[vortex] How to change the UDP data area in the 3c59x.c driver level ? Solved.

[Marcus Ramos]

Well, never mind about this. I discovered that a second, different UDP
segment was travelling through the driver and my code was unintentionally
changing its contents, thus causing the problem. Now that I filtered for
the correct UDP segment only, my code works perfectly well.

Marcus.

Marcus Ramos wrote:

> Hello,
>
> I am working on the development of an application that needs to
> timestamp UDP packets right before they are put on the wire, that means,
> on the driver level. I've modified the 3c59x.c source code at function
> boomerang_xmit_start() in order to simulate the insertion of a timestamp
> in the data area of the UDP packet. This means I am trying to change the
> contents of (skb->data)+42 to (skb->data)+53, corresponding to the 12
> bytes of data that follow the Ethernet/IP/UDP headers in the packet to
> be transmitted.
>
> However, doing so causes my application to behave completely strange, as
> if I were not allowed to change the contents of the UDP data segment. Of
> course I set (skb->data)+40 and (skb->data)+41 to zeroes, as to inform
> that no checksum will be used (at the least for the time being). That
> "should" prevent the transmission routines to misbehave because the
> previously calculated UDP checksum wouldn't match anymore the new
> contents of the UDP data area.
>
> Writing zeroes to the UDP checksum area does not cause any problem.
> Writing anything to the data area of the UDP segment causes big problem.
> Will somebody have an idea of what might be happening ? I would be most
> grateful for any hints or comments.
>
> Thanks in advance,
> Marcus.
>
> _______________________________________________
> vortex mailing list
> vortex@scyld.com
> http://www.scyld.com/mailman/listinfo/vortex