[PATCH] s390 - mark IPv6 support for QETH as broken

[PATCH] s390 - mark IPv6 support for QETH as broken

[Bastian Blank]

[-- Attachment #1.1: Type: text/plain, Size: 284 bytes --]

Hi Andrew

The attached patch marks IPv6 support for QETH broken, it is known to
need an extra patch to compile which was submitted one year ago but
never accepted.

Bastian

-- 
The heart is not a logical organ.
		-- Dr. Janet Wallace, "The Deadly Years", stardate 3479.4

[-- Attachment #1.2: diff --]
[-- Type: text/plain, Size: 598 bytes --]

diff -urN kernel-source-2.6.7.orig/drivers/s390/net/Kconfig kernel-source-2.6.7/drivers/s390/net/Kconfig
--- kernel-source-2.6.7.orig/drivers/s390/net/Kconfig	2004-06-16 05:19:53.000000000 +0000
+++ kernel-source-2.6.7/drivers/s390/net/Kconfig	2004-07-09 13:47:13.000000000 +0000
@@ -71,7 +71,7 @@
 
 config QETH_IPV6
 	bool "IPv6 support for gigabit ethernet"
-	depends on (QETH = IPV6) || (QETH && IPV6 = 'y')
+	depends on ((QETH = IPV6) || (QETH && IPV6 = 'y')) && BROKEN
 	help
 	  If CONFIG_QETH is switched on, this option will include IPv6
 	  support in the qeth device driver.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[Christoph Hellwig]

On Fri, Jul 09, 2004 at 04:06:30PM +0200, Bastian Blank wrote:
> Hi Andrew
> 
> The attached patch marks IPv6 support for QETH broken, it is known to
> need an extra patch to compile which was submitted one year ago but
> never accepted.

If it doesn't even compile BROKEN isn't enough.  In that case cometely
uncomment the option in Kconfig.

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[Adrian Bunk]

On Fri, Jul 09, 2004 at 03:09:47PM +0100, Christoph Hellwig wrote:
> On Fri, Jul 09, 2004 at 04:06:30PM +0200, Bastian Blank wrote:
> > Hi Andrew
> > 
> > The attached patch marks IPv6 support for QETH broken, it is known to
> > need an extra patch to compile which was submitted one year ago but
> > never accepted.
> 
> If it doesn't even compile BROKEN isn't enough.  In that case cometely
> uncomment the option in Kconfig.

Why shouldn't it be enough?

BROKEN and BROKEN_ON_SMP are the way non-compiling drivers as well as 
in any other way broken drivers are currently handled in 2.6 .

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[Andrew Morton]

Bastian Blank <bastian@waldi.eu.org> wrote:
>
>  The attached patch marks IPv6 support for QETH broken, it is known to
>  need an extra patch to compile which was submitted one year ago but
>  never accepted.

Well fixing it would be the preferable approach.  Where is the "extra
patch" and what was the complaint?

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[Christian Borntraeger]

Andrew Morton wrote:
> Bastian Blank <bastian@waldi.eu.org> wrote:
> >  The attached patch marks IPv6 support for QETH broken, it is known to
> >  need an extra patch to compile which was submitted one year ago but
> >  never accepted.
>
> Well fixing it would be the preferable approach.  Where is the "extra
> patch" and what was the complaint?

http://oss.sgi.com/projects/netdev/archive/2003-02/msg00061.html

The problem is that on s390 several virtual servers share the PHY and MAC 
part of the card and therefore have the same MAC address. Unfortunately 
IPv6 uses this MAC address to build an address. Now all virtual servers 
have the same auto configured IPv6 address - which is bad.
The proposed solution was to enable network card drivers to define their own 
ipv6 address auto configuration. 
If I recall correctly Dave Miller was not sure about rfc compliance. 

cheers

Christian

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[YOSHIFUJI Hideaki / 吉藤英明]

In article <200407092115.18097.linux-kernel@borntraeger.net> (at Fri, 9 Jul 2004 21:15:17 +0200), Christian Borntraeger <linux-kernel@borntraeger.net> says:

> Andrew Morton wrote:
> > Bastian Blank <bastian@waldi.eu.org> wrote:
> > >  The attached patch marks IPv6 support for QETH broken, it is known to
> > >  need an extra patch to compile which was submitted one year ago but
> > >  never accepted.
> >
> > Well fixing it would be the preferable approach.  Where is the "extra
> > patch" and what was the complaint?
> 
> http://oss.sgi.com/projects/netdev/archive/2003-02/msg00061.html

I don't see why it does NOT compile as you said.
It tried to add scheme for driver-specific interface identifier 
(including EUI-64 based on hardware address) generation scheme 
(in addition to the current ont; device-type-specific one).

