Bugzilla – Bug 1465
problems with grid configuration in a mesh simulation
Last modified: 2016-09-28 00:19:00 UTC
Description posted here: http://mailman.isi.edu/pipermail/ns-developers/2012-June/010479.html
The problem seems to be still here, see https://groups.google.com/forum/#!topic/ns-3-users/q_ktE2vZdds Moreover, I double checked and on my computer the network *seems* to be formed, but no packets are actually exchanged. Checking the cap files, the ARP request arrives at the intended node, but it doesn't reply.
(In reply to Tommaso Pecorella from comment #1) > The problem seems to be still here, see > https://groups.google.com/forum/#!topic/ns-3-users/q_ktE2vZdds > > Moreover, I double checked and on my computer the network *seems* to be > formed, but no packets are actually exchanged. > Checking the cap files, the ARP request arrives at the intended node, but it > doesn't reply. I think it is due to the problem reported in bug 737, which was marked invalid a while back but which I think probably ought to be reopened and fixed based on Kirill's patch there or something similar. The problem is that the relaying nodes do not defer and backoff; they immediately forward, causing a repeatable ARP collision on a grid topology. The ARP eventually fails.
(In reply to Tom Henderson from comment #2) > (In reply to Tommaso Pecorella from comment #1) > > The problem seems to be still here, see > > https://groups.google.com/forum/#!topic/ns-3-users/q_ktE2vZdds > > > > Moreover, I double checked and on my computer the network *seems* to be > > formed, but no packets are actually exchanged. > > Checking the cap files, the ARP request arrives at the intended node, but it > > doesn't reply. > > > I think it is due to the problem reported in bug 737, which was marked > invalid a while back but which I think probably ought to be reopened and > fixed based on Kirill's patch there or something similar. > > The problem is that the relaying nodes do not defer and backoff; they > immediately forward, causing a repeatable ARP collision on a grid topology. > The ARP eventually fails. Sadly it's not that easy. ARP request is received correctly, but ARP reply can't be delivered because the node can't find a route to the originator. Basically it's a bug in HWMP.
Another interesting fact. Beacons gave a duration of 60 us. Path req. have ZERO duration
I must correct myself. It *is* a collision issue, but it's not ARP. The colliding packets are path requests. I Still have no idea how to fix it for good. I tried adding some jitter but no luck. As a side note, HWMP uses random variables but no stream assignments(if I'm right). (In reply to Tommaso Pecorella from comment #3) > (In reply to Tom Henderson from comment #2) > > (In reply to Tommaso Pecorella from comment #1) > > > The problem seems to be still here, see > > > https://groups.google.com/forum/#!topic/ns-3-users/q_ktE2vZdds > > > > > > Moreover, I double checked and on my computer the network *seems* to be > > > formed, but no packets are actually exchanged. > > > Checking the cap files, the ARP request arrives at the intended node, but it > > > doesn't reply. > > > > > > I think it is due to the problem reported in bug 737, which was marked > > invalid a while back but which I think probably ought to be reopened and > > fixed based on Kirill's patch there or something similar. > > > > The problem is that the relaying nodes do not defer and backoff; they > > immediately forward, causing a repeatable ARP collision on a grid topology. > > The ARP eventually fails. > > Sadly it's not that easy. ARP request is received correctly, but ARP reply > can't be delivered because the node can't find a route to the originator. > Basically it's a bug in HWMP. (In reply to Tom Henderson from comment #2) > (In reply to Tommaso Pecorella from comment #1) > > The problem seems to be still here, see > > https://groups.google.com/forum/#!topic/ns-3-users/q_ktE2vZdds > > > > Moreover, I double checked and on my computer the network *seems* to be > > formed, but no packets are actually exchanged. > > Checking the cap files, the ARP request arrives at the intended node, but it > > doesn't reply. > > > I think it is due to the problem reported in bug 737, which was marked > invalid a while back but which I think probably ought to be reopened and > fixed based on Kirill's patch there or something similar. > > The problem is that the relaying nodes do not defer and backoff; they > immediately forward, causing a repeatable ARP collision on a grid topology. > The ARP eventually fails.
the patch to bug 2369 (changeset 12345:a94d790ef6e5) changed the behavior so that backoff occurs if the request for access arrives during AIFS interval, but allows immediate access if the medium has been idle for time >= AIFS. See discussion and documentation in bug 2369. This behavior solves the grid configuration issue (because backoff will now be invoked upon relaying a frame), so am marking this as resolved fixed.
*** Bug 1605 has been marked as a duplicate of this bug. ***