Linux Ethernet Bridging - Mar 2023 - [Bridge] [PATCH v2 net-next 6/6] selftests: forwarding: add dynamic FDB test

On Sun, Mar 26, 2023 at 05:41:06PM +0200, Hans Schultz
wrote:
> On Mon, Mar 20, 2023 at 10:44, Ido Schimmel <idosch at nvidia.com>
wrote:
> >> +	$MZ $swp1 -c 1 -p 128 -t udp "sp=54321,dp=12345" \
> >> +		-a $mac -b `mac_get $h2` -A 192.0.2.1 -B 192.0.2.2 -q
> >> +	tc_check_packets "dev $swp2 egress" 1 1
> >> +	check_fail $? "Dynamic FDB entry did not age out"
> >
> > Shouldn't this be check_err()? After the FDB entry was aged you
want to
> > make sure that packets received via $swp1 with SMAC being $mac are no
> > longer forwarded by the bridge.
> 
> I was thinking that check_fail() will pass when tc_check_packets() does
> not see any packets, thus the test passing here when no packets are
forwarded?
What do you mean by "I was *thinking*"? How is it possible that you
are
submitting a selftest that you didn't bother running?!

I see you trimmed my earlier question: "Does this actually work?"

I tried it and it passed:

# ./bridge_locked_port.sh                         
TEST: Locked port ipv4                                              [ OK ]
TEST: Locked port ipv6                                              [ OK ]
TEST: Locked port vlan                                              [ OK ]
TEST: Locked port MAB                                               [ OK ]
TEST: Locked port MAB roam                                          [ OK ]
TEST: Locked port MAB configuration                                 [ OK ]
TEST: Locked port MAB FDB flush                                     [ OK ]

And I couldn't understand how that's even possible. Then I realized that
the entire test is dead code because the patch is missing this
fundamental hunk:

```
diff --git a/tools/testing/selftests/net/forwarding/bridge_locked_port.sh
b/tools/testing/selftests/net/forwarding/bridge_locked_port.sh
index dbc7017fd45d..5bf6b2aa1098 100755
--- a/tools/testing/selftests/net/forwarding/bridge_locked_port.sh
+++ b/tools/testing/selftests/net/forwarding/bridge_locked_port.sh
@@ -9,6 +9,7 @@ ALL_TESTS="
        locked_port_mab_roam
        locked_port_mab_config
        locked_port_mab_flush
+       locked_port_dyn_fdb
 "
 
 NUM_NETIFS=4
```

Which tells me that you didn't even try running it once. Now the test
failed as I expected:

# ./bridge_locked_port.sh                         
TEST: Locked port ipv4                                              [ OK ]
TEST: Locked port ipv6                                              [ OK ]
TEST: Locked port vlan                                              [ OK ]
TEST: Locked port MAB                                               [ OK ]
TEST: Locked port MAB roam                                          [ OK ]
TEST: Locked port MAB configuration                                 [ OK ]
TEST: Locked port MAB FDB flush                                     [ OK ]
TEST: Locked port dyn FDB                                           [FAIL]
        Packet not seen on egress after adding dynamic FDB

Fixed by:

```
@@ -336,7 +337,7 @@ locked_port_dyn_fdb()
        tc filter add dev $swp2 egress protocol ip pref 1 handle 1 flower \
                dst_ip 192.0.2.2 ip_proto udp dst_port 12345 action pass
 
-       $MZ $swp1 -c 1 -p 128 -t udp "sp=54321,dp=12345" \
+       $MZ $h1 -c 1 -p 128 -t udp "sp=54321,dp=12345" \
                -a $mac -b `mac_get $h2` -A 192.0.2.1 -B 192.0.2.2 -q
        tc_check_packets "dev $swp2 egress" 1 1
        check_err $? "Packet not seen on egress after adding dynamic
FDB"
```

Ran it again and it failed because of the second issue I pointed out:

# ./bridge_locked_port.sh 
TEST: Locked port ipv4                                              [ OK ]
TEST: Locked port ipv6                                              [ OK ]
TEST: Locked port vlan                                              [ OK ]
TEST: Locked port MAB                                               [ OK ]
TEST: Locked port MAB roam                                          [ OK ]
TEST: Locked port MAB configuration                                 [ OK ]
TEST: Locked port MAB FDB flush                                     [ OK ]
TEST: Locked port dyn FDB                                           [FAIL]
        Dynamic FDB entry did not age out

Fixed by:

```
@@ -346,7 +347,7 @@ locked_port_dyn_fdb()
        $MZ $swp1 -c 1 -p 128 -t udp "sp=54321,dp=12345" \
                -a $mac -b `mac_get $h2` -A 192.0.2.1 -B 192.0.2.2 -q
        tc_check_packets "dev $swp2 egress" 1 1
-       check_fail $? "Dynamic FDB entry did not age out"
+       check_err $? "Dynamic FDB entry did not age out"
 
        ip link set dev br0 type bridge ageing_time $ageing_time
        bridge link set dev $swp1 learning off locked off
```

# ./bridge_locked_port.sh 
TEST: Locked port ipv4                                              [ OK ]
TEST: Locked port ipv6                                              [ OK ]
TEST: Locked port vlan                                              [ OK ]
TEST: Locked port MAB                                               [ OK ]
TEST: Locked port MAB roam                                          [ OK ]
TEST: Locked port MAB configuration                                 [ OK ]
TEST: Locked port MAB FDB flush                                     [ OK ]
TEST: Locked port dyn FDB                                           [ OK ]

Sigh

On Tue, Mar 28, 2023 at 19:40, Ido Schimmel <idosch at nvidia.com>
wrote:
> On Sun, Mar 26, 2023 at 05:41:06PM +0200, Hans Schultz wrote:
>> On Mon, Mar 20, 2023 at 10:44, Ido Schimmel <idosch at
nvidia.com> wrote:
>> >> +	$MZ $swp1 -c 1 -p 128 -t udp "sp=54321,dp=12345" \
>> >> +		-a $mac -b `mac_get $h2` -A 192.0.2.1 -B 192.0.2.2 -q
>> >> +	tc_check_packets "dev $swp2 egress" 1 1
>> >> +	check_fail $? "Dynamic FDB entry did not age out"
>> >
>> > Shouldn't this be check_err()? After the FDB entry was aged
you want to
>> > make sure that packets received via $swp1 with SMAC being $mac are
no
>> > longer forwarded by the bridge.
>> 
>> I was thinking that check_fail() will pass when tc_check_packets() does
>> not see any packets, thus the test passing here when no packets are
forwarded?
>
> What do you mean by "I was *thinking*"? How is it possible that
you are
> submitting a selftest that you didn't bother running?!
>
Sorry, but I have sent you several emails telling you about the problems
I have with running the selftests due to changes in the phy etc. Maybe
you have just not received all those emails?

Have you checked spamfilters?

With the kernels now, I cannot even test with the software bridge and
selftests as the compile fails - probably due to changes in uapi headers
compared to what the packages my system uses expects.

On Tue, Mar 28, 2023 at 19:40, Ido Schimmel <idosch at nvidia.com>
wrote:
> On Sun, Mar 26, 2023 at 05:41:06PM +0200, Hans Schultz wrote:
>> On Mon, Mar 20, 2023 at 10:44, Ido Schimmel <idosch at
nvidia.com> wrote:
>> >> +	$MZ $swp1 -c 1 -p 128 -t udp "sp=54321,dp=12345" \
>> >> +		-a $mac -b `mac_get $h2` -A 192.0.2.1 -B 192.0.2.2 -q
>> >> +	tc_check_packets "dev $swp2 egress" 1 1
>> >> +	check_fail $? "Dynamic FDB entry did not age out"
>> >
>> > Shouldn't this be check_err()? After the FDB entry was aged
you want to
>> > make sure that packets received via $swp1 with SMAC being $mac are
no
>> > longer forwarded by the bridge.
>> 
>> I was thinking that check_fail() will pass when tc_check_packets() does
>> not see any packets, thus the test passing here when no packets are
forwarded?
>
> What do you mean by "I was *thinking*"? How is it possible that
you are
> submitting a selftest that you didn't bother running?!
>
> I see you trimmed my earlier question: "Does this actually work?"
>
> I tried it and it passed:
>
> # ./bridge_locked_port.sh                         
> TEST: Locked port ipv4                                              [ OK ]
> TEST: Locked port ipv6                                              [ OK ]
> TEST: Locked port vlan                                              [ OK ]
> TEST: Locked port MAB                                               [ OK ]
> TEST: Locked port MAB roam                                          [ OK ]
> TEST: Locked port MAB configuration                                 [ OK ]
> TEST: Locked port MAB FDB flush                                     [ OK ]
>
> And I couldn't understand how that's even possible. Then I realized
that
> the entire test is dead code because the patch is missing this
> fundamental hunk:
>
> ```
> diff --git a/tools/testing/selftests/net/forwarding/bridge_locked_port.sh
b/tools/testing/selftests/net/forwarding/bridge_locked_port.sh
> index dbc7017fd45d..5bf6b2aa1098 100755
> --- a/tools/testing/selftests/net/forwarding/bridge_locked_port.sh
> +++ b/tools/testing/selftests/net/forwarding/bridge_locked_port.sh
> @@ -9,6 +9,7 @@ ALL_TESTS="
>         locked_port_mab_roam
>         locked_port_mab_config
>         locked_port_mab_flush
> +       locked_port_dyn_fdb
>  "
>  
>  NUM_NETIFS=4
> ```
>
> Which tells me that you didn't even try running it once.
Not true, it reveals that I forgot to put it in the patch, that's all. As
I cannot run several of these tests because of memory constraints I link
the file to a copy in a rw area where I modify the list and just run one
of the subtests at a time. If I try to run the whole it always fails
after a couple of sub-tests with an error.

It seems to me that these scripts are quite memory consuming as they
accumulate memory consuption in relation to what is loaded along the
way. A major problem with my system.