netdev
[Top] [All Lists]

Re: skb_checksum_help

To: Thomas Graf <tgraf@xxxxxxx>
Subject: Re: skb_checksum_help
From: "David S. Miller" <davem@xxxxxxxxxxxxx>
Date: Mon, 24 Jan 2005 18:03:54 -0800
Cc: herbert@xxxxxxxxxxxxxxxxxxx, david@xxxxxxxxxxxxxxxx, kaber@xxxxxxxxx, netdev@xxxxxxxxxxx
In-reply-to: <20050125020118.GC31837@xxxxxxxxxxxxxx>
References: <20050123202715.281ac87c.davem@xxxxxxxxxxxxx> <20050124121610.GP23931@xxxxxxxxxxxxxx> <41F50B6C.6010107@xxxxxxxxxxxxxxxx> <20050124151510.GV23931@xxxxxxxxxxxxxx> <20050124225423.GA15405@xxxxxxxxxxxxxxxxxxx> <20050124234515.GA31837@xxxxxxxxxxxxxx> <20050125000759.GA15883@xxxxxxxxxxxxxxxxxxx> <20050124164049.3b939791.davem@xxxxxxxxxxxxx> <20050125014538.GB31837@xxxxxxxxxxxxxx> <20050125014838.GA16637@xxxxxxxxxxxxxxxxxxx> <20050125020118.GC31837@xxxxxxxxxxxxxx>
Sender: netdev-bounce@xxxxxxxxxxx
On Tue, 25 Jan 2005 03:01:18 +0100
Thomas Graf <tgraf@xxxxxxx> wrote:

> * Herbert Xu <20050125014838.GA16637@xxxxxxxxxxxxxxxxxxx> 2005-01-25 12:48
> > On Tue, Jan 25, 2005 at 02:45:38AM +0100, Thomas Graf wrote:
> > > 
> > > Still, one problem remains, even if ip_summed was incorrect,
> > > skb_checksum_help should never be called in dev_queue_xmit for the
> > > acenic driver since it sets NETIF_F_IP_CSUM. For some reason, features
> > > is set to 0 while checking for it.
> > 
> > Didn't he say that he cleared the features bits in the acenic driver?
> 
> He changed it back to the original version, I just checked with him.
> 
> > BTW, is he using acenic for both the LAN and the WAN? If he's only
> > using it on the WAN side, then the LAN side might not be CSUM capable.
> 
> The packets get routed back over the same physical interface.

I think I see what is happening, it's the virtual VLAN device which
has dev->features set to zero not the acenic's netdev struct.


<Prev in Thread] Current Thread [Next in Thread>