase suite Makefile.

Delete this as well.

Quoted 1 line
(quoted diff) +1 -0
+++ b/testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh

...

testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh line 28

Quoted 3 lines
testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +3 -0
+	if ! is_cgroup_subsystem_available_and_enabled "memory";then
+		tst_res TWARN "Either Kernel does not support MEMORY resource controller or feature not enabled"
+		tst_brk TCONF ignored "Skipping all memory cgroup testcases...."

TWARN followed by TCONF does not make sense. Also skip is redundant. Please, avoid using dots (even in comments).
This is enough:

if ! is_cgroup_subsystem_available_and_enabled "memory"; then
tst_res TCONF "Either Kernel does not support MEMORY resource controller or feature not enabled"
fi

...

testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh line 34

Quoted 5 lines
testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +5 -0
+
+	echo 3 > /proc/sys/vm/drop_caches
+	sleep 2
+	local mem_free=`cat /proc/meminfo | grep MemFree | awk '{ print $2 }'`
+	local swap_free=`cat /proc/meminfo | grep SwapFree | awk '{ print $2 }'`

Test started using awk, which might not be everywhere (even it's quite basic. TST_NEEDS_CMDS="awk" would be good for tests.

And also TST_NEEDS_CMDS="$TST_NEEDS_CMDS awk" would be good for cgroup_lib.sh.

Quoted 24 lines
testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +14 -1
+
+	MEM=$(( $mem_free + $swap_free / 2 ))
+	MEM=$(( $MEM / 1024 ))
+	RUN_TIME=$(( 15 * 60 ))
+
+	tst_res TINFO "Calculated available memory $MEM MB"
 }
 
+do_cleanup()
+{
+	if [ -e /dev/memcg ]; then
+		umount /dev/memcg 2> /dev/null
+		rmdir /dev/memcg 2> /dev/null
+	fi
+}
 
 do_mount()
 {
-	cleanup;
+	do_cleanup
 
 	mkdir /dev/memcg 2> /dev/null
 	mount -t cgroup -omemory memcg /dev/memcg

testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh do_mount() · line 73

Quoted 10 lines
testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +3 -1
@@ -65,62 +65,53 @@ do_mount()
 # $4 - How long does this test run ? in second
 run_stress()
 {
-	do_mount;
+	local i
+
+	do_mount
 
 	for i in $(seq 0 $(($1-1)))
 	do

Please keep do on the same line (in memcg_stress_test.sh on several times):
for i in $(seq 0 $(($1-1))); do

testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh do_mount() · line 74

Quoted 3 lines
testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +0 -2
 		mkdir /dev/memcg/$i 2> /dev/null
-		./memcg_process_stress $2 $3 &
-		eval pid$i=$!

Eval is evil. And eval with local is even more evil. Even if it works on dash, using eval in whole function is just bad. Could you please rewrite it into single variable? You can just separate pids with space.
+ maybe describe variables instead of using positional variables, e.g.:
local cgroups="$1"
local mem="$2"
...

Quoted 18 lines
testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +4 -2
+		memcg_process_stress $2 $3 &
+		eval local pid$i=$!
 
 		eval echo \$pid$i > /dev/memcg/$i/tasks
 	done
 
 	for i in $(seq 0 $(($1-1)))
 	do
-		eval /bin/kill -s SIGUSR1 \$pid$i 2> /dev/null
+		eval kill -USR1 \$pid$i 2> /dev/null
 	done
 
 	sleep $4
 
 	for i in $(seq 0 $(($1-1)))
 	do
-		eval /bin/kill -s SIGKILL \$pid$i 2> /dev/null
+		eval kill -KILL \$pid$i 2> /dev/null

testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh do_mount() · line 99

Quoted 10 lines
testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +1 -1
 		eval wait \$pid$i
 
 		rmdir /dev/memcg/$i 2> /dev/null
 	done
 
-	cleanup;
+	do_cleanup
 }
 
 testcase_1()

