On 2/13/09 2:06 PM, "Steffen Maier" <ma...@linux.vnet.ibm.com> wrote:
> As far as I was told, certain versions of qeth device drivers required > the setting of a MAC address with real OSAs, because all virtual > machines would otherwise take the one MAC address of the OSA and would > end up talking with the same address to the outside no more being > distinguishable. Recent driver versions have code which generates a > unique MAC address when using real OSAs and there is no more need to > specify a MAC address. I faintly remember that being a "feature" of the layer 3-only OSA microcode though. Shouldn't that be something that can be determined and a suitable message issued? One problem remains, though. In environments with GuestLAN or VSWITCH, there is no need to specify a MAC address. How would a user specify the fact that he doesn't want to set the MAC address in the parmfile? I think I would argue that if you have a backlevel device like the situation above with the real OSA, that should be the exception case, not the rule that needs a special parm to circumvent - ie, if you have one of the "broken" devices, then you need the special parm. GLAN/VSWITCH devices already have MACs, so the normal logic should apply. The analagous case is how Intel distributions handle power management stuff like ACPI. They assume it works unless told otherwise with the "noacpi" kernel parm. It's up to you not to shoot yourself in the foot. ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390