2018-05-12 11:01:29

by liuqifa

Subject: [PATCH] ipvlan: flush arp table when mac address changed

From: Keefe Liu <[email protected]>

When master device's mac has been changed, the
commit <32c10bbfe914> "ipvlan: always use the current L2
addr of the master" makes the IPVlan devices's mac changed
also, but it doesn't flush the IPVlan's arp table.

Signed-off-by: Keefe Liu <[email protected]>
---
drivers/net/ipvlan/ipvlan_main.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ipvlan/ipvlan_main.c b/drivers/net/ipvlan/ipvlan_main.c
index 450eec2..a1edfe1 100644
--- a/drivers/net/ipvlan/ipvlan_main.c
+++ b/drivers/net/ipvlan/ipvlan_main.c
@@ -7,6 +7,8 @@
*
*/

+#include <net/neighbour.h>
+#include <net/arp.h>
#include "ipvlan.h"

static unsigned int ipvlan_netid __read_mostly;
@@ -792,8 +794,10 @@ static int ipvlan_device_event(struct notifier_block *unused,
break;

case NETDEV_CHANGEADDR:
- list_for_each_entry(ipvlan, &port->ipvlans, pnode)
+ list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
ether_addr_copy(ipvlan->dev->dev_addr, dev->dev_addr);
+ neigh_changeaddr(&arp_tbl, ipvlan->dev);
+ }
break;

case NETDEV_PRE_TYPE_CHANGE:
--
1.8.3.1

2018-05-14 07:40:10

by Paolo Abeni

Subject: Re: [PATCH] ipvlan: flush arp table when mac address changed

Hi,

On Sat, 2018-05-12 at 19:00 +0800, [email protected] wrote:
> From: Keefe Liu <[email protected]>
>
> When master device's mac has been changed, the
> commit <32c10bbfe914> "ipvlan: always use the current L2
> addr of the master" makes the IPVlan devices's mac changed
> also, but it doesn't flush the IPVlan's arp table.
>
> Signed-off-by: Keefe Liu <[email protected]>
> ---
> drivers/net/ipvlan/ipvlan_main.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ipvlan/ipvlan_main.c b/drivers/net/ipvlan/ipvlan_main.c
> index 450eec2..a1edfe1 100644
> --- a/drivers/net/ipvlan/ipvlan_main.c
> +++ b/drivers/net/ipvlan/ipvlan_main.c
> @@ -7,6 +7,8 @@
> *
> */
>
> +#include <net/neighbour.h>
> +#include <net/arp.h>
> #include "ipvlan.h"
>
> static unsigned int ipvlan_netid __read_mostly;
> @@ -792,8 +794,10 @@ static int ipvlan_device_event(struct notifier_block *unused,
> break;
>
> case NETDEV_CHANGEADDR:
> - list_for_each_entry(ipvlan, &port->ipvlans, pnode)
> + list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
> ether_addr_copy(ipvlan->dev->dev_addr, dev->dev_addr);
> + neigh_changeaddr(&arp_tbl, ipvlan->dev);
> + }

Why don't using:

call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev);

instead?

that is what other stacked device - bridge and vlans - are currently
doing in the same scenario.

Thanks,

Paolo

2018-05-14 08:49:57

by liuqifa

Subject: 答复: [PATCH] ipvlan: flush arp table when mac address changed

Hi,
>
> Hi,
>
> On Sat, 2018-05-12 at 19:00 +0800, [email protected] wrote:
> > From: Keefe Liu <[email protected]>
> >
> > When master device's mac has been changed, the commit <32c10bbfe914>
> > "ipvlan: always use the current L2 addr of the master" makes the
> > IPVlan devices's mac changed also, but it doesn't flush the IPVlan's
> > arp table.
> >
> > Signed-off-by: Keefe Liu <[email protected]>
> > ---
> > drivers/net/ipvlan/ipvlan_main.c | 6 +++++-
> > 1 file changed, 5 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/ipvlan/ipvlan_main.c
> > b/drivers/net/ipvlan/ipvlan_main.c
> > index 450eec2..a1edfe1 100644
> > --- a/drivers/net/ipvlan/ipvlan_main.c
> > +++ b/drivers/net/ipvlan/ipvlan_main.c
> > @@ -7,6 +7,8 @@
> > *
> > */
> >
> > +#include <net/neighbour.h>
> > +#include <net/arp.h>
> > #include "ipvlan.h"
> >
> > static unsigned int ipvlan_netid __read_mostly; @@ -792,8 +794,10 @@
> > static int ipvlan_device_event(struct notifier_block *unused,
> > break;
> >
> > case NETDEV_CHANGEADDR:
> > - list_for_each_entry(ipvlan, &port->ipvlans, pnode)
> > + list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
> > ether_addr_copy(ipvlan->dev->dev_addr, dev-
> >dev_addr);
> > + neigh_changeaddr(&arp_tbl, ipvlan->dev);
> > + }
>
> Why don't using:
>
> call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev);
>
> instead?
>
> that is what other stacked device - bridge and vlans - are currently doing in
> the same scenario.
>
> Thanks,
>
> Paolo
>
Yes, I agre with you, this is a better solution.

Thanks

Keefe