VLAN patches

VLAN patches

[Andersson Björn]

[-- Attachment #1: Type: text/plain, Size: 641 bytes --]

Hi,
I hope you are the right receiver of 8021q-patches.
We are running SuSe 8.0, i.e kernel 2.4.18.

- When addding several egress-mappings we get stuck in an eternal
loop if they end up in the same hash-bucket, see vlan_dev.c.path.

- If we try to remove a vlan with VID 0, ifconfig stops working
completly.
We fixed it with vlan.c.patch.


-- 
* Björn Andersson                   bjorn.andersson@ebc.ericsson.se *
* Ericsson Enterprise AB            NA/EBC/BEES/DNCC                *
* Nacka Strand                      tel: +46-8-422 3512             *
* 13 189 Stockholm, Sweden          fax: +46-8-422 1010             *

[-- Attachment #2: vlan.c.patch --]
[-- Type: text/plain, Size: 342 bytes --]

--- linux-2.4.18.SuSE/net/8021q/vlan.c.orig	Wed Mar 27 13:57:17 2002
+++ linux-2.4.18.SuSE/net/8021q/vlan.c	Wed Sep 18 13:19:13 2002
@@ -207,7 +207,7 @@
 #endif
 
 	/* sanity check */
-	if ((vlan_id >= VLAN_VID_MASK) || (vlan_id <= 0))
+	if ((vlan_id >= VLAN_VID_MASK) || (vlan_id < 0))
 		return -EINVAL;
 
 	spin_lock_bh(&vlan_group_lock);

[-- Attachment #3: vlan_dev.c.patch --]
[-- Type: text/plain, Size: 269 bytes --]

--- linux-2.4.18.SuSE/net/8021q/vlan_dev.c.orig	Wed Mar 27 13:57:17 2002
+++ linux-2.4.18.SuSE/net/8021q/vlan_dev.c	Wed Sep 18 13:19:52 2002
@@ -570,6 +570,7 @@
 					dev_put(dev);
 					return 0;
 				}
+				mp = mp->next;
 			}
 
 			/* Create a new mapping then. */

Re: [VLAN] VLAN patches

[Matti Aarnio]

On Mon, Sep 30, 2002 at 10:23:44AM +0200, Andersson Björn wrote:
> Hi,
> I hope you are the right receiver of 8021q-patches.
> We are running SuSe 8.0, i.e kernel 2.4.18.

  NETDEV isn't right place.  I resent your letter to:
          vlan@Scry.WANfear.com

  (which is strictly posting-by-subscribers-only / others-via-moderator
   list...)

> -- 
> * Björn Andersson                   bjorn.andersson@ebc.ericsson.se *

/Matti Aarnio

Re: [VLAN] VLAN patches

[Ben Greear]

Matti Aarnio wrote:
> On Mon, Sep 30, 2002 at 10:23:44AM +0200, Andersson Björn wrote:
> 
>>Hi,
>>I hope you are the right receiver of 8021q-patches.
>>We are running SuSe 8.0, i.e kernel 2.4.18.
> 
> 
>   NETDEV isn't right place.  I resent your letter to:
>           vlan@Scry.WANfear.com
> 
>   (which is strictly posting-by-subscribers-only / others-via-moderator
>    list...)

It is nice to get VLAN questions and patches to the vlan list, but
I see no reason to restrict patches to that list.  Dave is doing a
lot of the VLAN patching and up-keep these days, and I'm not even
sure he reads the vlan list....

So, I personally would be happy to have patches sent to the vlan
list, and cc'd to netdev and/or lkml.  Questions about using/configuring
vlans should most likely remain on the vlan mailing list though...

Thanks,
Ben

> 
> 
>>-- 
>>* Björn Andersson                   bjorn.andersson@ebc.ericsson.se *
> 
> 
> /Matti Aarnio
> 
> 


-- 
Ben Greear <greearb@candelatech.com>       <Ben_Greear AT excite.com>
President of Candela Technologies Inc      http://www.candelatech.com
ScryMUD:  http://scry.wanfear.com     http://scry.wanfear.com/~greear

Re: VLAN patches

[Ben Greear]

Dave, please apply this patch if you haven't already,
it seems correct to me.

Andersson Björn wrote:
> Hi,
> I hope you are the right receiver of 8021q-patches.
> We are running SuSe 8.0, i.e kernel 2.4.18.
> 
> - When addding several egress-mappings we get stuck in an eternal
> loop if they end up in the same hash-bucket, see vlan_dev.c.path.

> ------------------------------------------------------------------------
> 
> --- linux-2.4.18.SuSE/net/8021q/vlan_dev.c.orig	Wed Mar 27 13:57:17 2002
> +++ linux-2.4.18.SuSE/net/8021q/vlan_dev.c	Wed Sep 18 13:19:52 2002
> @@ -570,6 +570,7 @@
>  					dev_put(dev);
>  					return 0;
>  				}
> +				mp = mp->next;
>  			}
>  
>  			/* Create a new mapping then. */


-- 
Ben Greear <greearb@candelatech.com>       <Ben_Greear AT excite.com>
President of Candela Technologies Inc      http://www.candelatech.com
ScryMUD:  http://scry.wanfear.com     http://scry.wanfear.com/~greear

Re: VLAN patches

[Ben Greear]

This patch looks good too, though the (vlan_id < 0) test
is redundant since vlan_id is an unsigned number.  For
clarity of code, I wouldn't mind if it stayed in though.

