openssh unix dev - Jun 2008 - ibuf_empty delayed efd

I'm seeing something unusual in 5.0p1.  Let me start by saying that I'm
on kind of an oddball
system (HP NonStop).

What I'm seeing is that at the end of an scp session, the server gets
stuck in a loop.
First I see a shutdown failure, followed by looping on an "ibuf_empty
delayed efd 9/(0)" condition.

This may have to do with some minor semantic differences in the way the
NonStop socket layer works,
as opposed to stock XPG4 or BSD sockets.  What I'm really concerned
about is the looping.

Debugging output excerpted below.

-- begin excerpt
debug2: channel 0: rcvd adjust 131072
debug2: channel 0: read<=0 rfd 7 len 0
debug2: channel 0: read failed
debug2: channel 0: close_read
debug2: channel 0: input open -> drain
debug2: channel 0: ibuf_empty delayed efd 9/(0)
debug2: channel 0: read 0 from efd 9
debug2: channel 0: ibuf_empty delayed efd 9/(0)
debug2: notify_done: reading
debug1: Received SIGCHLD.
debug1: session_by_pid: pid 620560410
debug1: session_exit_message: session 0 channel 0 pid 620560410
debug2: channel 0: request exit-status confirm 0
debug1: session_exit_message: release channel 0
debug2: channel 0: write failed
debug2: channel 0: close_write
debug2: channel 0: chan_shutdown_write: shutdown() failed for fd7:
Socket is not connected
debug2: channel 0: output open -> closed
debug2: channel 0: read 0 from efd 9
debug2: channel 0: ibuf_empty delayed efd 9/(0)
debug2: channel 0: read 0 from efd 9
debug2: channel 0: ibuf_empty delayed efd 9/(0)
debug2: channel 0: read 0 from efd 9
-- end excerpt

----
Scott Neugroschl
XYPRO Technologies
scott_n at xypro.com
805-583-2874 x121

On Tue, 10 Jun 2008, Scott Neugroschl wrote:
> I'm seeing something unusual in 5.0p1.  Let me start by saying that
I'm
> on kind of an oddball
> system (HP NonStop).
> 
> What I'm seeing is that at the end of an scp session, the server gets
> stuck in a loop.
> First I see a shutdown failure, followed by looping on an "ibuf_empty
> delayed efd 9/(0)" condition.
I'm not sure what is going wrong here, but could you please try a
snapshot? We fixed a long-existing race in the efd handling @ close
case and the fix may serendibitously fix your case :)

Snapshots are at: http://www.mindrot.org/openssh_snap/

-d