[PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[Francois-Xavier Le Bail]

- Use ipv6_chk_acast_addr_src() in inet6_bind().

Signed-off-by: Francois-Xavier Le Bail <fx.lebail@yahoo.com>
---
v2: ipv6_chk_acast_addr_src() was previously added.

 net/ipv6/af_inet6.c |    4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c
index c921d5d..68b81e9 100644
--- a/net/ipv6/af_inet6.c
+++ b/net/ipv6/af_inet6.c
@@ -347,7 +347,9 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
 				if (!(inet->freebind || inet->transparent) &&
 				    !ipv6_chk_addr(net, &addr->sin6_addr,
-						   dev, 0)) {
+						   dev, 0) &&
+				    !ipv6_chk_acast_addr_src(net, dev,
+							     &addr->sin6_addr)) {
 					err = -EADDRNOTAVAIL;
 					goto out_unlock;
 				}

Re: [PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[Hannes Frederic Sowa]

Hi!

[added Cc list from old thread]

On Wed, Feb 12, 2014 at 02:38:51PM +0100, Francois-Xavier Le Bail wrote:
> - Use ipv6_chk_acast_addr_src() in inet6_bind().
> 
> Signed-off-by: Francois-Xavier Le Bail <fx.lebail@yahoo.com>

I would agree with the change but would like to see some people from the old
thread about this change to agree with it, too.

> ---
> v2: ipv6_chk_acast_addr_src() was previously added.
> 
>  net/ipv6/af_inet6.c |    4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c
> index c921d5d..68b81e9 100644
> --- a/net/ipv6/af_inet6.c
> +++ b/net/ipv6/af_inet6.c
> @@ -347,7 +347,9 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
>  			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
>  				if (!(inet->freebind || inet->transparent) &&
>  				    !ipv6_chk_addr(net, &addr->sin6_addr,
> -						   dev, 0)) {
> +						   dev, 0) &&
> +				    !ipv6_chk_acast_addr_src(net, dev,
> +							     &addr->sin6_addr)) {
>  					err = -EADDRNOTAVAIL;
>  					goto out_unlock;
>  				}

Thanks,

  Hannes

Re: [PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[David Miller]

From: Hannes Frederic Sowa <hannes@stressinduktion.org>
Date: Thu, 13 Feb 2014 01:16:24 +0100

> [added Cc list from old thread]
> 
> On Wed, Feb 12, 2014 at 02:38:51PM +0100, Francois-Xavier Le Bail wrote:
>> - Use ipv6_chk_acast_addr_src() in inet6_bind().
>> 
>> Signed-off-by: Francois-Xavier Le Bail <fx.lebail@yahoo.com>
> 
> I would agree with the change but would like to see some people from the old
> thread about this change to agree with it, too.

I hear crickets... instead of other reviews :-)

Re: [PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[Christoph Paasch]

On 17/02/14 - 00:15:45, David Miller wrote:
> From: Hannes Frederic Sowa <hannes@stressinduktion.org>
> Date: Thu, 13 Feb 2014 01:16:24 +0100
> 
> > [added Cc list from old thread]
> > 
> > On Wed, Feb 12, 2014 at 02:38:51PM +0100, Francois-Xavier Le Bail wrote:
> >> - Use ipv6_chk_acast_addr_src() in inet6_bind().
> >> 
> >> Signed-off-by: Francois-Xavier Le Bail <fx.lebail@yahoo.com>
> > 
> > I would agree with the change but would like to see some people from the old
> > thread about this change to agree with it, too.
> 
> I hear crickets... instead of other reviews :-)

I don't have strong opinions for or against this patch.

I was only involved in the original thread because F-X claimed that
draft-iab-anycast-arch-implications (now RFC 7094) allows the use of anycast
addresses for TCP, which is not what RFC 7094 is saying. There is no
recommendation concerning TCP in the RFC and the situation is rather unclear.


Cheers,
Christoph

Re: [PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[Alexey Kuznetsov]

Hello!

On Mon, Feb 17, 2014 at 1:32 PM, Christoph Paasch
<christoph.paasch@uclouvain.be> wrote:
> I don't have strong opinions for or against this patch.
>
> I was only involved in the original thread because F-X claimed that
> draft-iab-anycast-arch-implications (now RFC 7094) allows the use of anycast
> addresses for TCP, which is not what RFC 7094 is saying. There is no
> recommendation concerning TCP in the RFC and the situation is rather unclear.

The same is here.

Using anycast as source or bind address, why not?

Use of anycast with TCP? Logically impossible, ergo prohibited.
If someone wants to play with fire, the option can be left hidden behind
a sysctl disabled by default.

Re: [PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[David Miller]

From: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Date: Mon, 17 Feb 2014 14:02:02 +0400

> Hello!
> 
> On Mon, Feb 17, 2014 at 1:32 PM, Christoph Paasch
> <christoph.paasch@uclouvain.be> wrote:
>> I don't have strong opinions for or against this patch.
>>
>> I was only involved in the original thread because F-X claimed that
>> draft-iab-anycast-arch-implications (now RFC 7094) allows the use of anycast
>> addresses for TCP, which is not what RFC 7094 is saying. There is no
>> recommendation concerning TCP in the RFC and the situation is rather unclear.
> 
> The same is here.
> 
> Using anycast as source or bind address, why not?
> 
> Use of anycast with TCP? Logically impossible, ergo prohibited.
> If someone wants to play with fire, the option can be left hidden behind
> a sysctl disabled by default.