Andersson Björn wrote:
> Hi,
> I hope you are the right receiver of 8021q-patches.
> We are running SuSe 8.0, i.e kernel 2.4.18.

> - If we try to remove a vlan with VID 0, ifconfig stops working
> completly.
> We fixed it with vlan.c.patch.
> 
> 
> 
> 
> ------------------------------------------------------------------------
> 
> --- linux-2.4.18.SuSE/net/8021q/vlan.c.orig	Wed Mar 27 13:57:17 2002
> +++ linux-2.4.18.SuSE/net/8021q/vlan.c	Wed Sep 18 13:19:13 2002
> @@ -207,7 +207,7 @@
>  #endif
>  
>  	/* sanity check */
> -	if ((vlan_id >= VLAN_VID_MASK) || (vlan_id <= 0))
> +	if ((vlan_id >= VLAN_VID_MASK) || (vlan_id < 0))
>  		return -EINVAL;
>  
>  	spin_lock_bh(&vlan_group_lock);


-- 
Ben Greear <greearb@candelatech.com>       <Ben_Greear AT excite.com>
President of Candela Technologies Inc      http://www.candelatech.com
ScryMUD:  http://scry.wanfear.com     http://scry.wanfear.com/~greear

Re: VLAN patches

[David S. Miller]

   From: Ben Greear <greearb@candelatech.com>
   Date: Sat, 05 Oct 2002 12:27:15 -0700

   This patch looks good too, though the (vlan_id < 0) test
   is redundant since vlan_id is an unsigned number.  For
   clarity of code, I wouldn't mind if it stayed in though.
   
VLAN ID zero is illegal.  You should not use ifconfig
with a VID of zero.

Re: VLAN patches

[Ben Greear]

David S. Miller wrote:
>    From: Ben Greear <greearb@candelatech.com>
>    Date: Sat, 05 Oct 2002 12:27:15 -0700
> 
>    This patch looks good too, though the (vlan_id < 0) test
>    is redundant since vlan_id is an unsigned number.  For
>    clarity of code, I wouldn't mind if it stayed in though.
>    
> VLAN ID zero is illegal.  You should not use ifconfig
> with a VID of zero.
> 

As someone mentioned, vlan of 0 may be used to do a priority-only
type of VLAN.  I don't know how much this makes sense though...

If we do decide to restrict it, the right place to restrict is in
the creation clause, not the deletion code, as exists now.

-- 
Ben Greear <greearb@candelatech.com>       <Ben_Greear AT excite.com>
President of Candela Technologies Inc      http://www.candelatech.com
ScryMUD:  http://scry.wanfear.com     http://scry.wanfear.com/~greear

Re: VLAN patches

[David S. Miller]

   From: Ben Greear <greearb@candelatech.com>
   Date: Sat, 05 Oct 2002 21:31:45 -0700

   As someone mentioned, vlan of 0 may be used to do a priority-only
   type of VLAN.  I don't know how much this makes sense though...
   
   If we do decide to restrict it, the right place to restrict is in
   the creation clause, not the deletion code, as exists now.

I think we should, please submit a patch which denies it on
both config and delete.

My most recent reading of 802.1q made it very clear that VID 0
is special and should not be assigned to any interface.

Re: VLAN patches

[David S. Miller]

   From: Ben Greear <greearb@candelatech.com>
   Date: Sat, 05 Oct 2002 12:23:36 -0700

   Dave, please apply this patch if you haven't already,
   it seems correct to me.
   
Applied to both 2.4.x and 2.5.x sources, thanks.

Re: VLAN patches

[jamal]

On Sat, 5 Oct 2002, David S. Miller wrote:

> I think we should, please submit a patch which denies it on
> both config and delete.
>

A packet with VLANid 0 and an 802.1p tag > 0 is legal. I
think its known as a "priority tagged" packet (not 100% sure
about the term). Therefore VLANid 0 MUST be accepted and ability to send
it should be there.
BTW, what about VLANid 0xFFF?

cheers,
jamal

Re: VLAN patches

[David S. Miller]

   From: jamal <hadi@cyberus.ca>
   Date: Sun, 6 Oct 2002 18:47:50 -0400 (EDT)
   
   A packet with VLANid 0 and an 802.1p tag > 0 is legal. I
   think its known as a "priority tagged" packet (not 100% sure
   about the term). Therefore VLANid 0 MUST be accepted and ability to send
   it should be there.

Great, I stand corrected, please send me a patch which therefore
accepts VID 0 on create and destroy.

   BTW, what about VLANid 0xFFF?

