openssh unix dev - Jul 2003 - sshd takes 15 minutes to start

Hi all,

We are running OpenSSH v3.4p1 on three ES340 Alpha which run Tru64 v5.1A.  
The last couple of weeks, the system hangs at bootup for 15 minutes at the
point where it is bring up sshd.  When I manually stop and start sshd, it still
takes 15 minutes.  I manually ran sshd with the -ddd level 3 debug option.
 sshd hanges at the message which states it is creating a randomized seed.  It
also tkaes 15 minutes to run ssh_keygen.

The only tings we have done lately is to upgrad the ES40's firmware in
preporation to going to Tru64 5.1B.  We have also added gigabit NICs to set up
a private netowrk for use of NIS, NFS, and other intranetwork traffic.

I have tried compiling OpenSSH v3.6.1p2 with the same results.  OpenSSH was
compiled with GCC v3.3.  We are also running OpenSSL v0.6.9g and zlib v1.4.
 Does anyone here have any ideas what could be going on?

Any info will be appreciated.

Thanks,
Steven
-- 
                   ___     ____________
              <<<((__O\   (__<>___<>__ \   ____    
Don't get rattled by
Steven Whatley       \ \_(__<>___<>__)\O\_/O___>-<  what I
say.  It's just
swhatley at hal-pc.org   \O__<>___<>___<>)\___/        my
opinion.

swhatley at hal-pc.org wrote:
> We are running OpenSSH v3.4p1 on three ES340 Alpha which run Tru64 v5.1A.
> The last couple of weeks, the system hangs at bootup for 15 minutes at the
> point where it is bring up sshd.  When I manually stop and start sshd, it
still
> takes 15 minutes.  I manually ran sshd with the -ddd level 3 debug option.
>  sshd hanges at the message which states it is creating a randomized seed. 
It
> also tkaes 15 minutes to run ssh_keygen.
It sounds like one of the processes run by ssh-rand-helper is now hanging
for some reason.  This will most likely be one of the binaries replaced in
your upgrade.

The random helper is supposed to time out any command that takes too long
(>100 ms from memory) but there was a problem in earlier versions of
OpenSSH on some platforms where that did not occur.

To identify the command that's hanging, run ssh-rand-helper on its own
with a "-vv" argument.  You will be able to see each command as
it's run
and identify the one that's hanging.  Once you've found it (or them),
comment those commands out of ssh_prngs_cmds.

Upgrading sshd will also resolve the problem as the timeout bug has been
fixed.

-- 
Darren Tucker (dtucker at zip.com.au)
GPG key 8FF4FA69 / D9A3 86E9 7EEE AF4B B2D4  37C9 C982 80C7 8FF4 FA69
    Good judgement comes with experience. Unfortunately, the experience
usually comes from bad judgement.

Steven Whatley wrote:
[snip detail]
> Does any of this make any sense?
Yes.  Some of the netstats take a while but can be timed out by the random
helper (which uses SIGINT, not SIGTERM as I previously said) but the one
that hangs and can't be interrupted is netstat -ia ("Displays the state
of
configured interfaces") which means it's probably hung in an ioctl
syscall
and can't be interrupted even with a SIGKILL.

-- 
Darren Tucker (dtucker at zip.com.au)
GPG key 8FF4FA69 / D9A3 86E9 7EEE AF4B B2D4  37C9 C982 80C7 8FF4 FA69
    Good judgement comes with experience. Unfortunately, the experience
usually comes from bad judgement.