Bug #2312: CHILD_INHERIT does not work when run nfdc register - NFD - NDN project issue tracking system

Actions

Copy link

Bug #2312

closed

CHILD_INHERIT does not work when run nfdc register

Added by Yanbiao Li almost 10 years ago. Updated almost 10 years ago.

Status:

Closed

Priority:

High

Assignee:

Vince Lehman

Category:

RIB

Target version:

v0.3

Start date:

12/19/2014

Due date:

% Done:

90%

Estimated time:

Description

A route with prefix /localhop/nfd/rib is already exists. (/localhop/nfd/rib nexthops={faceid=259 (cost=0)})

After register a prefix /localhop/nfd (using nfdc register), a new route is generated (/localhop/nfd nexthops={faceid=262 (cost=0)}), in which CHILD_INHERIT is default set.

However, the previous route doesn't inherit the face of the second route.

here follows the full screenshot:

LiYanBiaos-MacBook-Pro:NFD lybmath$ nfd-start
LiYanBiaos-MacBook-Pro:NFD lybmath$ nfd-status -b
FIB:
  /localhost/nfd nexthops={faceid=1 (cost=0)}
  /localhop/nfd/rib nexthops={faceid=259 (cost=0)}
  /localhost/nfd/rib nexthops={faceid=259 (cost=0)}
LiYanBiaos-MacBook-Pro:NFD lybmath$ nfdc register /localhop/nfd udp://spurs.cs.ucla.edu
Successful in name registration: ControlParameters(Name: /localhop/nfd, FaceId: 262, Origin: 255, Cost: 0, Flags: 1, )
LiYanBiaos-MacBook-Pro:NFD lybmath$ nfd-status -b
FIB:
  /localhost/nfd nexthops={faceid=1 (cost=0)}
  /localhop/nfd/rib nexthops={faceid=259 (cost=0)}
  /localhop/nfd nexthops={faceid=262 (cost=0)}
  /localhost/nfd/rib nexthops={faceid=259 (cost=0)}
LiYanBiaos-MacBook-Pro:NFD lybmath$

Related issues 1 (1 open — 0 closed)

Actions

Copy link

Updated by Alex Afanasyev almost 10 years ago

I think the only problem is that /localhop/nfd prefix is registered by NRD directly to FIB. I believe NRD should register it in its own FIB, so the child inherit flag can take effect.

Actions

Copy link

Updated by Alex Afanasyev almost 10 years ago

Description updated (diff)
Assignee set to Yanbiao Li
Target version set to v0.3

Actions

Copy link

Updated by Alex Afanasyev almost 10 years ago

Assignee changed from Yanbiao Li to Vince Lehman

Actions

Copy link

Updated by Alex Afanasyev almost 10 years ago

Priority changed from Normal to High

Actions

Copy link

Updated by Alex Afanasyev almost 10 years ago

I'm raising the priority as it is relatively critical issue that needs to be resolved asap.

However, I'm not quite sure about how exactly this should be solved. Options:

hard-code that when NRD starts it adds RIB entry for /localhost/nfd, /localhost/nfd/rib, and /localhop/nfd
when NRD starts, read current NFD FaceDataset and add all the existing routes to RIB.

I prefer the second approach, though it is more complicated.

Actions

Copy link

Updated by Vince Lehman almost 10 years ago

Status changed from New to In Progress

So for the second approach:

Request the FIB DataSet
When the full DataSet is received, enumerate each FibEntry
enumerate each NextHopRecord in FibEntry
Register a route in the RIB with FibEntry.name, NextHopRecord.FaceId, and NextHopRecord.Cost

Is this correct? What route flags should each entry have, if any?

Actions

Copy link

Updated by Junxiao Shi almost 10 years ago

I disagree with the second approach in note-5. It's unnecessarily complex, causes additional delay, and may cause inconsistency.

To solve this particular Bug, RIB daemon just need to insert these Routes (hard coded), or even just the last Route:

Name	FaceId	Origin	ChildInherit	Capture
/localhost/nfd	internal	app	no	yes
/localhost/nfd/rib	nrd	app	yes	no
/localhop/nfd/rib	nrd	app	yes	yes

Actions

Copy link

Updated by Alex Afanasyev almost 10 years ago

Hmm. I think both options are actually wrong and unnecessary.

The only real problem is with "localhop" prefix, which is actually registered by NRD itself. I think, right now it uses FIB management protocol directly. Instead, it should add record to RIB and use FIB updater to add actual record to NFD's FIB.

Actions

Copy link

Updated by Junxiao Shi almost 10 years ago

I agree with note-8.

This Bug needs to wait after #1941. I won't agree with any hack.

Actions

Copy link

#10

Updated by Alex Afanasyev almost 10 years ago

This issue is (partially) independent of #1941 and does not require any hacking. It is just instead of calling registerPrefix, it should call something internal and then the overload setInterestFilter.

I could be ok waiting, if everything in #1941 can be finished this week. If there are more things, then I would push for some fix asap.

Actions

Copy link

#11

Updated by Vince Lehman almost 10 years ago

Status changed from In Progress to Code review
% Done changed from 0 to 90

I think finishing #1941 this week would be very optimistic; I have pushed a patch to Gerrit which uses the current codebase to do what note-8 suggests.

Actions

Copy link

#12