linux-mips-fnet
[Top] [All Lists]

Re: udp checksum

To: Rabeeh Khoury <rabeeh@galileo.co.il>
Subject: Re: udp checksum
From: Ralf Baechle <ralf@uni-koblenz.de>
Date: Sat, 29 Jan 2000 03:15:29 +0100
Cc: "linux-mips@fnet.fr" <linux-mips@fnet.fr>
In-reply-to: <38919BAA.6C7EBE80@galileo.co.il>
References: <3890DC5A.725151A1@galileo.co.il> <20000128010842.D11868@uni-koblenz.de> <38919BAA.6C7EBE80@galileo.co.il>
On Fri, Jan 28, 2000 at 03:37:46PM +0200, Rabeeh Khoury wrote:

> The Ethernet port is part of our new chip - GT96100 which is System
> controller + 2 ethernet ports + 8 Serial ports (which can be configured
> as framerelay ports , ISDN and many more protocols) + 2 PCI and other more.
> 
> so the ethernet port is an embedded port in the chip. as i have noted ; i
> managed to ramp up NFS over the system with full functionality. meaning
> that the packets are sent and received correct but the crc checksum is
> wrong in the udp (it uses functions included from include/asm-mips/checksum)

CRC is used to protect the Ethernet frame, a one's complement sum is used
to protect the IP packets.  I assume you mean the latter, don't you?

> could it be that the error in checksum calculations is that i'm using big
> mips configuration ?

Improbable.  The code is definately working fine for big endian machines.
I'm running it on various big endian machines, among them a diskless
machine with good results.

Have you verified that the packet is actually being sent correct except
the bad checksum?  At times it can be very tricky to get the cache
coherence thing right and the sympthoms with network drivers tend to
be things like corrupted packets.

  Ralf

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