It's great!
I should have read the comment. A new patch will be published soon.
From: Ilya Maximets <i.maxim...@ovn.org>
Date: 2023-03-01 19:34:01
To: Faicker Mo <faicker...@ucloud.cn>
Cc: d...@openvswitch.org,i.maxim...@ovn.org,Eelco Chaudron
<echau...@redhat.com>,Simon Horman <simon.hor...@corigine.com>
Subject: Re: [ovs-dev] [PATCH v7] netdev-offload-tc: del ufid mapping if device
not exist>On 3/1/23 03:12, Faicker Mo wrote:
>> The device may be deleted and added with ifindex changed.
>> The tc rules on the device will be deleted if the device is deleted.
>> The func tc_del_filter will fail when flow del. The mapping of
>> ufid to tc will not be deleted.
>> The traffic will trigger the same flow(with same ufid) to put to tc
>> on the new device. Duplicated ufid mapping will be added.
>> If the hashmap is expanded, the old mapping entry will be the first entry,
>> and now the dp flow can't be deleted.
>>
>>
>> Signed-off-by: Faicker Mo <faicker...@ucloud.cn>
>> ---
>> v2:
>> - Add tc offload test case
>> v3:
>> - No change
>> v4:
>> - No change
>> v5:
>> - No change
>> v6:
>> - No change
>> v7:
>> - Minor fix for test case and rebased
>> ---
>> lib/netdev-offload-tc.c | 3 ++-
>> tests/system-offloads-traffic.at | 46 ++++++++++++++++++++++++++++++++
>> 2 files changed, 48 insertions(+), 1 deletion(-)
>>
>>
>> diff --git a/lib/netdev-offload-tc.c b/lib/netdev-offload-tc.c
>> index 4fb9d9f21..dd2020cad 100644
>> --- a/lib/netdev-offload-tc.c
>> +++ b/lib/netdev-offload-tc.c
>> @@ -276,8 +276,9 @@ del_filter_and_ufid_mapping(struct tcf_id *id, const
>> ovs_u128 *ufid,
>> }
>>
>> err = tc_del_flower_filter(id);
>> - if (!err) {
>> + if (!err || err == ENODEV) {
>> del_ufid_tc_mapping(ufid);
>> + return 0;
>> }
>> return err;
>> }
>> diff --git a/tests/system-offloads-traffic.at
>> b/tests/system-offloads-traffic.at
>> index 7558812eb..9b2f12ddb 100644
>> --- a/tests/system-offloads-traffic.at
>> +++ b/tests/system-offloads-traffic.at
>> @@ -690,3 +690,49 @@
>> OVS_CHECK_ACTIONS([check_pkt_len(size=200,gt(5),le(check_pkt_len(size=100,gt(5),
>>
>> OVS_TRAFFIC_VSWITCHD_STOP
>> AT_CLEANUP
>> +
>> +AT_SETUP([offloads - delete ufid mapping if device not exist - offloads
>> enabled])
>> +OVS_TRAFFIC_VSWITCHD_START([], [], [-- set Open_vSwitch .
>> other_config:hw-offload=true])
>> +
>> +AT_CHECK([ovs-ofctl add-flow br0 "actions=normal"])
>> +
>> +ADD_NAMESPACES(at_ns0, at_ns1, at_ns2)
>> +# avoid unexpected ipv6 flow
>> +AT_CHECK([sysctl -w net.ipv6.conf.br0.disable_ipv6=1], [0], [ignore])
>> +NS_CHECK_EXEC([at_ns0], [sysctl -w net.ipv6.conf.all.disable_ipv6=1], [0],
>> [ignore])
>> +NS_CHECK_EXEC([at_ns1], [sysctl -w net.ipv6.conf.all.disable_ipv6=1], [0],
>> [ignore])
>> +NS_CHECK_EXEC([at_ns2], [sysctl -w net.ipv6.conf.all.disable_ipv6=1], [0],
>> [ignore])
>> +
>> +ADD_VETH(p0, at_ns0, br0, "10.1.1.1/24", "aa:1a:54:e9:c5:56")
>> +ADD_VETH(p1, at_ns1, br0, "10.1.1.2/24")
>> +
>> +NS_CHECK_EXEC([at_ns0], [ping -q -c 2 -i 0.2 10.1.1.2 | FORMAT_PING], [0],
>> [dnl
>> +2 packets transmitted, 2 received, 0% packet loss, time 0ms
>> +])
>> +sleep 1
>> +
>> +#delete and add interface ovs-p0/p0
>> +AT_CHECK([ip link del dev ovs-p0])
>> +AT_CHECK([ip link add p0 type veth peer name ovs-p0 || return 77])
>> +AT_CHECK([ip link set p0 netns at_ns0])
>> +AT_CHECK([ip link set dev ovs-p0 up])
>> +NS_CHECK_EXEC([at_ns0], [ip addr add dev p0 "10.1.1.1/24"])
>> +NS_CHECK_EXEC([at_ns0], [ip link set dev p0 up])
>> +NS_CHECK_EXEC([at_ns0], [ip link set dev p0 address "aa:1a:54:e9:c5:56"])
>> +
>> +sleep 12
>
>Hi. AFAICT, this sleep is added to wait for flow expiration in the datapath.
>But it's a very large seep time. We should be able to avoid it.
>
>I see the comment on the very first version of this patch from Eelco:
>
>"
> I did not test the script, but the (long) sleeps are not really
> desired. Why are they needed, and can they maybe be replaced with
> “ovs-appctl revalidator/wait|purge”
>"
>
>But I didn't find a reply to it.
>
>Did you try using 'revalidator/purge' appctl here instead? At least for
>the first 12 second sleep. The second one seems to be more specific on
>what it is trying to do, while the first one is just waiting for removal
>of flows for the initial random traffic.
>
>If nothing can be done, then we should decrease the 'max-idle' parameter
>from the default 10 seconds to something more reasonable, so revalidators
>will remove flows faster.
>
>> +#flows to trigger the hmap expand once
>
>nit: 'dnl' is a better way of adding comments to the test (instead of '#').
>
>> +ADD_VETH(p2, at_ns2, br0, "10.1.1.3/24")
>> +NS_CHECK_EXEC([at_ns0], [ping -q -c 2 -i 0.2 10.1.1.2 | FORMAT_PING], [0],
>> [dnl
>> +2 packets transmitted, 2 received, 0% packet loss, time 0ms
>> +])
>> +NS_CHECK_EXEC([at_ns0], [ping -q -c 2 -i 0.2 10.1.1.3 | FORMAT_PING], [0],
>> [dnl
>> +2 packets transmitted, 2 received, 0% packet loss, time 0ms
>> +])
>> +
>> +sleep 12
>> +AT_CHECK([test $(ovs-appctl dpctl/dump-flows | grep -c "eth_type(0x0800)")
>> -eq 0], [0], [ignore])
>> +
>> +OVS_TRAFFIC_VSWITCHD_STOP(["/could not open network device ovs-p0/d
>> +/on nonexistent port/d"])
>> +AT_CLEANUP
>
_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
