07-19-2019 01:26 PM
Dear Community,
I am currently studying for the CCNP-ROUTE exam in an attempt to pass it before the February 2020 deadline. I am currently studying section 1.3b that has to do with IPv4 and IPv6 fragmentation and I had a couple of questions:
1) When the originating host decides that it needs to fragment the IPv4 packet, how does the host itself know that a downstream MTU is smaller than some of the packet sizes and that the packets need to be fragmented.
2) If an IPv4 packet that is not fragmented needs to be fragmented in transit, or is fragmented already and needs to be fragmented again, can an intermediate router in the path perform the fragmentation? if so, under what circumstances would this occur?
I know these questions may be a little "in the weeds" for the exam but I always try to be as prepared as possible. I appreciate any feedback you can provide regarding the fragmentation process of IPv4.
Thanks!
Solved! Go to Solution.
07-19-2019 04:41 PM - edited 08-07-2019 11:12 AM
#1 It would send its IP packets out with the don't-fragment bit set. If, along the path, some IP hop cannot forward the MTU received, i.e. it needs to fragment, it wouldn't fragment (because of the bit setting) but instead send a IGMP message to the source IP noting fragmentation is needed (later versions will include in the message how large the MTU can be).
When the sender receives this message, it normally decrease the MTU to fit (for the destination), it doesn't generally send/generate fragments.
If there's no don't-fragment bit, the transit router will fragment the IP packet and the sending host "doesn't know".
#2 Yes. It's just like the first instance. If don't-fragment isn't set, the already too large fragmented packet will be fragmented again. If the don't-fragment is set, to get to the next smaller transit hop, the packet would not have a larger MTU then all the upstream hops support. (I.e. an earlier too large packet might have already caused the source to reduce its MTU.)
For example the source send 1500 bytes with DF set. Along the way a hop can only handle 1000, so the source re-sends the data using 1000 byte packets, still with DF set. Then further along he path another hop can only handle 750 bytes. So, again, the source re-sends the data using 750 bytes.
07-19-2019 02:21 PM
07-19-2019 04:41 PM - edited 08-07-2019 11:12 AM
#1 It would send its IP packets out with the don't-fragment bit set. If, along the path, some IP hop cannot forward the MTU received, i.e. it needs to fragment, it wouldn't fragment (because of the bit setting) but instead send a IGMP message to the source IP noting fragmentation is needed (later versions will include in the message how large the MTU can be).
When the sender receives this message, it normally decrease the MTU to fit (for the destination), it doesn't generally send/generate fragments.
If there's no don't-fragment bit, the transit router will fragment the IP packet and the sending host "doesn't know".
#2 Yes. It's just like the first instance. If don't-fragment isn't set, the already too large fragmented packet will be fragmented again. If the don't-fragment is set, to get to the next smaller transit hop, the packet would not have a larger MTU then all the upstream hops support. (I.e. an earlier too large packet might have already caused the source to reduce its MTU.)
For example the source send 1500 bytes with DF set. Along the way a hop can only handle 1000, so the source re-sends the data using 1000 byte packets, still with DF set. Then further along he path another hop can only handle 750 bytes. So, again, the source re-sends the data using 750 bytes.
08-06-2019 02:11 PM
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide