On 1/9/26 1:35 PM, Frode Nordahl wrote:
> On 1/9/26 10:59, Frode Nordahl wrote:
>> The tc command from iproute2 changed its rounding behavior in commit
>> d947f365602b ("tc: Fix rounding in tc_calc_xmittime and tc_calc_xmitsize.").
>> This caused the "QoS - basic configuration", "QoS - 64bit" and
>> "Ingress Policing - 64-bit" tests to fail because they were matching exact
>> burst and cburst values in tc output.
>>
>> For consistency and as a proactive measure in case of further changes to
>> tc, the following tests are also updated, although they are currently not
>> directly affected:
>> "offloads - set ingress_policing_rate and ingress_policing_burst - offloads
>> disabled"
>> "offloads - set ingress_policing_rate and ingress_policing_burst - offloads
>> enabled"
>>
>> The rounding fix means that burst and cburst values may differ slightly
>> from previous versions. For example, values that were previously 750000
>> might now be 749999 or similar variations.
>>
>> To maintain compatibility with both old and new versions of tc, the test
>> assertions now use pattern matching with dots that:
>> - Matches the most significant digit of the value
>> - Uses dots to match any character for remaining digits
>> - Maintains the correct total number of digits
>> - Preserves the unit suffix (e.g., 'b' for bytes)
>>
>> For example, '375000b' now matches '3.....b' which accepts any 6-digit
>> value starting with 3, allowing for rounding differences while still
>> validating the general magnitude is correct.
>>
>> This follows the same approach as the related OVN patch:
>> https://mail.openvswitch.org/pipermail/ovs-dev/2025-December/428593.html
>>
>> Reported-at: https://launchpad.net/bugs/2129005
>> Assisted-by: claude-sonnet-4.5, GitHub Copilot CLI
>> Signed-off-by: Frode Nordahl <[email protected]>
>> ---
>> tests/system-offloads-traffic.at | 14 ++++++--------
>> tests/system-traffic.at | 12 ++++++------
>> 2 files changed, 12 insertions(+), 14 deletions(-)
>>
>> diff --git a/tests/system-offloads-traffic.at
>> b/tests/system-offloads-traffic.at
>> index 492d5b689..e38221101 100644
>> --- a/tests/system-offloads-traffic.at
>> +++ b/tests/system-offloads-traffic.at
>> @@ -106,11 +106,10 @@ AT_CHECK([ovs-vsctl set interface ovs-p0
>> ingress_policing_burst=10])
>> AT_CHECK([ovs-vsctl --columns=other_config list open], [0], [dnl
>> other_config : {hw-offload="false"}
>> ])
>> +m4_define([POLICE_CONF], [rate 100Kbit burst 1...b])
>> AT_CHECK([tc -o -s -d filter show dev ovs-p0 ingress |
>> - sed -n 's/.*\(rate [[0-9]]*[[a-zA-Z]]* burst [[0-9]]*[[a-zA-Z]]*\).*/\1/;
>> T; p; q'],
>> - [0],[dnl
>> -rate 100Kbit burst 1280b
>> -])
>> + sed -n 's/.*\(rate [[0-9]]*[[a-zA-Z]]* burst [[0-9]]*[[a-zA-Z]]*\).*/\1/;
>> T; p; q' |
>> + grep POLICE_CONF])
>> AT_CHECK([tc -s -d filter show dev ovs-p0 ingress |
>> grep -E "basic|matchall" > /dev/null], [0])
>> OVS_TRAFFIC_VSWITCHD_STOP
>> @@ -128,11 +127,10 @@ AT_CHECK([ovs-vsctl set interface ovs-p0
>> ingress_policing_burst=10])
>> AT_CHECK([ovs-vsctl --columns=other_config list open], [0], [dnl
>> other_config : {hw-offload="true"}
>> ])
>> +m4_define([POLICE_CONF], [rate 100Kbit burst 1...b])
>> AT_CHECK([tc -o -s -d filter show dev ovs-p0 ingress |
>> - sed -n 's/.*\(rate [[0-9]]*[[a-zA-Z]]* burst [[0-9]]*[[a-zA-Z]]*\).*/\1/;
>> T; p; q'],
>> - [0],[dnl
>> -rate 100Kbit burst 1280b
>> -])
>> + sed -n 's/.*\(rate [[0-9]]*[[a-zA-Z]]* burst [[0-9]]*[[a-zA-Z]]*\).*/\1/;
>> T; p; q' |
>> + grep POLICE_CONF])
>> AT_CHECK([tc -o -s -d filter show dev ovs-p0 ingress | grep matchall |
>> sed -n 's/.*\(matchall\).*/\1/; T; p; q'], [0], [dnl
>> matchall
>> diff --git a/tests/system-traffic.at b/tests/system-traffic.at
>> index 58a46af0a..93264e8ce 100644
>> --- a/tests/system-traffic.at
>> +++ b/tests/system-traffic.at
>> @@ -3141,8 +3141,8 @@ OVS_WAIT_UNTIL([tc qdisc show dev ovs-tap0 | grep -q
>> htb])
>> OVS_WAIT_UNTIL([tc qdisc show dev ovs-tap1 | grep -q htb])
>>
>> dnl Check the configuration.
>> -m4_define([HTB_CONF0], [rate 2Mbit ceil 3Mbit burst 375000b cburst 375000b])
>> -m4_define([HTB_CONF1], [rate 4Mbit ceil 5Gbit burst 500000b cburst 500000b])
>> +m4_define([HTB_CONF0], [rate 2Mbit ceil 3Mbit burst 3.....b cburst 3.....b])
>> +m4_define([HTB_CONF1], [rate 4Mbit ceil 5Gbit burst 5.....b cburst 5.....b])
>> AT_CHECK([tc class show dev ovs-tap0 | grep -q 'class htb .* HTB_CONF0'])
>> AT_CHECK([tc class show dev ovs-tap0 | grep -q 'class htb .* HTB_CONF1'])
>> AT_CHECK([tc class show dev ovs-tap1 | grep -q 'class htb .* HTB_CONF0'])
>> @@ -3172,7 +3172,7 @@ AT_CHECK([ovs-vsctl set port ovs-p0 qos=@qos -- set
>> port ovs-p1 qos=@qos dnl
>> OVS_WAIT_UNTIL([tc qdisc show dev ovs-p0 | grep -q htb])
>> OVS_WAIT_UNTIL([tc qdisc show dev ovs-p1 | grep -q htb])
>>
>> -m4_define([HTB_CONF], [rate 40Gbit ceil 50Gbit burst 620000b cburst
>> 618750b])
>> +m4_define([HTB_CONF], [rate 40Gbit ceil 50Gbit burst 6.....b cburst
>> 6.....b])
>> AT_CHECK([tc class show dev ovs-p0 | grep -q 'class htb .* HTB_CONF'])
>> AT_CHECK([tc class show dev ovs-p1 | grep -q 'class htb .* HTB_CONF'])
>>
>> @@ -3189,10 +3189,10 @@ ADD_VETH(p0, ns0, br0, "10.1.1.1/24")
>> AT_CHECK([ovs-vsctl set interface ovs-p0 ingress_policing_rate=50000000])
>> AT_CHECK([ovs-vsctl set interface ovs-p0 ingress_policing_burst=400000])
>>
>> +m4_define([POLICE_CONF], [rate 50Gbit burst 7.......b])
>> AT_CHECK([tc -o -s -d filter show dev ovs-p0 ingress |
>> - sed -n 's/.*\(rate [[0-9]]*[[a-zA-Z]]* burst [[0-9]]*[[a-zA-Z]]*\).*/\1/;
>> T; p; q'],
>> - [0],[dnl
>> -rate 50Gbit burst 74500000b
>> + sed -n 's/.*\(rate [[0-9]]*[[a-zA-Z]]* burst [[0-9]]*[[a-zA-Z]]*\).*/\1/;
>> T; p; q' |
>> + grep POLICE_CONF])
>
> It seems I somehow managed to botch this by missing some quoting and
> ignoring of output (somehow there were no failures in my forks CI it
> must not have run this test or something like that).
>
> I'll send another revision, apologies for the noise.
>
It's weird. The test does fail in CI, but it somehow takes the whole suite
with it and after all of this the CI lane reports success for some reason...
https://github.com/ovsrobot/ovs/actions/runs/20851728783/job/59908210017#step:15:2830
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
