On Friday, 01/05/2007 at 10:23 CST, Tom Duerbusch
<[EMAIL PROTECTED]> wrote:
> I don't see why you would have had the problem in the first place.  I've
been
> connecting SLES9 (31 and 64 bit) to a vswitch under both z/VM 5.1 and
z/VM 5.2
> without any problem.

There are timing issues with the VSWITCH on z/VM 5.1 that were addressed
in z/VM 5.2.  In some circumstances on z/VM 5.1 you will see CP issue
HCP2833E with code E005 (address already registered on the OSA) when it is
already owned by *this* guest.  A factoid that, while true, doesn't hurt
anything.

When the subnet mask on an interface is changed in Linux, the entire
interface is brought down and back up (this is all handled by the
platform-independent layers).  That means unregistering and re-registering
the IP address in the OSA.  The speed at which Linux is able to perform
these operations relative to the responsiveness of both the OSA and the
system itself affect the probability of getting the message.  In z/VM 5.2
we redesigned CP to serialize IP [un]registration on the OSA to avoid the
entire mess.

Now, if your chosen subnet mask is the same as default class mask based on
the IP address, then Linux is smart enough not to bother changing the
netmask and, so, the problem never arises.  For example, those of you with
class C 192.168.what.ever addresses with masks of 255.255.255.0 won't see
it.  OTOH, if you have a class A 10-dot with subnet mask of 255.255.255.0,
then you have the potential to hit this condition.

It is unfortunate that Linux manages the adapters and routing tables as a
single widget.  If you want a widget with different attributes, delete the
old one and add a new one with the new attributes.  Even on Intel this
model opens a window where someone can sneak in and legitimately take your
IP address just because you have a non-default subnet mask or you manually
change the subnet mask with ifconfig.  The Law of Unintended Consequences
applies.

Sorry to give a long-winded answer, but I just finished researching the
reason for this rather elusive problem and wanted to get it out into the
Public Domain.

So, if you get HCP2833E E005 on a VSWITCH, don't panic.  Your connection
is probably still good ("But I *know* no one else is using this IP
addy!").

Alan Altmark
z/VM Development
IBM Endicott

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to