[PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[Mike Rapoport]

The multicast and broadcast packets may have RTCF_LOCAL set in rt_flags
and therefore will be sent out bypassing encapsulation. This breaks
delivery of packets sent to the vxlan multicast group.
Disabling encapsulation bypass for multicasts and broadcasts fixes the
issue.

Signed-off-by: Mike Rapoport <mike.rapoport@ravellosystems.com>
---
 drivers/net/vxlan.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c
index 725aba3..97a306c 100644
--- a/drivers/net/vxlan.c
+++ b/drivers/net/vxlan.c
@@ -1014,7 +1014,8 @@ static netdev_tx_t vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
 	}
 
 	/* Bypass encapsulation if the destination is local */
-	if (rt->rt_flags & RTCF_LOCAL) {
+	if (rt->rt_flags & RTCF_LOCAL &&
+	    !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
 		struct vxlan_dev *dst_vxlan;
 
 		ip_rt_put(rt);
-- 
1.8.1.5

Re: [PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[Cong Wang]

On Sun, 14 Apr 2013 at 09:21 GMT, Mike Rapoport <mike.rapoport@ravellosystems.com> wrote:
> The multicast and broadcast packets may have RTCF_LOCAL set in rt_flags
> and therefore will be sent out bypassing encapsulation. This breaks
> delivery of packets sent to the vxlan multicast group.
> Disabling encapsulation bypass for multicasts and broadcasts fixes the
> issue.
>

This could probably fix the regression that I reported:
http://marc.info/?l=linux-netdev&m=136550149118954&w=2

I will test it tomorrow.

Re: [PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[David Miller]

From: Mike Rapoport <mike.rapoport@ravellosystems.com>
Date: Sun, 14 Apr 2013 12:21:51 +0300

> The multicast and broadcast packets may have RTCF_LOCAL set in rt_flags
> and therefore will be sent out bypassing encapsulation. This breaks
> delivery of packets sent to the vxlan multicast group.
> Disabling encapsulation bypass for multicasts and broadcasts fixes the
> issue.
> 
> Signed-off-by: Mike Rapoport <mike.rapoport@ravellosystems.com>

This still won't handle the case of encapsulations occuring in different
namespaces or virtual machines.

I really want this fixed correctly.

Re: [PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[Mike Rapoport]

On Sun, Apr 14, 2013 at 10:39 PM, David Miller <davem@davemloft.net> wrote:
> From: Mike Rapoport <mike.rapoport@ravellosystems.com>
> Date: Sun, 14 Apr 2013 12:21:51 +0300
>
>> The multicast and broadcast packets may have RTCF_LOCAL set in rt_flags
>> and therefore will be sent out bypassing encapsulation. This breaks
>> delivery of packets sent to the vxlan multicast group.
>> Disabling encapsulation bypass for multicasts and broadcasts fixes the
>> issue.
>>
>> Signed-off-by: Mike Rapoport <mike.rapoport@ravellosystems.com>
>
> This still won't handle the case of encapsulations occuring in different
> namespaces or virtual machines.
>
> I really want this fixed correctly.

I'm sorry, but I don't quite follow you. The patch does not aim to
reduce the amount of encapsulation. On the contrary, it prevents some
packets from being encapsulated when they should not be...

-- 
Sincerely yours,
Mike.

Re: [PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[David Miller]

From: Mike Rapoport <mike.rapoport@ravellosystems.com>
Date: Mon, 15 Apr 2013 08:31:15 +0300

> On Sun, Apr 14, 2013 at 10:39 PM, David Miller <davem@davemloft.net> wrote:
>> From: Mike Rapoport <mike.rapoport@ravellosystems.com>
>> Date: Sun, 14 Apr 2013 12:21:51 +0300
>>
>>> The multicast and broadcast packets may have RTCF_LOCAL set in rt_flags
>>> and therefore will be sent out bypassing encapsulation. This breaks
>>> delivery of packets sent to the vxlan multicast group.
>>> Disabling encapsulation bypass for multicasts and broadcasts fixes the
>>> issue.
>>>
>>> Signed-off-by: Mike Rapoport <mike.rapoport@ravellosystems.com>
>>
>> This still won't handle the case of encapsulations occuring in different
>> namespaces or virtual machines.
>>
>> I really want this fixed correctly.
> 
> I'm sorry, but I don't quite follow you. The patch does not aim to
> reduce the amount of encapsulation. On the contrary, it prevents some
> packets from being encapsulated when they should not be...

I'm saying you need to make sure the bypass is avoided in these
situations as well.

There is an ongoing discussion about this on the list between Cong
Wang and the author of the bypass changes.

