[PATCH 001/001] smsc95xx: Fix hard_header_len

This device may require up-to 12 bytes of headroom. Assign the needed_headroom
but stop lying about hard_header_len which is used in other places throughout
the kernel.

Without this patch it is not possible to use "IFB + tc filter ... action ... 
dev ifb0" as per
http://www.linuxfoundation.org/collaborate/workgroups/networking/ifb.

Tested using Raspberry Pi as a router with trafficshaping and IFB.
---
 drivers/net/usb/smsc95xx.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/usb/smsc95xx.c b/drivers/net/usb/smsc95xx.c
index 0aee6c0..02293dd 100644
--- a/drivers/net/usb/smsc95xx.c
+++ b/drivers/net/usb/smsc95xx.c

@@ -1071,8 +1071,8 @@ static int smsc95xx_bind(struct usbnet *dev, struct usb_interface *intf)
 	dev->net->netdev_ops = &smsc95xx_netdev_ops;
 	dev->net->ethtool_ops = &smsc95xx_ethtool_ops;
 	dev->net->flags |= IFF_MULTICAST;
-	dev->net->hard_header_len += SMSC95XX_TX_OVERHEAD_CSUM;
-	dev->hard_mtu = dev->net->mtu + dev->net->hard_header_len;
+	dev->net->needed_headroom = SMSC95XX_TX_OVERHEAD_CSUM;
+	dev->hard_mtu = dev->net->mtu + dev->net->hard_header_len + dev->net->needed_headroom;
 	return 0;
 }
 

-- 
1.7.4.1.433.gcd306

Hi James,

Thanks for spotting this, I don't think I've ever tested this device with IMQ.

-       dev->net->hard_header_len += SMSC95XX_TX_OVERHEAD_CSUM;
-       dev->hard_mtu = dev->net->mtu + dev->net->hard_header_len;
+       dev->net->needed_headroom = SMSC95XX_TX_OVERHEAD_CSUM;
+       dev->hard_mtu = dev->net->mtu + dev->net->hard_header_len +
dev->net->needed_headroom;

I'm unsure what we *should* be setting hard_header_len and
needed_headroom to.  David?

Thanks,
--
Steve Glendinning

On Thu, 2012-08-16 at 10:57 +0100, Steve Glendinning wrote:

Hi James,

Thanks for spotting this, I don't think I've ever tested this device with IMQ.

quoted

-       dev->net->hard_header_len += SMSC95XX_TX_OVERHEAD_CSUM;
-       dev->hard_mtu = dev->net->mtu + dev->net->hard_header_len;
+       dev->net->needed_headroom = SMSC95XX_TX_OVERHEAD_CSUM;
+       dev->hard_mtu = dev->net->mtu + dev->net->hard_header_len +
dev->net->needed_headroom;

I'm unsure what we *should* be setting hard_header_len and
needed_headroom to.  David?

hard_header_len is set to ETH_HLEN by alloc_etherdev() (in the
ether_setup() callback).  Any extra headroom you want before the
Ethernet header should indeed be specified in needed_headroom.

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

hard_header_len is set to ETH_HLEN by alloc_etherdev() (in the
ether_setup() callback).  Any extra headroom you want before the
Ethernet header should indeed be specified in needed_headroom.

Thanks Ben, looks like this patch is doing it the right way then.

Acked-By: Steve Glendinning <steve.glendinning@shawell.net>


-- 
Steve Glendinning

From: Steve Glendinning <redacted>
Date: Fri, 17 Aug 2012 09:35:04 +0100

quoted

hard_header_len is set to ETH_HLEN by alloc_etherdev() (in the
ether_setup() callback).  Any extra headroom you want before the
Ethernet header should indeed be specified in needed_headroom.

Thanks Ben, looks like this patch is doing it the right way then.

Acked-By: Steve Glendinning <steve.glendinning@shawell.net>

This patch needs to be submitted with a proper signoff.

quoted

Acked-By: Steve Glendinning <steve.glendinning@shawell.net>

This patch needs to be submitted with a proper signoff.

I think Steve has been in "the delivery path" so he could sign it off,
I took his Acked-By as saying the patch was okay. In case it helps ...

Signed-off-by: James Betts <redacted>

Thanks.

From: James Betts <redacted>
Date: Tue, 21 Aug 2012 06:33:07 +0000

quoted

quoted

Acked-By: Steve Glendinning <steve.glendinning@shawell.net>

This patch needs to be submitted with a proper signoff.

I think Steve has been in "the delivery path" so he could sign it off,
I took his Acked-By as saying the patch was okay. In case it helps ...

Signed-off-by: James Betts <redacted>

The original author needs to properly freshly submit the patch
with their own signoff included in the commit message.