I'd rename it to test1.

testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh do_mount() · line 101

Quoted 3 lines
testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +1 -1
 {
-	run_stress 150 $(( ($mem-150) / 150 )) 5 $RUN_TIME
+	tst_res TINFO "testcase 1 started...it will run for $RUN_TIME secs"

Can you avoid dots, will?
Maybe describe it a bit.
I'd print test info in run_stress(), something like this:
tst_res TINFO "Testing $cgroups cgroups, using $mem, run for $RUN_TIME secs"

Kind regards,
Petr

#3 lore raw

PATCH v3 1/1 memcg_stress_test.sh: ported to newlib

Hi Cristian,

I suggest following cleanup of your patch (see bellow).

This change, with my other cleanup patch [1] + your patch [2] can be seen on [3].

Kind regards,
Petr

[1] https://patchwork.ozlabs.org/patch/1032923/
[2] https://patchwork.ozlabs.org/patch/1020887/
[3] https://github.com/pevik/ltp/commits/christian/memcg_stress_test.sh.v3.fixes

kernel/controllers/cgroup_lib.sh +1 -1
index c164932fa..7918b5636 100644
@@ -33,7 +33,7 @@ is_cgroup_subsystem_available_and_enabled()
 	[ $# -eq 0 ] && tst_brk TBROK "is_cgroup_subsystem_available_and_enabled: subsystem not defined"
 
 	val=$(grep -w $subsystem /proc/cgroups | awk '{ print $4 }')
-	[ "x$val" = "x1" ] && return 0
+	[ "$val" = "1" ] && return 0
 
 	return 1
 }
kernel/controllers/memcg/stress/Makefile +0 -3
index 773363cfc..a9678bf3b 100644
@@ -1,9 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0-or-later
 # Copyright (C) 2009, Cisco Systems Inc.
 # Author: Ngie Cooper, September 2009
-#
-#    kernel/controllers/memcg/stress testcase suite Makefile.
-#
 
 top_srcdir		?= ../../../../..
 
kernel/controllers/memcg/stress/memcg_process_stress.c +0 -1
index 6af550012..422deaeee 100644
@@ -1,7 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 /*
  * Copyright (c) 2009 FUJITSU LIMITED
- *
  * Author: Li Zefan <[email protected]>
  */
 
kernel/controllers/memcg/stress/memcg_stress_test.sh +17 -25

105 lines not on this page · raw message

#4 lore raw

PATCH v3 1/1 memcg_stress_test.sh: ported to newlib

Hi

On 29/01/2019 17:50, Petr Vorel wrote:> Hi Cristian,

Quoted 7 lines

I suggest following cleanup of your patch (see bellow).

This change, with my other cleanup patch [1] + your patch [2] can be seen on [3].

Kind regards,
Petr

Thanks for the review and the cleanup. (next I was going to rework following your today's advises, but you've been faster...)

I'll test asap.

Thanks

Regards

Cristian

________________________________
From: Petr Vorel <[email protected]>
Sent: 29 January 2019 17:50:19
To: Cristian Marussi
Cc: [email protected]
Subject: Re: [LTP] [PATCH v3] memcg_stress_test.sh: ported to newlib

Hi Cristian,

I suggest following cleanup of your patch (see bellow).

This change, with my other cleanup patch [1] + your patch [2] can be seen on [3].

Kind regards,
Petr

[1] https://patchwork.ozlabs.org/patch/1032923/
[2] https://patchwork.ozlabs.org/patch/1020887/
[3] https://github.com/pevik/ltp/commits/christian/memcg_stress_test.sh.v3.fixes

kernel/controllers/cgroup_lib.sh +1 -1
index c164932fa..7918b5636 100644
@@ -33,7 +33,7 @@ is_cgroup_subsystem_available_and_enabled()
         [ $# -eq 0 ] && tst_brk TBROK "is_cgroup_subsystem_available_and_enabled: subsystem not defined"
 