Your guess is as good as mine ;-)

Re: VLAN patches

[Ben Greear]

David S. Miller wrote:
>    From: jamal <hadi@cyberus.ca>
>    Date: Sun, 6 Oct 2002 18:47:50 -0400 (EDT)
>    
>    A packet with VLANid 0 and an 802.1p tag > 0 is legal. I
>    think its known as a "priority tagged" packet (not 100% sure
>    about the term). Therefore VLANid 0 MUST be accepted and ability to send
>    it should be there.
> 
> Great, I stand corrected, please send me a patch which therefore
> accepts VID 0 on create and destroy.

It already accepts on create, you just need to add the patch that
was already sent to allow for delete.

> 
>    BTW, what about VLANid 0xFFF?

I think it is reserved, but my spec is 3 years old, and I don't know
where it is right now, so whatever it is now, let's not touch it :)

> 
> Your guess is as good as mine ;-)
> 


-- 
Ben Greear <greearb@candelatech.com>       <Ben_Greear AT excite.com>
President of Candela Technologies Inc      http://www.candelatech.com
ScryMUD:  http://scry.wanfear.com     http://scry.wanfear.com/~greear

Re: VLAN patches

[David S. Miller]

   From: Ben Greear <greearb@candelatech.com>
   Date: Sun, 06 Oct 2002 20:35:38 -0700
   
   It already accepts on create, you just need to add the patch that
   was already sent to allow for delete.

Done.

Re: VLAN patches

[David S. Miller]

   From: Eran Mann <emann@mrv.com>
   Date: Mon, 07 Oct 2002 09:09:38 +0200
   
   Regarding the 0 VID it is indeed used for priority-only frames. 
   Shouldn't it be supported by alowing the user to configure a priority 
   map for the ethernet device (rather than creating another user-visible 
   device?

I don't have any opinion about this, besides the fact that if
the device based interface we have now can do whatever the user
wants we should not add a new configuration mechanism just to provide
some alias that effectively accomplishes the same task.

Re: VLAN patches

[Eran Mann]

According to 802.1Q (which is BTW available freely from IEEE's site) 
0xFFF VID is "Reserved for implementation use. This VID value shall not 
be configured as a PVID, configured in any Filtering Database entry, 
used in any Management operation, or transmitted in a tag header.".

Regarding the 0 VID it is indeed used for priority-only frames. 
Shouldn't it be supported by alowing the user to configure a priority 
map for the ethernet device (rather than creating another user-visible 
device?

Ben Greear wrote:
> David S. Miller wrote:
>>    From: jamal 
>>       A packet with VLANid 0 and an 802.1p tag > 0 is legal. I
>>    think its known as a "priority tagged" packet (not 100% sure
>>    about the term). Therefore VLANid 0 MUST be accepted and ability to 
>>    send it should be there.
>>
>> Great, I stand corrected, please send me a patch which therefore
>> accepts VID 0 on create and destroy.
 >
> It already accepts on create, you just need to add the patch that
> was already sent to allow for delete.
> 
>>    BTW, what about VLANid 0xFFF?
> 
> I think it is reserved, but my spec is 3 years old, and I don't know
> where it is right now, so whatever it is now, let's not touch it :)

-- 
Eran Mann
Senior Software Engineer
MRV International
Tel: 972-4-9936297
Fax: 972-4-9890430
www.mrv.com

Re: VLAN patches

[Eran Mann]

David S. Miller wrote:
>    From: Eran Mann 
>    
>    Regarding the 0 VID it is indeed used for priority-only frames. 
>    Shouldn't it be supported by alowing the user to configure a priority 
>    map for the ethernet device (rather than creating another user-visible 
>    device?
> 
> I don't have any opinion about this, besides the fact that if
> the device based interface we have now can do whatever the user
> wants we should not add a new configuration mechanism just to provide
> some alias that effectively accomplishes the same task.
> 
I was thinking more in the line of:
# vconfig set_egress_map eth0 XX YY
etc...
(once CONFIG_VLAN_8021Q=y|m ofcourse)
-- 
Eran Mann
Senior Software Engineer
MRV International
Tel: 972-4-9936297
Fax: 972-4-9890430
www.mrv.com

Re: VLAN patches

[jamal]

On Mon, 7 Oct 2002, Eran Mann wrote:

> Regarding the 0 VID it is indeed used for priority-only frames.
> Shouldn't it be supported by alowing the user to configure a priority
> map for the ethernet device (rather than creating another user-visible
> device?
>

That or you complain and drop the packet when someone sends; btw
also VLANid > 0 and p tag 0 is legal - i guess that is the default
for linux.
On a side not:
I never liked Bens patches that much compared to the other VLAN
imp. -- would have prefered a marriage of the two.

cheers,
jamal

Re: VLAN patches

[David S. Miller]

   From: jamal <hadi@cyberus.ca>
   Date: Mon, 7 Oct 2002 08:05:07 -0400 (EDT)

   I never liked Bens patches that much compared to the other VLAN
   imp. -- would have prefered a marriage of the two.

There is a long thread deep in the archives for these lists
saying why Ben's code went it.

Please, let's not duplicate that conversation ok?

Re: VLAN patches

[Boro King]

please take me of the mailing list!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Re: VLAN patches

[Boro King]

how the fuck do i get of that mailing list ive been asking for weeks now 
fuck..  plz take me OFF!!!!!