[6lo] Router reboot - loss of state

["Pascal Thubert (pthubert)" <[e-mail]>]

Dear all :

There's one thing that was left unspecified in the RFC chain from RFC 6775, 8505, and 9010.  That's the case where the 6LR reboots and the 6LN is not aware of the event, maybe it was sleeping. In that case the 6LR forgets the registration and the 6LN might become unreachable till it reregisters. A router that knows the event will happen  goes could send a final RA but the 6LN might not hear it either, so the result is not deterministic. Anyway that does not cover the unintended reboot.

Usually the L2 detects a loss of association or something, that triggers the 6LN to reparent.  But that is not guaranteed to be available in all networks.
RPL has a method, the DTSN in the DIO (https://datatracker.ietf.org/doc/html/rfc6550#section-6.3.1<https://jpn01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Frfc6550%23section-6.3.1&data=05%7C01%7Cklaus.hueske%40renesas.com%7Cc6095f0e524742e74d5108da2de3ab9f%7C53d82571da1947e49cb4625a166a4a2a%7C0%7C1%7C637872753125005927%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=5PLxb2Vqb3sMtbjFx7onAM8eT1CP1Uqhf3Cc%2FmnpAEI%3D&reserved=0>). A new sequence indicates that the child that sees it needs to send its state, DAO in this case. The child will eventually see a DIO, and when it sees it, the child will know that the sequence was incremented. Though the text in RFC 6550 does not list all the cases when that is useful, a reboot in storing mode is certainly one.

But this only requires resending  DAO messages and has no effect on ND. https://datatracker.ietf.org/doc/html/draft-chakrabarti-nordmark-6man-efficient-nd-07#section-6.3<https://jpn01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Fdraft-chakrabarti-nordmark-6man-efficient-nd-07%23section-6.3&data=05%7C01%7Cklaus.hueske%40renesas.com%7Cc6095f0e524742e74d5108da2de3ab9f%7C53d82571da1947e49cb4625a166a4a2a%7C0%7C1%7C637872753125005927%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=7L53ZM1wt%2BMbYAZEoMPssWOBXiDpi9XlNjGBCx%2Bu2iY%3D&reserved=0> has the same operation with a new RA option, the RAO, which also has a sequence counter, the router epoch; but the draft was stalled at 6MAN and the function is still missing. My suggestion is to fix that gap sooner than later.

The fast path is to integrate the option in the multicast draft. The slow path is to make yet another RFC. What would you guys prefer?

Keep safe

Pascal