On Wed, Aug 29, 2007 at 01:42:36AM -0700, Darren Reed wrote:
> if it isn't a hard hang, can you telnet/rlogin/ssh in?
> is it just the console that dies?

Yes. I started up sshd and I can get in. When the console is
hanging the ssh connection is perceptibly slower to respond -
about 5 seconds to get the password prompt, as opposed to the
normal fraction of a second.

I don't see anything unusual as far as processes go. And it
isn't just the keyboard but the screen as well - e.g. there
is no message on the screen about the root login, and nothing
while halting the system over the network. I have to do stop-a
or power-cycle to get the screen back to normal.

When I issue a "reboot" over ssh, the connection is closed as
expected, but then the machine hangs and refuses connections;
stop-a still works then. In mdb, ::ps shows powerd and picld
running along with reboot.
 
> also, if you can STOP-A (or L1-A), do a "sync" on a hung
> system to create a crash dump that can be analyzed further.

Yes, I did, but as I said I'm no coredump genius... sched is
always on the CPU, findstack shows usbkbm_unpack_usb_packet,
which is all probably normal.

I now created four: one just after boot, before the hang occurs;
one while the console is stuck; one after an ssh reboot, when it
gives "connection refused"; and finally the fourth for a good
measure while the computer is working normally (power management
disabled). They are large, but in case anyone cares to look:
  http://www.math.niu.edu/~behr/pfilcores/

Thanks a lot for the response, Darren.

-- 
Eric Behr          | NIU Mathematical Sciences      | (815) 753 6727
[EMAIL PROTECTED]  | http://www.math.niu.edu/~behr/ |  fax: 753 1112

Reply via email to