On 7/13/22 10:14, Roi Dayan wrote:
>
>
> On 2022-07-04 4:46 PM, Roi Dayan wrote:
>>
>>
>> On 2022-06-30 11:34 AM, Tao Liu wrote:
>>> When netdev first opened by netdev_open(..., NULL, ...), netdev_class sets
>>> to system by default, and auto_classified sets to true.
>>>
>>> If netdev reopens by netdev_open(..., "system", ...), auto_classified
>>> should be cleared. This will be used in next patch to fix lag issue.
>>>
>>> Fixes: 8c2c225e481d ("netdev: Fix netdev_open() to track and recreate
>>> classless interfaces")
>>> Signed-off-by: Tao Liu <thomas....@ucloud.cn>
>>> ---
>>> lib/netdev.c | 41 +++++++++++++++++++++++------------------
>>> 1 file changed, 23 insertions(+), 18 deletions(-)
>>>
>>> diff --git a/lib/netdev.c b/lib/netdev.c
>>> index 8305f6c..b97408b 100644
>>> --- a/lib/netdev.c
>>> +++ b/lib/netdev.c
>>> @@ -387,25 +387,30 @@ netdev_open(const char *name, const char *type,
>>> struct netdev **netdevp)
>>> ovs_mutex_lock(&netdev_mutex);
>>> netdev = shash_find_data(&netdev_shash, name);
>>> - if (netdev &&
>>> - type && type[0] && strcmp(type, netdev->netdev_class->type)) {
>>> -
>>> - if (netdev->auto_classified) {
>>> - /* If this device was first created without a classification
>>> type,
>>> - * for example due to routing or tunneling code, and they keep
>>> a
>>> - * reference, a "classified" call to open will fail. In this
>>> case
>>> - * we remove the classless device, and re-add it below. We
>>> remove
>>> - * the netdev from the shash, and change the sequence, so
>>> owners of
>>> - * the old classless device can release/cleanup. */
>>> - if (netdev->node) {
>>> - shash_delete(&netdev_shash, netdev->node);
>>> - netdev->node = NULL;
>>> - netdev_change_seq_changed(netdev);
>>> - }
>>> + if (netdev && type && type[0]) {
>>> + if (strcmp(type, netdev->netdev_class->type)) {
>>> +
>>> + if (netdev->auto_classified) {
>>> + /* If this device was first created without a
>>> classification
>>> + * type, for example due to routing or tunneling code, and
>>> they
>>> + * keep a reference, a "classified" call to open will fail.
>>> + * In this case we remove the classless device, and re-add
>>> it
>>> + * below. We remove the netdev from the shash, and change
>>> the
>>> + * sequence, so owners of the old classless device can
>>> + * release/cleanup. */
>>> + if (netdev->node) {
>>> + shash_delete(&netdev_shash, netdev->node);
>>> + netdev->node = NULL;
>>> + netdev_change_seq_changed(netdev);
>>> + }
>>> - netdev = NULL;
>>> - } else {
>>> - error = EEXIST;
>>> + netdev = NULL;
>>> + } else {
>>> + error = EEXIST;
>>> + }
>>> + } else if (netdev->auto_classified) {
>>> + /* If netdev reopens of type "system", clear auto_classified */
>>> + netdev->auto_classified = false;
>>> }
>>> }
>>
>> this fix you issue with lag?
>> netdev_open() is not being called more than once?
>>
>> i'll take a little bit more this week and test a bit before acking.
>
> this is ok by me. didnt hit an issue.
>
> Acked-by: Roi Dayan <r...@nvidia.com>
Thanks! Applied and backported down to 2.13.
Best regards, Ilya Maximets.
_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
