Subject: Re: bge(4) tcp4csum problem -- driver problem or chip problem???
To: Jason Thorpe <>
From: Ignatios Souvatzis <>
List: tech-kern
Date: 12/19/2006 21:19:05
On Tue, Dec 19, 2006 at 09:23:47AM -0800, Jason Thorpe wrote:
> On Dec 19, 2006, at 9:01 AM, Brian Buhrow wrote:
> >	Hello.  During my tests, the only variable I changed was the tcp4csum
> >value with ifconfig on the bge(4) interface.  Ip4csum and udp4csum  
> >were
> >enabled during all tests, and remain so on this production machine.  I
> >didn't think to try disabling ip4csum while leaving tcp4csum  
> >enabled.  Are
> >these settings not hierarchical?  That is, if you enable tcp4csum,  
> >are you
> >not, by implication, also enableing ip4csum checking?
> No.  The IP header checksum and the TCP header+payload checksum are  
> independent.

Ok. _I_, otoh, tested only tcp4csum without ip4csum; I'll do the other
test tomorrow if I can.

But anyway - my test was for a ftp get operation; I fetched a couple of
megabytes without seeing tcp checksum errors increase.

seal your e-mail: