netdev
[Top] [All Lists]

Re: IPv6: the same address can be added multiple times

To: Daniel Roesen <dr@xxxxxxx>
Subject: Re: IPv6: the same address can be added multiple times
From: "David S. Miller" <davem@xxxxxxxxxx>
Date: Thu, 3 May 2001 17:30:33 -0700 (PDT)
Cc: Pekka Savola <pekkas@xxxxxxxxxx>, netdev@xxxxxxxxxxx, usagi-users@xxxxxxxxxxxxxx
In-reply-to: <20010503123033.A1378@xxxxxxxxxxxxxxxxxxxxxxxx>
References: <Pine.LNX.4.33.0105031202080.13012-100000@xxxxxxxxxx> <20010503123033.A1378@xxxxxxxxxxxxxxxxxxxxxxxx>
Sender: owner-netdev@xxxxxxxxxxx
Daniel Roesen writes:
 > On Thu, May 03, 2001 at 12:04:10PM +0300, Pekka Savola wrote:
 > > It looks like a check or two in kernel is missing,
 > 
 > Nod, if I understand net/ipv6/addrconf.c ipv6_add_addr() correctly,
 > the check is simply missing. It looks like it unconditionally adds
 > the address to the interface.

Please send me a patch for this when someone comes up with it.

I cannot see any logical reason for the current behavior.  Can
anyone else?  Maybe I could see some weird case where you'd be
able to assign different route characteristics to two instances
of the same interface address.  ... no it doesn't make any sense.

I have no opinion about matching the ifconfig behavior of BSD.
Maybe it would be nice to return no error in this case and silently
act as if it happened successfully, but you would need to make sure
all the other correctness checks were made before we returned
success.  That may be more work than it is worth.

Later,
David S. Miller
davem@xxxxxxxxxx

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