SSH login takes very long time...sometimes

[Rostislav Krasny]

On Thu, 23 Feb 2006 02:08:17 +0900
Hajimu UMEMOTO <ume at freebsd.org> wrote:

> Hi,
> 
> >>>>> On Wed, 22 Feb 2006 02:44:30 +0200
> >>>>> Rostislav Krasny <rosti.bsd at gmail.com> said:
> 
> rosti> On Tue, 21 Feb 2006 19:59:59 +0300
> rosti> Yar Tikhiy <yar at comp.chem.msu.su> wrote:
> 
> rosti> I forgot that a "search" resolver(5) parameter is useless for reverse
> rosti> resolving. But that "doubling" of name->IP requests with an empty (or
> rosti> root, according to resolver(5)) domain in the "search" is still a bug,
> rosti> IMHO. Although it shouldn't affect the sshd.
> 
> I looked BIND9's resolver, and took the related part into our
> resolver.  However, it seems to me that there is still same issue in
> BIND9's resolver.  So, I change more bit.  Please try the following
> patch and let me know the result:

Your patch fixed the problem, thank you. But during my tests I've found
another form of "doubling" bug in getaddrinfo(). To test the getaddrinfo()
behavior I used a program that is attached to this email.
If hints.ai_socktype == 0 then the getaddrinfo() does the lookup twice,
even if DNS is reachable. If the latter is the case, returned linked
list is twice as large than it should be. The hints.ai_socktype seems
to have the same influence when hints.ai_family is PF_INET6 or
PF_UNSPEC too.