         val=$(grep -w $subsystem /proc/cgroups | awk '{ print $4 }')
-       [ "x$val" = "x1" ] && return 0
+       [ "$val" = "1" ] && return 0
 
         return 1
 }
kernel/controllers/memcg/stress/Makefile +0 -3
index 773363cfc..a9678bf3b 100644
@@ -1,9 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0-or-later
 # Copyright (C) 2009, Cisco Systems Inc.
 # Author: Ngie Cooper, September 2009
-#
-#    kernel/controllers/memcg/stress testcase suite Makefile.
-#
 
 top_srcdir              ?= ../../../../..
 
kernel/controllers/memcg/stress/memcg_process_stress.c +0 -1
index 6af550012..422deaeee 100644
@@ -1,7 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 /*
  * Copyright (c) 2009 FUJITSU LIMITED
- *
  * Author: Li Zefan <[email protected]>
  */
 
kernel/controllers/memcg/stress/memcg_stress_test.sh +17 -25

105 lines not on this page · raw message

IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you. -------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20190129/95a19364/attachment.html>

#5 lore raw

PATCH v3 1/1 memcg_stress_test.sh: ported to newlib

Hi Cristian,

Quoted 1 line

Thanks for the review and the cleanup. (next I was going to rework following your today's advises, but you've been faster...)

I'm sorry, I'll be patient next time :).
I was thinking to add these changes to your commit and merge patchset, but as I introduced 2 regressions in your patchset last time (I test it of course), I was careful.

Quoted 1 line

I'll test asap.

Thanks a lot!

Kind regards,
Petr

#6 lore raw

PATCH v3 1/1 memcg_stress_test.sh: ported to newlib

Hi

Quoted 10 lines · Petr Vorel

Hi Cristian,

Quoted 1 line

Thanks for the review and the cleanup. (next I was going to rework following your today's advises, but you've been faster...)

I'm sorry, I'll be patient next time :).
I was thinking to add these changes to your commit and merge patchset, but as I introduced 2 regressions in your patchset last time (I test it of course), I was careful.

Quoted 1 line

I'll test asap.

Thanks a lot!

Tested from you branch with ToT as:

10df48aa7 (HEAD -> pevik_ltp/christian/memcg_stress_test.sh.v3.fixes) memcg_stress_test.sh: fix memory usage
ebd6efaa6 memcg_stress_test.sh: Further cleanup
02c520928 memcg_stress_test.sh: ported to newlib
fc544f842 controllers/mem_process.c: Fix comparison warning
...

on 4k/64k pages defocnfig....looks fine to me.

Thanks a lot for the cleanup.

Regards

Cristian

Quoted 5 lines

Kind regards,
Petr

#7 lore raw

PATCH v3 1/1 memcg_stress_test.sh: ported to newlib

Hi Cristian,

Quoted 9 lines

Tested from you branch with ToT as:

10df48aa7 (HEAD -> pevik_ltp/christian/memcg_stress_test.sh.v3.fixes) memcg_stress_test.sh: fix memory usage
ebd6efaa6 memcg_stress_test.sh: Further cleanup
02c520928 memcg_stress_test.sh: ported to newlib
fc544f842 controllers/mem_process.c: Fix comparison warning
...

Quoted 3 lines

on 4k/64k pages defocnfig....looks fine to me.

Thanks a lot for the cleanup.

Thanks a lot for testing!
Pushed (with your Tested-by on my commit).

I did few cleanup changes in my commit (TINFO messages, fix missing local pid, redirect stderr for wait to make it quiet, move TPASS message into run_stress()).

I wonder how to suppress kill message for the first cgroup pid:
/opt/ltp/testcases/bin/memcg_stress_test.sh: line 65: 14750 User defined signal 1 memcg_process_stress $mem_size $interval

This pid is the only one killed by -USR1, but line 65 is not in fist kill loop. Therefore it cannot be killed by -KILL (second kill loop).