> The problem is that on s390 several virtual servers share the PHY and MAC 
> part of the card and therefore have the same MAC address. Unfortunately 
> IPv6 uses this MAC address to build an address. Now all virtual servers 
> have the same auto configured IPv6 address - which is bad.
> The proposed solution was to enable network card drivers to define their own 
> ipv6 address auto configuration. 

I think it would be good to store interface identifier when
initializing interface, by RTM_NEWLINK attribute probably.

-- 
Hideaki YOSHIFUJI @ USAGI Project <yoshfuji@linux-ipv6.org>
GPG FP: 9022 65EB 1ECF 3AD1 0BDF  80D8 4807 F894 E062 0EEA

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[Bastian Blank]

[-- Attachment #1: Type: text/plain, Size: 785 bytes --]

On Fri, Jul 09, 2004 at 12:03:36PM -0700, Andrew Morton wrote:
> Bastian Blank <bastian@waldi.eu.org> wrote:
> >
> >  The attached patch marks IPv6 support for QETH broken, it is known to
> >  need an extra patch to compile which was submitted one year ago but
> >  never accepted.
> 
> Well fixing it would be the preferable approach.  Where is the "extra
> patch" and what was the complaint?

The original submission is recorded on
http://marc.theaimsgroup.com/?l=linux-net&m=104551077013011&w=2. And the
complaint was that it puts '"ipv6 stuff" into the generic netdevice
structure'. I don't know if this can be solved another way.

Bastian

-- 
Earth -- mother of the most beautiful women in the universe.
		-- Apollo, "Who Mourns for Adonais?" stardate 3468.1

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[David S. Miller]

On Fri, 9 Jul 2004 21:22:53 +0200
Bastian Blank <bastian@waldi.eu.org> wrote:

> The original submission is recorded on
> http://marc.theaimsgroup.com/?l=linux-net&m=104551077013011&w=2. And the
> complaint was that it puts '"ipv6 stuff" into the generic netdevice
> structure'. I don't know if this can be solved another way.

Put it in the inet6device private area.

It's been a year, and you haven't put forth the effort to look
for solutions like that?

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[Bastian Blank]

[-- Attachment #1: Type: text/plain, Size: 749 bytes --]

On Fri, Jul 09, 2004 at 12:30:05PM -0700, David S. Miller wrote:
> On Fri, 9 Jul 2004 21:22:53 +0200
> Bastian Blank <bastian@waldi.eu.org> wrote:
> 
> > The original submission is recorded on
> > http://marc.theaimsgroup.com/?l=linux-net&m=104551077013011&w=2. And the
> > complaint was that it puts '"ipv6 stuff" into the generic netdevice
> > structure'. I don't know if this can be solved another way.
> 
> Put it in the inet6device private area.

Where is it declared?

> It's been a year, and you haven't put forth the effort to look
> for solutions like that?

Err, I'm not the submitter of the original patch. I just wanted to have
the feature which needs that code marked as broken.

Bastian

-- 
Beam me up, Scotty!

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[YOSHIFUJI Hideaki / 吉藤英明]

In article <20040709194753.GB11138@wavehammer.waldi.eu.org> (at Fri, 9 Jul 2004 21:47:53 +0200), Bastian Blank <bastian@waldi.eu.org> says:

> Err, I'm not the submitter of the original patch. I just wanted to have
> the feature which needs that code marked as broken.

It is NOT broken at all.

-- 
Hideaki YOSHIFUJI @ USAGI Project <yoshfuji@linux-ipv6.org>
GPG FP: 9022 65EB 1ECF 3AD1 0BDF  80D8 4807 F894 E062 0EEA

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[Bastian Blank]

[-- Attachment #1: Type: text/plain, Size: 281 bytes --]

On Sat, Jul 10, 2004 at 05:00:28AM +0900, YOSHIFUJI Hideaki / 吉藤英明 wrote:
> It is NOT broken at all.