Agreed.

Re: [PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[Hannes Frederic Sowa]

On Mon, Feb 17, 2014 at 02:19:49PM -0500, David Miller wrote:
> From: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
> Date: Mon, 17 Feb 2014 14:02:02 +0400
> 
> > Hello!
> > 
> > On Mon, Feb 17, 2014 at 1:32 PM, Christoph Paasch
> > <christoph.paasch@uclouvain.be> wrote:
> >> I don't have strong opinions for or against this patch.
> >>
> >> I was only involved in the original thread because F-X claimed that
> >> draft-iab-anycast-arch-implications (now RFC 7094) allows the use of anycast
> >> addresses for TCP, which is not what RFC 7094 is saying. There is no
> >> recommendation concerning TCP in the RFC and the situation is rather unclear.
> > 
> > The same is here.
> > 
> > Using anycast as source or bind address, why not?
> > 
> > Use of anycast with TCP? Logically impossible, ergo prohibited.
> > If someone wants to play with fire, the option can be left hidden behind
> > a sysctl disabled by default.
> 
> Agreed.

If a knob seems necessary I would vote for per address flags on those
anycast addresses (via setsockopt and netlink) so an application can
decide itself if it wants to do so.

Greetings,

  Hannes

Re: [PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[François-Xavier Le Bail]

> From: David Miller <davem@davemloft.net>

> From: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
>>  On Mon, Feb 17, 2014 at 1:32 PM, Christoph Paasch
>>  <christoph.paasch@uclouvain.be> wrote:
>>>  I don't have strong opinions for or against this patch.
>>> 
>>>  I was only involved in the original thread because F-X claimed that
>>>  draft-iab-anycast-arch-implications (now RFC 7094) allows the use of 
> anycast
>>>  addresses for TCP, which is not what RFC 7094 is saying. There is no
>>>  recommendation concerning TCP in the RFC and the situation is rather 
> unclear.
>> 
>>  The same is here.
>> 
>>  Using anycast as source or bind address, why not?
>> 
>>  Use of anycast with TCP? Logically impossible, ergo prohibited.
>>  If someone wants to play with fire, the option can be left hidden behind
>>  a sysctl disabled by default.
> 
> Agreed.

If I understand well the reviews, no change was ask for this patch.
The sysctl was asked in case of patch for TCP.
So, why the status is now "Changes Requested" ? 

Re: [PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[David Miller]

From: François-Xavier Le Bail <fx.lebail@yahoo.com>
Date: Mon, 17 Feb 2014 11:42:45 -0800 (PST)

> If I understand well the reviews, no change was ask for this patch.
> The sysctl was asked in case of patch for TCP.
> So, why the status is now "Changes Requested" ? 

This change is in generic bind() code, therefore it allows binding TCP
sockets with anycast, therefore you need to make a change.

Re: [PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[François-Xavier Le Bail]

----- Original Message -----
> From: David Miller <davem@davemloft.net>
> To: fx.lebail@yahoo.com
> Cc: kuznet@ms2.inr.ac.ru; christoph.paasch@uclouvain.be; hannes@stressinduktion.org; netdev@vger.kernel.org; eric.dumazet@gmail.com
> Sent: Monday, February 17, 2014 8:48 PM
> Subject: Re: [PATCH net-next v2] IPv6: enable bind() to assign an anycast address
> 
> From: François-Xavier Le Bail <fx.lebail@yahoo.com>
> 
> Date: Mon, 17 Feb 2014 11:42:45 -0800 (PST)
> 
>>  If I understand well the reviews, no change was ask for this patch.
>>  The sysctl was asked in case of patch for TCP.
>>  So, why the status is now "Changes Requested" ? 
> 
> This change is in generic bind() code, therefore it allows binding TCP
> sockets with anycast, therefore you need to make a change.

Understood, thanks.

[PATCH net-next v2] IPv6: enable bind() to assign an anycast address

[Francois-Xavier Le Bail]

- Use ipv6_chk_acast_addr_src() in inet6_bind().

Signed-off-by: Francois-Xavier Le Bail <fx.lebail@yahoo.com>
---
Tested with SOCK_DGRAM and SOCK_STREAM sockets.
Tested with link-local and global anycast addresses.

 net/ipv6/af_inet6.c |    4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c
index c921d5d..68b81e9 100644
--- a/net/ipv6/af_inet6.c
+++ b/net/ipv6/af_inet6.c
@@ -347,7 +347,9 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
 				if (!(inet->freebind || inet->transparent) &&
 				    !ipv6_chk_addr(net, &addr->sin6_addr,
-						   dev, 0)) {
+						   dev, 0) &&
+				    !ipv6_chk_acast_addr_src(net, dev,
+							     &addr->sin6_addr)) {
 					err = -EADDRNOTAVAIL;
 					goto out_unlock;
 				}