Re: TCP checksum and device driver eri0

From: Casper Dik (Casper.Dik_at_Sun.COM)
Date: 06/29/03

  • Next message: Scott Howard: "Re: TCP checksum and device driver eri0" 
    To: Giovanni Porcelli <porcelli@unipmn.it>
Date: Sun, 29 Jun 2003 10:58:31 +0200

    >During a snoop session on a 280R with solaris 9 and network interface
    >eri0, I noticed that the TCP checksum field of every outgoing packet was
    >incorrect.

    The eri0 interface and some other Solaris interfaces employ
    hardware checksumming; as a result, packets with correct checksums
    never exist in the kernel, they're generated in the hardware.

    >The same packet snooped on the network has the correct checksum value.
    >I believe that something happens between the snoop point on the machine
    >and the network interface eri0 output.
    >I also tried on other machines: on a Sun 450R with solaris 9 (but with
    >hme0) the TCP checksum snooped locally is correct, also on a Sun 250
    >with solaris 8 and hme0 is correct.

    No hardware checksum for hme.

    >With this last result I believe that the responsible is the eri0 driver.
    >It is true?
    >If I want to snoop locally *but* with the correct checksum value, how
    >can I do?

    You can disable hardware checksums:

            set ip:dohwcksum = 0

    in /etc/system

    Casper

  • Next message: Scott Howard: "Re: TCP checksum and device driver eri0"