[LTP] [PATCH] containers/netns/sysfs: check for known iface names, explicitly

[Jiri Jaburek]

The original code tries to check that no existing host interfaces
made it into the new netns. This needs exceptions for ifaces that
are created automatically, ie. 'lo', 'dummy0' and (on newer kernels)
'ip_vti0', which is not a long-term sustainable solution.

It also hade a corner case of always PASSing if the host had no
network interfaces of its own - it only checked for "no extra ifaces
in the new ns".

This change makes it use explicit pre-created interface names in
both the host and the new ns, avoiding any (existing) ones.

Signed-off-by: Jiri Jaburek <jjaburek@redhat.com>
---
 testcases/kernel/containers/netns/netns_sysfs.sh | 33 +++++++++---------------
 1 file changed, 12 insertions(+), 21 deletions(-)
 mode change 100755 => 100644 testcases/kernel/containers/netns/netns_sysfs.sh

diff --git a/testcases/kernel/containers/netns/netns_sysfs.sh b/testcases/kernel/containers/netns/netns_sysfs.sh
old mode 100755
new mode 100644
index 4c6e7e4..a887fa6
--- a/testcases/kernel/containers/netns/netns_sysfs.sh
+++ b/testcases/kernel/containers/netns/netns_sysfs.sh
@@ -25,7 +25,8 @@
 TCID="netns_sysfs"
 TST_TOTAL=3
 NS_TYPE="net,mnt"
-DUMMYDEV="dummy_test0"
+DUMMYDEV_HOST="dummy_test0"
+DUMMYDEV="dummy_test1"
 . test.sh
 
 setns_check
@@ -36,6 +37,7 @@ fi
 cleanup()
 {
 	tst_rmdir
+	ip link del $DUMMYDEV_HOST 2>/dev/null
 	ip link del $DUMMYDEV 2>/dev/null
 	kill -9 $NS_HANDLE 2>/dev/null
 }
@@ -48,9 +50,8 @@ if [ $? -eq 1 ]; then
 fi
 TST_CLEANUP=cleanup
 
-# exclude dummy0 (dummy1, etc.) from comparison as it gets automatically created
-# by the dummy device driver upon insmod/modprobe (during ip link add)
-ls /sys/class/net | grep -v 'dummy[0-9]\+' >sysfs_before
+ip link add $DUMMYDEV_HOST type dummy || \
+	tst_brkm TBROK "failed to add a new (host) dummy device"
 
 ns_exec $NS_HANDLE $NS_TYPE mount --make-rprivate /sys
 ns_exec $NS_HANDLE $NS_TYPE ip link add $DUMMYDEV type dummy || \
@@ -59,7 +60,7 @@ ns_exec $NS_HANDLE $NS_TYPE mount -t sysfs none /sys 2>/dev/null
 
 
 # TEST CASE #1
-ns_exec $NS_HANDLE $NS_TYPE test -d /sys/class/net/$DUMMYDEV
+ns_exec $NS_HANDLE $NS_TYPE test -e /sys/class/net/$DUMMYDEV
 if [ $? -eq 0 ]; then
 	tst_resm TPASS "sysfs in new namespace has $DUMMYDEV interface"
 else
@@ -69,27 +70,17 @@ fi
 
 # TEST CASE #2
 res=0
-for d in $(ns_exec $NS_HANDLE $NS_TYPE ls /sys/class/net/); do
-	case "$d" in
-		lo|$DUMMYDEV)
-			;;
-		*)
-			tst_resm TINFO "sysfs in new namespace should not contain: $d"
-			res=1
-			;;
-	esac
-done
-if [ $res -eq 0 ]; then
-	tst_resm TPASS "sysfs in new namespace has only lo and $DUMMYDEV interfaces"
+ns_exec $NS_HANDLE $NS_TYPE test -e /sys/class/net/$DUMMYDEV_HOST
+if [ $? -ne 0 ]; then
+	tst_resm TPASS "sysfs in new namespace does not have $DUMMYDEV_HOST interface"
 else
-	tst_resm TFAIL "sysfs in new namespace has more than lo and $DUMMYDEV interfaces"
+	tst_resm TFAIL "sysfs in new namespace contains $DUMMYDEV_HOST interface"
 fi
 
 
 # TEST CASE #3
-ls /sys/class/net | grep -v 'dummy[0-9]\+' >sysfs_after
-diff sysfs_before sysfs_after
-if [ $? -eq 0 ]; then
+test -e /sys/class/net/$DUMMYDEV
+if [ $? -ne 0 ]; then
 	tst_resm TPASS "sysfs not affected by a separate namespace"
 else
 	tst_resm TFAIL "sysfs affected by a separate namespace"
-- 
2.4.3