[j-nsp] Re: ICMPv6 "packet too big" not being emitted

[Hannes Gredler]

Daniel Roesen wrote:
> On Tue, Jan 10, 2006 at 07:46:24PM +0100, Daniel Roesen wrote:
> 
> The problematic router's backbone uplinks are indeed unnumbered ATM
> PVCs, unlike the other routers I've tested which have POS/GigE uplinks.
> The "mtu exceeded" counter of "show pfe statistics ip6 icmp" is
> confirmed to NOT increase when receiving packets which are too large for
> the tunnel.
> 
> Juniper: I'm pretty sure you have a serious problem with handling
> packets which would need fragmentation, being received on unnumbered
> links, in IPv4 and IPv6. Would you eventually care to open a PR about
> that this time, even if not submitted via sales channels or support
> contract? Should be in your best interest to get it fixed. :-)

daniel,

have opened PR67314
tx for the report,

/hannes

---

 >Number:         67314
 >Category:       sw-kern-net
 >Synopsis:       generation of ICMP frag-needed but DF set broken if packet enters on unnumbered link
 >Severity:       minor
 >Priority:       medium
 >Responsible:    bug-sw-kern-net
 >State:          open
 >Class:          sw-bug
 >Target:
 >Keywords:       ICMP DF unnumbered
 >Blocking-Release:
 >Submitter-Id:   jtac
 >Arrival-Date:   Wed Jan 11 03:56:15 -0800 2006
 >Closed-Date:
 >Last-Modified:
 >Originator:     Hannes Gredler
 >Release:        7.5-20060111.0