[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ntp-hackers
Subject:    [ntp:hackers] Re: ntp-dev and getaddrinfo()
From:       "David L. Mills" <mills () udel ! edu>
Date:       2003-08-12 23:39:50
[Download RAW message or body]

Harlan,

When Dennis Fergusson wrote the original code in 1988, resolvers weren't
as reliable as now. All things considered, using an inline resolver and
avoiding the fork might not be such a bad idea. After about a minute of
DNS wait, make the configuration down and toss a message to the log. For
ephemeral associations, the same error path as now used for crypto
failure, including a new kiss code NDNS.

Doing this will weed out a lot of code in a lot of hard to find places,
plus making sure remote configuration continues to work.

Dave 

Harlan Stenn wrote:
> 
> It looks like the IPv6 changes have altered the config file processing
> in that instead of calling the code in ntp_intres.c to handle DNS/rDNS
> lookups thru a forked process we now call getaddrinfo() in the main
> code.
> 
> This means ntpd can block, which is a Bad Thing (last I checked).
> 
> I believe this must be fixed before we release 4.2.0.
> 
> Choices include:
> 
> - decide it's OK if ntpd blocks while making resolver calls
> - move the getaddrinfo() calls to ntp_intres.c and change the calling
>   code to use the forked resolver
> - implement the lwres library and go to an async resolver
> 
> Thoughts/comments?
> 
> Harlan
_______________________________________________
hackers mailing list
hackers@ntp.org
http://mailman.ntp.org/mailman/listinfo/hackers
[prev in list] [next in list] [prev in thread] [next in thread]