[clue-tech] Linux pppd with lcp-echo-failure option doesn't detect dead peer?

[David L. Anselmi]

Jim Ockers wrote:
> Hi CLUEbies,
>
> I've got a weird one. We run pppd on Linux with generally very good
> success. We have a 15 second lcp-echo-interval and 3 lcp-echo-failure to
> detect a dead peer. I ran into a situation where the peer was dead for
> several hours but pppd was running normally and sending out its lcp
> echo-requests and not getting any replies, but it did not die!

So you're saying that lcp-echo is for detecting connection failures and that pppd should die when 
they happen.  Is that the documented behavior or more an assumption that might be mistaken?  Not 
doubting you, just asking whether you've considered your assumptions.

> I did some googling and I think I am the only person in the world that
> this has ever happened to. I tried looking to see what bugs were fixed
> in pppd 2.4.5 but I don't have good bug tracking mojo I guess. Does this
> seem like it might be a bug in pppd?

Given that it works as you expect I'd say it's a bug in pppd.  Probably need to hook up gdb and see 
where the internal state is getting wonked.

Of course, if it never happens again perhaps it was just spurious cosmic ray muons.  Something like 
the lcp-echo-failure register had its high bit flipped.

I think it would be worth filing, just to document that it happened.  Don't suppose you have a core, 
log entries, or a packet trace to support your story?

Dave