Kind regards,
Petr

#8 lore raw

PATCH v3 1/1 memcg_stress_test.sh: ported to newlib

Hi Petr

Quoted 17 lines · Petr Vorel

Hi Cristian,

Quoted 9 lines

Tested from you branch with ToT as:

10df48aa7 (HEAD -> pevik_ltp/christian/memcg_stress_test.sh.v3.fixes) memcg_stress_test.sh: fix memory usage
ebd6efaa6 memcg_stress_test.sh: Further cleanup
02c520928 memcg_stress_test.sh: ported to newlib
fc544f842 controllers/mem_process.c: Fix comparison warning
...

Quoted 3 lines

on 4k/64k pages defocnfig....looks fine to me.

Thanks a lot for the cleanup.

Thanks a lot for testing!
Pushed (with your Tested-by on my commit).

Quoted 9 lines

I did few cleanup changes in my commit (TINFO messages, fix missing local pid, redirect stderr for wait to make it quiet, move TPASS message into run_stress()).

I wonder how to suppress kill message for the first cgroup pid:
/opt/ltp/testcases/bin/memcg_stress_test.sh: line 65: 14750 User defined signal 1 memcg_process_stress $mem_size $interval

This pid is the only one killed by -USR1, but line 65 is not in fist kill loop. Therefore it cannot be killed by -KILL (second kill loop).

? ... I saw no error messages in my testing :D

Regards

Cristian

#9 lore raw

PATCH v3 1/1 memcg_stress_test.sh: ported to newlib

In reply to Petr Vorel
Quoted 11 lines · Petr Vorel

Hi Cristian,

Quoted 9 lines

Tested from you branch with ToT as:
10df48aa7 (HEAD -> pevik_ltp/christian/memcg_stress_test.sh.v3.fixes) memcg_stress_test.sh: fix memory usage
ebd6efaa6 memcg_stress_test.sh: Further cleanup
02c520928 memcg_stress_test.sh: ported to newlib
fc544f842 controllers/mem_process.c: Fix comparison warning
...
on 4k/64k pages defocnfig....looks fine to me.
Thanks a lot for the cleanup.

Quoted 8 lines

Thanks a lot for testing!
Pushed (with your Tested-by on my commit).

I did few cleanup changes in my commit (TINFO messages, fix missing local pid, redirect stderr for wait to make it quiet, move TPASS message into run_stress()).

I wonder how to suppress kill message for the first cgroup pid:
/opt/ltp/testcases/bin/memcg_stress_test.sh: line 65: 14750 User defined signal 1 memcg_process_stress $mem_size $interval

Hi Petr, Cristian

I also saw the same message occasionally when running memcg_stress_test, as below:
-----------------------------------------------------------------------
...
/opt/ltp/testcases/bin/memcg_stress_test.sh: line 65: 11262 User defined
signal 1 memcg_process_stress $mem_size $interval
/opt/ltp/testcases/bin/memcg_stress_test.sh: line 65: 11264 User defined
signal 1 memcg_process_stress $mem_size $interval
/opt/ltp/testcases/bin/memcg_stress_test.sh: line 65: 11267 User defined
signal 1 memcg_process_stress $mem_size $interval
...
-----------------------------------------------------------------------

It is possible for memcg_process_stress to be killed by SIGUSR1(default action), because it may
have received SIGUSR1 before completing the specified initialization of SIGUSR1 signal action.

Perpahs, we should sleep a few seconds to wait for the completion of initializing SIGUSR1 signal action. :-)

Please see my patch for details:
-----------------------------------------------------------------------

(quoted diff) +0 -0

a/testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh b/testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh index 5b19cc2..697a973 100755

testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +2 -0
@@ -81,6 +81,13 @@ run_stress()
                 pids="$! $pids"
         done
 
+       # If memcg_process_stress has received SIGUSR1 before completing the
+       # specified initialization of SIGUSR1 signal action, 

memcg_process_stress
+ # will be killed by SIGUSR1(i.e. default action). So we should
wait for
+ # the completion of initializing SIGUSR1 signal action by
sleeping a few
+ # seconds.
+ sleep 2
+
for pid in $pids; do
kill -USR1 $pid 2> /dev/null
done
-----------------------------------------------------------------------

Best Regards,
Xiao Yang

Quoted 7 lines

This pid is the only one killed by -USR1, but line 65 is not in fist kill loop. Therefore it cannot be killed by -KILL (second kill loop).

