Skip to contentSkip to conversation

cgroup: fix mount errors on cgroup subsys

PATCH New Li Wang 2 participants 4 messages Nov 23, 2016 – Nov 24, 2016

lore
Share & tools
lore

Li Wang

author
#0 lore raw

PATCH 1/1 cgroup: fix mount errors on cgroup subsys

This patch is on target to fix tiny errors:

mount: xxx is already mounted or /opt/ltp/testcases/bin/cgroup busy

cgroup_regression_test    3  TFAIL  :  ltpapicmd.c:190: Failed to mount cpu subsys
cgroup_regression_test    4  TCONF  :  ltpapicmd.c:190: CONFIG_LOCKDEP is not enabled
mount: xxx is already mounted or /opt/ltp/testcases/bin/cgroup busy
cgroup_regression_test    5  TFAIL  :  ltpapicmd.c:190: mount net_prio and pids failed
cgroup_regression_test    6  TCONF  :  ltpapicmd.c:190: CONFIG_CGROUP_NS

mount: xxx is already mounted or /opt/ltp/testcases/bin/cgroup busy
cgroup_regression_test 7 TFAIL : ltpapicmd.c:190: failed to mount net_prio

Test Environment: 23G RAM, 24 CPUS, x86_64, RHEL7.3GA

Signed-off-by Li Wang <[email protected]>
---
 .../controllers/cgroup/cgroup_regression_test.sh   | 46 +++++++++++++++++++++-
 1 file changed, 44 insertions(+), 2 deletions(-)
testcases/kernel/controllers/cgroup/cgroup_regression_test.sh +44 -2
index 30d0dbf..971243f 100755
@@ -181,10 +181,17 @@ test_3()
 		return
 	fi
 
+	local cpu=cpu
+
+	# For rhel7.3 or later, cpu is always binding with cpuacct subsys
+	if grep -w cpu /proc/mounts ; then
+		cpu=$(basename $(grep -w cpu /proc/mounts | cut -d ' ' -f 2))
+	fi
+
 	# Run the test for 30 secs
-	mount -t cgroup -o cpu xxx cgroup/
+	mount -t cgroup -o $cpu xxx cgroup/
 	if [ $? -ne 0 ]; then
-		tst_resm TFAIL "Failed to mount cpu subsys"
+		tst_resm TFAIL "Failed to mount $cpu subsys"
 		failed=1
 		return
 	fi

@@ -262,6 +269,20 @@ test_5()
 	subsys1=`tail -n 1 /proc/cgroups | awk '{ print $1 }'`
 	subsys2=`tail -n 2 /proc/cgroups | head -1 | awk '{ print $1 }'`
 
+	# On rhel7.3, cgroup subsystem is mounted automatically
+	# Here just unmout these two target temporarily for test
+	if grep -w $subsys1 /proc/mounts ; then
+		subsys1_orig=$(basename $(grep -w $subsys1 /proc/mounts | cut -d ' ' -f 2))
+		subsys1_path=$(grep -w $subsys1 /proc/mounts | cut -d ' ' -f 2)
+		umount $subsys1_path
+	fi
+
+	if grep -w $subsys2 /proc/mounts ; then
+		subsys2_orig=$(basename $(grep -w $subsys2 /proc/mounts | cut -d ' ' -f 2))
+		subsys2_path=$(grep -w $subsys2 /proc/mounts | cut -d ' ' -f 2)
+		umount $subsys2_path
+	fi
+
 	mount -t cgroup -o $subsys1,$subsys2 xxx cgroup/
 	if [ $? -ne 0 ]; then
 		tst_resm TFAIL "mount $subsys1 and $subsys2 failed"

@@ -298,6 +319,17 @@ test_5()
 	wait $!
 	rmdir cgroup/0
 	umount cgroup/
+
+	# waiting for cgroup/ unmout successfuly
+	tst_sleep 100ms
+
+	if [ -n "$subsys1_path" ]; then
+		mount -t cgroup -o $subsys1_orig cgroup $subsys1_path
+	fi
+
+	if [ -n "$subsys2_path" ]; then
+		mount -t cgroup -o $subsys2_orig cgroup $subsys2_path
+	fi
 }
 
 #---------------------------------------------------------------------------

@@ -411,6 +443,12 @@ test_7()
 
 	subsys=`tail -n 1 /proc/cgroups | awk '{ print $1 }'`
 
+	if grep -w $subsys /proc/mounts ; then
+		subsys_orig=$(basename $(grep -w $subsys /proc/mounts | cut -d ' ' -f 2))
+		subsys_path=$(grep -w $subsys /proc/mounts | cut -d ' ' -f 2)
+		umount $subsys_path
+	fi
+
 	# remount to add new subsystems to the hierarchy
 	i=1
 	while [ $i -le 2 ] ; do

@@ -426,6 +464,10 @@ test_7()
 		: $(( i += 1 ))
 	done
 
+	if [ -n "$subsys_path" ]; then
+		mount -t cgroup -o $subsys_orig cgroup $subsys_path
+	fi
+
 	tst_resm TPASS "no kernel bug was found"
 }
Signature · 1 line
1.8.3.1
#1 lore raw

PATCH 1/1 cgroup: fix mount errors on cgroup subsys

Hi!

