On Tue, Sep 22, 2015 at 02:38:45PM +0200, Stanislav Kinsburskiy wrote:
>
>
> 22.09.2015 13:28, Andrew Vagin пишет:
> >On Tue, Sep 22, 2015 at 01:20:11PM +0200, Stanislav Kinsburskiy wrote:
> >>
> >>22.09.2015 13:15, Andrew Vagin пишет:
> >>>On Tue, Sep 22, 2015 at 11:40:03AM +0200, Stanislav Kinsburskiy wrote:
> >>>>22.09.2015 11:36, Andrew Vagin пишет:
> >>>>>On Mon, Sep 21, 2015 at 07:44:31PM +0400, Stanislav Kinsburskiy wrote:
> >>>>>>From: Stanislav Kinsburskiy <skinsbur...@parallels.com>
> >>>>>>
> >>>>>>This functions are mostly duplicating existent code in per-net hooks,
> >>>>>>except
> >>>>>>some initialization and locking, which are performed in rtnetlink layer.
> >>>>>>
> >>>>>>https://jira.sw.ru/browse/PSBM-23441
> >>>>>>
> >>>>>>Signed-off-by: Stanislav Kinsburskiy <skinsbur...@parallels.com>
> >>>>>>---
> >>>>>> drivers/net/venetdev.c | 49
> >>>>>> ++++++++++++++++++++++++++++++++++++++++++++++++
> >>>>>> 1 file changed, 49 insertions(+)
> >>>>>>
> >>>>>>diff --git a/drivers/net/venetdev.c b/drivers/net/venetdev.c
> >>>>>>index 50b3c91..261be76 100644
> >>>>>>--- a/drivers/net/venetdev.c
> >>>>>>+++ b/drivers/net/venetdev.c
> >>>>>>@@ -749,6 +749,53 @@ static void venet_setup(struct net_device *dev)
> >>>>>> SET_ETHTOOL_OPS(dev, &venet_ethtool_ops);
> >>>>>> }
> >>>>>>+static void venet_dellink(struct net_device *dev, struct list_head
> >>>>>>*head)
> >>>>>>+{
> >>>>>>+ struct ve_struct *env = dev->nd_net->owner_ve;
> >>>>>>+
> >>>>>>+ venet_ext_clean(env);
> >>>>>>+ veip_stop(env);
> >>>>>>+
> >>>>>>+ env->_venet_dev = NULL;
> >>>>>>+ unregister_netdevice_queue(dev, head);
> >>>>>>+}
> >>>>>>+
> >>>>>>+static int venet_newlink(struct net *src_net, struct net_device *dev,
> >>>>>>+ struct nlattr *tb[], struct nlattr *data[])
> >>>>>>+{
> >>>>>>+ struct ve_struct *env = src_net->owner_ve;
> >>>>>>+ int err;
> >>>>>>+
> >>>>>>+ if (env->ve_netns && src_net != env->ve_netns) {
> >>>>>>+ /* Don't create venet-s in sub net namespaces */
> >>>>>>+ return 0;
> >>>Are you sure that we can return 0 here? Looks like EEXIST.
> >>
> >>It's a former logic, so I preserved it.
> >>Introduced by Andrew Vagin here:
> >>4504338aaf4c7deb2285fc192ddfebf711f10129
> >I don't return 0 from venet_newlink...
>
> Makes sense. Should I return -ENOTSUPP?
ENOTSUPP maybe even better that EEXIST
>
> >
> >>>>>>+ }
> >>>>>>+
> >>>>>>+ if (env->veip)
> >>>>>Where is env->veip set to NULL, when venet is removed?
> >>>>>
> >>>>This is missing, thanks. I'll add.
> >>>>
> >>>>>>+ return -EEXIST;
> >>>>>>+
> >>>>>>+ env->ve_netns = src_net;
> >>>>>Why do we set ve_netns here?
> >>>>To prevent venet creation in nested network namespaces within containers.
> >>>>
> >>>>>>+
> >>>>>>+ err = veip_start(env);
> >>>>>>+ if (err)
> >>>>>>+ goto err;
> >>>>>>+
> >>>>>>+ dev->features |= NETIF_F_NETNS_LOCAL;
> >>>>>>+
> >>>>>>+ err = register_netdevice(dev);
> >>>>>>+ if (err)
> >>>>>>+ goto err_stop;
> >>>>>>+
> >>>>>>+ env->_venet_dev = dev;
> >>>>>>+ return 0;
> >>>>>>+
> >>>>>>+err_stop:
> >>>>>>+ veip_stop(env);
> >>>>>>+err:
> >>>>>>+ env->ve_netns = NULL;
> >>>>>>+ return err;
> >>>>>>+}
> >>>>>>+
> >>>>>> #ifdef CONFIG_PROC_FS
> >>>>>> static void veaddr_seq_print(struct seq_file *m, struct ve_struct *ve)
> >>>>>> {
> >>>>>>@@ -1207,6 +1254,8 @@ static const struct nla_policy
> >>>>>>venet_policy[VENET_INFO_MAX + 1] = {
> >>>>>> static struct rtnl_link_ops venet_link_ops = {
> >>>>>> .kind = "venet",
> >>>>>> .priv_size = sizeof(struct veip_struct),
> >>>>>>+ .newlink = venet_newlink,
> >>>>>>+ .dellink = venet_dellink,
> >>>>>> .setup = venet_setup,
> >>>>>> .changelink = venet_changelink,
> >>>>>> .policy = venet_policy,
> >>>>>>
> >>>>>>_______________________________________________
> >>>>>>Devel mailing list
> >>>>>>Devel@openvz.org
> >>>>>>https://lists.openvz.org/mailman/listinfo/devel
>
_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