Kind regards,
Petr

#10 lore raw

PATCH v3 1/1 memcg_stress_test.sh: ported to newlib

Hi Xiao

Quoted 18 lines · Xiao Yang
Quoted 1 line · Petr Vorel

Hi Cristian,

Hi Petr, Cristian

I also saw the same message occasionally when running memcg_stress_test, as below:
-----------------------------------------------------------------------
...
/opt/ltp/testcases/bin/memcg_stress_test.sh: line 65: 11262 User defined
signal 1 memcg_process_stress $mem_size $interval
/opt/ltp/testcases/bin/memcg_stress_test.sh: line 65: 11264 User defined
signal 1 memcg_process_stress $mem_size $interval
/opt/ltp/testcases/bin/memcg_stress_test.sh: line 65: 11267 User defined
signal 1 memcg_process_stress $mem_size $interval
...
-----------------------------------------------------------------------

Quoted 7 lines

It is possible for memcg_process_stress to be killed by SIGUSR1(default action), because it may
have received SIGUSR1 before completing the specified initialization of SIGUSR1 signal action

Perpahs, we should sleep a few seconds to wait for the completion of initializing SIGUSR1 signal action. :-)

Mmm...yes that's a possibility....but I was noticing that we spawn a lot of bg processes in a loop before starting signaling all of them in the following loop (i.e. after all the forks)...but I just realized than then we signal them starting from the most recently spawned...not the oldest (in reverse order)....so maybe it could be enough to start signalling from the oldest.

Signaling in the natural-born order...

(quoted diff) +0 -0

b/testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh index 5b19cc292..aa91f77b2 100755

testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +1 -1
@@ -78,7 +78,7 @@ run_stress()
                mkdir /dev/memcg/$i 2> /dev/null
                memcg_process_stress $mem_size $interval &
                echo $! > /dev/memcg/$i/tasks
-               pids="$! $pids"
+               pids="$pids $!"
        done
 
        for pid in $pids; do

I understand that this approach is NO more deterministic than your sleeping solution, anyway.
IMHO the only real deterministic solution would be to somehow explicitly "sync" at the end of process-spawn loop, with the test script waiting for all the spawned child to be ready....but it is maybe a bit overkill for the issue we face. Maybe it is worth trying signaling in the correct order and then sleeping a bit and see if it solves...I haven't seen this issue at all on my local setup on ARM.

Regards

Thanks

Cristian

Quoted 41 lines

Please see my patch for details:
-----------------------------------------------------------------------

(quoted diff) +0 -0

a/testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh b/testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh index 5b19cc2..697a973 100755

testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +2 -0
@@ -81,6 +81,13 @@ run_stress()
                 pids="$! $pids"
         done
 
+       # If memcg_process_stress has received SIGUSR1 before completing the
+       # specified initialization of SIGUSR1 signal action, 

memcg_process_stress
+ # will be killed by SIGUSR1(i.e. default action). So we should
wait for
+ # the completion of initializing SIGUSR1 signal action by
sleeping a few

testcases/kernel/controllers/memcg/stress/memcg_stress_test.sh +3 -1
+       # seconds.
+       sleep 2
+
         for pid in $pids; do
                 kill -USR1 $pid 2> /dev/null
         done
-----------------------------------------------------------------------

Best Regards,
Xiao Yang

Quoted 7 lines

This pid is the only one killed by -USR1, but line 65 is not in fist kill loop. Therefore it cannot be killed by -KILL (second kill loop).

Kind regards,
Petr