It does not compile at all.

Bastian

-- 
Dismissed.  That's a Star Fleet expression for, "Get out."
		-- Capt. Kathryn Janeway, Star Trek: Voyager, "The Cloud"

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[YOSHIFUJI Hideaki / 吉藤英明]

In article <20040709200445.GD11138@wavehammer.waldi.eu.org> (at Fri, 9 Jul 2004 22:04:45 +0200), Bastian Blank <bastian@waldi.eu.org> says:

> On Sat, Jul 10, 2004 at 05:00:28AM +0900, YOSHIFUJI Hideaki / 吉藤英明 wrote:
> > It is NOT broken at all.
> 
> It does not compile at all.

Okay... Let's remove generate_eui64 part instead.
Something like this. Thanks.

Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>

Index: drivers/s390/net/Kconfig
===================================================================
RCS file: /cvsroot/usagi/usagi/kernel/linux26/drivers/s390/net/Kconfig,v
retrieving revision 1.1.1.6
diff -u -u -r1.1.1.6 Kconfig
--- linux26/drivers/s390/net/Kconfig	28 Apr 2004 10:12:42 -0000	1.1.1.6
+++ linux26/drivers/s390/net/Kconfig	9 Jul 2004 20:31:06 -0000
@@ -75,7 +75,11 @@
 	help
 	  If CONFIG_QETH is switched on, this option will include IPv6
 	  support in the qeth device driver.
-	
+
+config QETH_IPV6_IFID
+	bool "Broken private inteface identifier support"
+	depends on ((QETH = IPV6) || (QETH && IPV6 = 'y')) && BROKEN
+
 config QETH_VLAN
 	bool "VLAN support for gigabit ethernet"
 	depends on (QETH = VLAN_8021Q) || (QETH && VLAN_8021Q = 'y')
Index: drivers/s390/net/qeth_main.c
===================================================================
RCS file: /cvsroot/usagi/usagi/kernel/linux26/drivers/s390/net/qeth_main.c,v
retrieving revision 1.1.1.5
diff -u -u -r1.1.1.5 qeth_main.c
--- linux26/drivers/s390/net/qeth_main.c	17 Jun 2004 01:22:21 -0000	1.1.1.5
+++ linux26/drivers/s390/net/qeth_main.c	9 Jul 2004 20:31:07 -0000
@@ -4732,6 +4732,7 @@
 }
 
 #ifdef CONFIG_QETH_IPV6
+#ifdef CONFIG_QETH_IPV6_IFID
 int
 qeth_ipv6_generate_eui64(u8 * eui, struct net_device *dev)
 {
@@ -4751,6 +4752,7 @@
 
 }
 #endif
+#endif
 
 static void
 qeth_get_mac_for_ipm(__u32 ipm, char *mac, struct net_device *dev)
@@ -5159,10 +5161,10 @@
 #ifdef CONFIG_QETH_IPV6
 	/*IPv6 address autoconfiguration stuff*/
 	card->dev->dev_id = card->info.unique_id & 0xffff;
+#ifdef CONFIG_QETH_IPV6_IFID
 	if (!(card->info.unique_id & UNIQUE_ID_NOT_BY_CARD))
 		card->dev->generate_eui64 = qeth_ipv6_generate_eui64;
-
-
+#endif
 #endif
 	dev->hard_header_parse = NULL;
 	dev->set_mac_address = NULL;

--yoshfuji

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[David S. Miller]

On Fri, 9 Jul 2004 21:47:53 +0200
Bastian Blank <bastian@waldi.eu.org> wrote:

> On Fri, Jul 09, 2004 at 12:30:05PM -0700, David S. Miller wrote:
> > On Fri, 9 Jul 2004 21:22:53 +0200
> > Bastian Blank <bastian@waldi.eu.org> wrote:
> > 
> > > The original submission is recorded on
> > > http://marc.theaimsgroup.com/?l=linux-net&m=104551077013011&w=2. And the
> > > complaint was that it puts '"ipv6 stuff" into the generic netdevice
> > > structure'. I don't know if this can be solved another way.
> > 
> > Put it in the inet6device private area.
> 
> Where is it declared?

struct inet6_dev is declared in include/net/if_inet6.h

I used grep to discover how the ipv6 device provate information
is managed, you could have as well :-)  I didn't know any of this
stuff offhand, I had to look it up.