Re: [PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[Mike Rapoport]

On Mon, Apr 15, 2013 at 8:45 AM, David Miller <davem@davemloft.net> wrote:
> From: Mike Rapoport <mike.rapoport@ravellosystems.com>
> Date: Mon, 15 Apr 2013 08:31:15 +0300
>
>> On Sun, Apr 14, 2013 at 10:39 PM, David Miller <davem@davemloft.net> wrote:
>>> From: Mike Rapoport <mike.rapoport@ravellosystems.com>
>>> Date: Sun, 14 Apr 2013 12:21:51 +0300
>>>
>>>> The multicast and broadcast packets may have RTCF_LOCAL set in rt_flags
>>>> and therefore will be sent out bypassing encapsulation. This breaks
>>>> delivery of packets sent to the vxlan multicast group.
>>>> Disabling encapsulation bypass for multicasts and broadcasts fixes the
>>>> issue.
>>>>
>>>> Signed-off-by: Mike Rapoport <mike.rapoport@ravellosystems.com>
>>>
>>> This still won't handle the case of encapsulations occuring in different
>>> namespaces or virtual machines.
>>>
>>> I really want this fixed correctly.
>>
>> I'm sorry, but I don't quite follow you. The patch does not aim to
>> reduce the amount of encapsulation. On the contrary, it prevents some
>> packets from being encapsulated when they should not be...
>
> I'm saying you need to make sure the bypass is avoided in these
> situations as well.

Ok, I see, thanks.

> There is an ongoing discussion about this on the list between Cong
> Wang and the author of the bypass changes.

Yes, I'm following that discussion.

-- 
Sincerely yours,
Mike.

Re: [PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[Mike Rapoport]

On Mon, Apr 15, 2013 at 8:45 AM, David Miller <davem@davemloft.net> wrote:
> From: Mike Rapoport <mike.rapoport@ravellosystems.com>
> Date: Mon, 15 Apr 2013 08:31:15 +0300
>
>> On Sun, Apr 14, 2013 at 10:39 PM, David Miller <davem@davemloft.net> wrote:
>>> From: Mike Rapoport <mike.rapoport@ravellosystems.com>
>>> Date: Sun, 14 Apr 2013 12:21:51 +0300
>>>
>>>> The multicast and broadcast packets may have RTCF_LOCAL set in rt_flags
>>>> and therefore will be sent out bypassing encapsulation. This breaks
>>>> delivery of packets sent to the vxlan multicast group.
>>>> Disabling encapsulation bypass for multicasts and broadcasts fixes the
>>>> issue.
>>>>
>>>> Signed-off-by: Mike Rapoport <mike.rapoport@ravellosystems.com>
>>>
>>> This still won't handle the case of encapsulations occuring in different
>>> namespaces or virtual machines.
>>>
>>> I really want this fixed correctly.
>>
>> I'm sorry, but I don't quite follow you. The patch does not aim to
>> reduce the amount of encapsulation. On the contrary, it prevents some
>> packets from being encapsulated when they should not be...
>
> I'm saying you need to make sure the bypass is avoided in these
> situations as well.

I've tried the cases with different namespaces and virtual machines.
Both worked for me.
Maybe I'm missing something, but I could not identify a case where
encapsulation occurs when it should not.

> There is an ongoing discussion about this on the list between Cong
> Wang and the author of the bypass changes.

And it seems that for Sridhar Samudrala it worked as well :)

-- 
Sincerely yours,
Mike.

Re: [PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[Cong Wang]

On Sun, 14 Apr 2013 at 09:21 GMT, Mike Rapoport <mike.rapoport@ravellosystems.com> wrote:
> The multicast and broadcast packets may have RTCF_LOCAL set in rt_flags
> and therefore will be sent out bypassing encapsulation. This breaks
> delivery of packets sent to the vxlan multicast group.
> Disabling encapsulation bypass for multicasts and broadcasts fixes the
> issue.
>
> Signed-off-by: Mike Rapoport <mike.rapoport@ravellosystems.com>

This fixes the regression I reported.

Tested-by: Cong Wang <xiyou.wangcong@gmail.com>

Thanks!

Re: [PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[Sridhar Samudrala]

On 4/15/2013 9:58 AM, Cong Wang wrote:
> On Sun, 14 Apr 2013 at 09:21 GMT, Mike Rapoport <mike.rapoport@ravellosystems.com> wrote:
>> The multicast and broadcast packets may have RTCF_LOCAL set in rt_flags
>> and therefore will be sent out bypassing encapsulation. This breaks
>> delivery of packets sent to the vxlan multicast group.
>> Disabling encapsulation bypass for multicasts and broadcasts fixes the
>> issue.
>>
>> Signed-off-by: Mike Rapoport <mike.rapoport@ravellosystems.com>
> This fixes the regression I reported.
>
> Tested-by: Cong Wang <xiyou.wangcong@gmail.com>
>
>
Acked-by: Sridhar Samudrala <sri@us.ibm.com>
Tested-by: Sridhar Samudrala <sri@us.ibm.com>

Thanks
Sridhar

Re: [PATCH net-next] vxlan: don't bypass encapsulation for multi- and broadcasts

[David Miller]

From: Cong Wang <xiyou.wangcong@gmail.com>
Date: Mon, 15 Apr 2013 16:58:02 +0000 (UTC)

> On Sun, 14 Apr 2013 at 09:21 GMT, Mike Rapoport <mike.rapoport@ravellosystems.com> wrote:
>> The multicast and broadcast packets may have RTCF_LOCAL set in rt_flags
>> and therefore will be sent out bypassing encapsulation. This breaks
>> delivery of packets sent to the vxlan multicast group.
>> Disabling encapsulation bypass for multicasts and broadcasts fixes the
>> issue.
>>
>> Signed-off-by: Mike Rapoport <mike.rapoport@ravellosystems.com>
> 
> This fixes the regression I reported.
> 
> Tested-by: Cong Wang <xiyou.wangcong@gmail.com>

Applied.