Quoted 18 lines
---
 .../controllers/cgroup/cgroup_regression_test.sh   | 46 +++++++++++++++++++++-
 1 file changed, 44 insertions(+), 2 deletions(-)
testcases/kernel/controllers/cgroup/cgroup_regression_test.sh +6 -0
index 30d0dbf..971243f 100755
@@ -181,10 +181,17 @@ test_3()
 		return
 	fi
 
+	local cpu=cpu
+
+	# For rhel7.3 or later, cpu is always binding with cpuacct subsys
+	if grep -w cpu /proc/mounts ; then
+		cpu=$(basename $(grep -w cpu /proc/mounts | cut -d ' ' -f 2))
+	fi

testcases/kernel/controllers/cgroup/cgroup_regression_test.sh test_3() · line 197

Quoted 10 lines
testcases/kernel/controllers/cgroup/cgroup_regression_test.sh +3 -2
+
 	# Run the test for 30 secs
-	mount -t cgroup -o cpu xxx cgroup/
+	mount -t cgroup -o $cpu xxx cgroup/
 	if [ $? -ne 0 ]; then
-		tst_resm TFAIL "Failed to mount cpu subsys"
+		tst_resm TFAIL "Failed to mount $cpu subsys"
 		failed=1
 		return
 	fi

Why can't we mount the cgroup only when it's not mounted yet and use the path it's mounted on otherwise? We do that for the functional tetscases already.

And I do not think that umounting system cgroups is a good idea either. I would say that the test should use whatever is mounted on the system and if that is not possible I would rather produce TCONF than fiddle with umounting cgroups from the standard system path.

Signature · 2 lines
Cyril Hrubis
[email protected]
author
#2 lore raw

PATCH 1/1 cgroup: fix mount errors on cgroup subsys

Quoted 25 lines · Cyril Hrubis

Hi!

Quoted 24 lines
---
 .../controllers/cgroup/cgroup_regression_test.sh   | 46 +++++++++++++++++++++-
 1 file changed, 44 insertions(+), 2 deletions(-)
testcases/kernel/controllers/cgroup/cgroup_regression_test.sh +8 -2
index 30d0dbf..971243f 100755
@@ -181,10 +181,17 @@ test_3()
              return
      fi
 
+     local cpu=cpu
+
+     # For rhel7.3 or later, cpu is always binding with cpuacct subsys
+     if grep -w cpu /proc/mounts ; then
+             cpu=$(basename $(grep -w cpu /proc/mounts | cut -d ' ' -f 2))
+     fi
+
      # Run the test for 30 secs
-     mount -t cgroup -o cpu xxx cgroup/
+     mount -t cgroup -o $cpu xxx cgroup/
      if [ $? -ne 0 ]; then
-             tst_resm TFAIL "Failed to mount cpu subsys"
Quoted 8 lines

testcases/kernel/controllers/cgroup/cgroup_regression_test.sh test_3() · line 197

Quoted 4 lines
testcases/kernel/controllers/cgroup/cgroup_regression_test.sh +1 -0
+             tst_resm TFAIL "Failed to mount $cpu subsys"
              failed=1
              return
      fi

Why can't we mount the cgroup only when it's not mounted yet and use the path it's mounted on otherwise? We do that for the functional tetscases already.

I thought that the original case including cgroup mounted test, so I unmount the system cgroup and do itself mount & testing there.

Quoted 5 lines

And I do not think that umounting system cgroups is a good idea either. I would say that the test should use whatever is mounted on the system and if that is not possible I would rather produce TCONF than fiddle with umounting cgroups from the standard system path.

Using system cgroups is not a bad idea, I hesitated about doing that or just doing tiny fix. now you help me make a decision.

Thanks for your comments, I will rewrite this PATCH.

Signature · 3 lines
Regards,
Li Wang
Email: [email protected]
#3 lore raw

PATCH 1/1 cgroup: fix mount errors on cgroup subsys

Hi!

Quoted 6 lines
Quoted 3 lines

Why can't we mount the cgroup only when it's not mounted yet and use the path it's mounted on otherwise? We do that for the functional tetscases already.

I thought that the original case including cgroup mounted test, so I unmount the system cgroup and do itself mount & testing there.

Well the original test was written at times where cgroups were shiny and new and were not mounted by default on some well defined location...

Signature · 2 lines
Cyril Hrubis
[email protected]