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[Christian Borntraeger]

David S. Miller wrote:
> On Fri, 9 Jul 2004 21:22:53 +0200
> Bastian Blank <bastian@waldi.eu.org> wrote:
> > The original submission is recorded on
> > http://marc.theaimsgroup.com/?l=linux-net&m=104551077013011&w=2. And
> > the complaint was that it puts '"ipv6 stuff" into the generic netdevice
> > structure'. I don't know if this can be solved another way.
>
> Put it in the inet6device private area.
>
> It's been a year, and you haven't put forth the effort to look
> for solutions like that?

I think its sensible to add the driver authors to cc. Hopefully we can agree 
on a nice solution.

cheers

Christian

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[Bastian Blank]

[-- Attachment #1: Type: text/plain, Size: 717 bytes --]

On Fri, Jul 09, 2004 at 09:50:22PM +0200, Christian Borntraeger wrote:
> David S. Miller wrote:
> > Put it in the inet6device private area.

struct net_device defines a void pointer named ip6_ptr. This clearly
defines that as opaque data. There is no other ipv6 specific field.

A simple solution will be adding a pointer to a public struct with ipv6
specific things.

> > It's been a year, and you haven't put forth the effort to look
> > for solutions like that?
> 
> I think its sensible to add the driver authors to cc. Hopefully we can agree 
> on a nice solution.

I hope so.

Bastian

-- 
Only a fool fights in a burning house.
		-- Kank the Klingon, "Day of the Dove", stardate unknown

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[YOSHIFUJI Hideaki / 吉藤英明]

In article <20040709123005.086fdfc5.davem@redhat.com> (at Fri, 9 Jul 2004 12:30:05 -0700), "David S. Miller" <davem@redhat.com> says:

> On Fri, 9 Jul 2004 21:22:53 +0200
> Bastian Blank <bastian@waldi.eu.org> wrote:
> 
> > The original submission is recorded on
> > http://marc.theaimsgroup.com/?l=linux-net&m=104551077013011&w=2. And the
> > complaint was that it puts '"ipv6 stuff" into the generic netdevice
> > structure'. I don't know if this can be solved another way.
> 
> Put it in the inet6device private area.

David, inet6device area is created when we create first address.
The first address is likely link-local address based on
mac address, and kernel detects address duplication and 
delete the inet6device.  Finally, we don't have inet6device.

> It's been a year, and you haven't put forth the effort to look
> for solutions like that?

I'd suggest following approach.
 1. support up/down via netlink interface.
 2. add IFLA_IFID (or something like that) (RTM_NEWLINK attribute)
    inet6 layer create inet6device without an address
    when receiving this.

User will be able to do
 # ip link eth0 up eui64 de:ad:be:ef:ca:fe:ba:be
or something like this.

-- 
Hideaki YOSHIFUJI @ USAGI Project <yoshfuji@linux-ipv6.org>
GPG FP: 9022 65EB 1ECF 3AD1 0BDF  80D8 4807 F894 E062 0EEA

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[David S. Miller]

On Sat, 10 Jul 2004 04:59:04 +0900 (JST)
YOSHIFUJI Hideaki / ^[$B5HF#1QL@^[(B <yoshfuji@linux-ipv6.org> wrote:

> David, inet6device area is created when we create first address.

We could change the ipv6 layer to allocate the private
layer much earlier.

Re: [PATCH] s390 - mark IPv6 support for QETH as broken

[YOSHIFUJI Hideaki / 吉藤英明]

In article <20040709130553.223bdc6b.davem@redhat.com> (at Fri, 9 Jul 2004 13:05:53 -0700), "David S. Miller" <davem@redhat.com> says:

> On Sat, 10 Jul 2004 04:59:04 +0900 (JST)
> YOSHIFUJI Hideaki / ^[$B5HF#1QL@^[(B <yoshfuji@linux-ipv6.org> wrote:
> 
> > David, inet6device area is created when we create first address.
> 
> We could change the ipv6 layer to allocate the private
> layer much earlier.

Okay. Yes, I think we need this anyway.

-- 
Hideaki YOSHIFUJI @ USAGI Project <yoshfuji@linux-ipv6.org>
GPG FP: 9022 65EB 1ECF 3AD1 0BDF  80D8 4807 F894 E062 0EEA