min/max mtu field in struct net_device

min/max mtu field in struct net_device

[Arend Van Spriel]

Just wondering. Does anyone have any ideas on how to backport the two
patches below. The struct net_device now holds mtu range which network
subsystem checks. For a number of drivers and mac80211 it means they no
longer have .ndo_change_mtu callback. My guess is that we need patches
in backport of some sort to tackle this.

Regards,
Arend

commit 61e84623ace35ce48975e8f90bbbac7557c43d61
Author: Jarod Wilson <jarod@redhat.com>
Date:   Fri Oct 7 22:04:33 2016 -0400

    net: centralize net_device min/max MTU checking

commit 9c22b4a34eddbaa5b5243c8cd27e31aa36e676e1
Author: Jarod Wilson <jarod@redhat.com>
Date:   Thu Oct 20 13:55:18 2016 -0400

    net: use core MTU range checking in wireless drivers
--
To unsubscribe from this list: send the line "unsubscribe backports" in

Re: min/max mtu field in struct net_device

[Johannes Berg]

On Mon, 2016-11-14 at 09:49 +0100, Arend Van Spriel wrote:
> Just wondering. Does anyone have any ideas on how to backport the two
> patches below. The struct net_device now holds mtu range which
> network subsystem checks. For a number of drivers and mac80211 it
> means they no longer have .ndo_change_mtu callback. My guess is that
> we need patches in backport of some sort to tackle this.

Yeah, I kinda saw this coming. Since in almost all cases using the new
min/max the ndo_change_mtu was actually removed, perhaps we can come up
with a way to spatch it back in?

johannes
--
To unsubscribe from this list: send the line "unsubscribe backports" in

Re: min/max mtu field in struct net_device

[Arend Van Spriel]

[-- Attachment #1: Type: text/plain, Size: 849 bytes --]

On 14-11-2016 10:30, Johannes Berg wrote:
> On Mon, 2016-11-14 at 09:49 +0100, Arend Van Spriel wrote:
>> Just wondering. Does anyone have any ideas on how to backport the two
>> patches below. The struct net_device now holds mtu range which
>> network subsystem checks. For a number of drivers and mac80211 it
>> means they no longer have .ndo_change_mtu callback. My guess is that
>> we need patches in backport of some sort to tackle this.
> 
> Yeah, I kinda saw this coming. Since in almost all cases using the new
> min/max the ndo_change_mtu was actually removed, perhaps we can come up
> with a way to spatch it back in?

Me too. This is what I got so far, but the callback is generated twice
for net/mac80211/iface.c for obvious reasons (data and monitor ops).
These are my first baby steps with spatch so not sure how to solve it.

Gr. AvS

[-- Attachment #2: 0073-netdevice-mtu-range.cocci --]
[-- Type: text/plain, Size: 395 bytes --]

@r1@
expression ndev, e1, e2;
identifier func;
@@
func(...) {
	<...
- 	ndev->min_mtu = e1;
- 	ndev->max_mtu = e2;
	...>
}
@@
expression r1.e1, r1.e2;
identifier OPS;
@@
+ static int __change_mtu(struct net_device *ndev, int new_mtu)
+ {
+	if (new_mtu < e1 || new_mtu > e2)
+ 		return -EINVAL;
+	ndev->mtu = new_mtu;
+ }
+
struct net_device_ops OPS = {
+ 	.ndo_change_mtu = __change_mtu,
	...
};

Re: min/max mtu field in struct net_device

[Johannes Berg]

> Me too. This is what I got so far, but the callback is generated
> twice
> for net/mac80211/iface.c for obvious reasons (data and monitor ops).
> These are my first baby steps with spatch so not sure how to solve
> it.

Oh, right, that's problematic - no easy way to link the two and know
which one is which ...

I guess we'll have to carry that as a manual patch. We should probably
also ifdef on the kernel version.

johannes
--
To unsubscribe from this list: send the line "unsubscribe backports" in

Fwd: Re: min/max mtu field in struct net_device

[Arend Van Spriel]

[-- Attachment #1: Type: text/plain, Size: 1428 bytes --]

Hi Julia,

Would you have any ideas on how to solve the problem below. The problem
occurs with source files that have more than one struct net_device_ops
defined, which is the case for net/mac80211/iface.c for one. Obviously I
want the callback definition to be added only once.

Regards,
Arend

-------- Forwarded Message --------
Subject: Re: min/max mtu field in struct net_device
Date: Mon, 14 Nov 2016 11:46:01 +0100
From: Arend Van Spriel <arend.vanspriel@broadcom.com>
To: Johannes Berg <johannes@sipsolutions.net>, backports@vger.kernel.org
<backports@vger.kernel.org>

On 14-11-2016 10:30, Johannes Berg wrote:
> On Mon, 2016-11-14 at 09:49 +0100, Arend Van Spriel wrote:
>> Just wondering. Does anyone have any ideas on how to backport the two
>> patches below. The struct net_device now holds mtu range which
>> network subsystem checks. For a number of drivers and mac80211 it
>> means they no longer have .ndo_change_mtu callback. My guess is that
>> we need patches in backport of some sort to tackle this.
> 
> Yeah, I kinda saw this coming. Since in almost all cases using the new
> min/max the ndo_change_mtu was actually removed, perhaps we can come up
> with a way to spatch it back in?

Me too. This is what I got so far, but the callback is generated twice
for net/mac80211/iface.c for obvious reasons (data and monitor ops).
These are my first baby steps with spatch so not sure how to solve it.

Gr. AvS


[-- Attachment #2: 0073-netdevice-mtu-range.cocci --]
[-- Type: text/plain, Size: 395 bytes --]

@r1@
expression ndev, e1, e2;
identifier func;
@@
func(...) {
	<...
- 	ndev->min_mtu = e1;
- 	ndev->max_mtu = e2;
	...>
}
@@
expression r1.e1, r1.e2;
identifier OPS;
@@
+ static int __change_mtu(struct net_device *ndev, int new_mtu)
+ {
+	if (new_mtu < e1 || new_mtu > e2)
+ 		return -EINVAL;
+	ndev->mtu = new_mtu;
+ }
+
struct net_device_ops OPS = {
+ 	.ndo_change_mtu = __change_mtu,
	...
};