openssh bugs - Feb 2005 - [Bug 52] ssh hangs on exit

http://bugzilla.mindrot.org/show_bug.cgi?id=52





------- Additional Comments From t8m at centrum.cz  2005-02-03 22:06 -------
Created an attachment (id=801)
 --> (http://bugzilla.mindrot.org/attachment.cgi?id=801&action=view)
Patch submitted by John Bowman (bowman at math.ualberta.ca)

This patch was attached to
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=39128

Also see https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=24293 a
regression which happened from another attempt to solve this hang.




------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

http://bugzilla.mindrot.org/show_bug.cgi?id=52


t8m at centrum.cz changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |t8m at centrum.cz






------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

http://bugzilla.mindrot.org/show_bug.cgi?id=52





------- Additional Comments From djm at mindrot.org  2005-02-03 22:49 -------
1. That doesn't look like a fix - it appears to be a workaround by adding a
exit
delay.

2. We don't have access to that Redhat regression bug, it is limited to
insiders
only (why?). Guessing from the bugs filed on either side of it, it looks like it
was filed way back in early 2001, so it may have been related to my early and
misguided attempt to fix the issue. Those hacks were backed out before the
release of openssh-2.5.0p1.

Anyone trying to fix this bug should read comment #1, comment #2 and Marc Aurele
La France's patches. So far, Marc seems to have struck closest to the root
cause
of this issue.



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

http://bugzilla.mindrot.org/show_bug.cgi?id=52





------- Additional Comments From tsi at ualberta.ca  2005-02-04 02:19 -------
Well, I haven't had a lot of time to spend on this, but the hangs I'm
still
seeing after my patch are intermittent and difficult to duplicate. 
"High-ish"
loads seem to trigger them, after, for example, whatever's been put in the
background has had a chance to increase the load average.  So there's
definitely
a timing issue here somewhere.

The "child" sshd is left sitting in select() with SIGCHLD blocked
after its own
child (the shell or whatever) has been reaped.  The few forays into the code
I've had time to do so far have yet to explain, to me at least, how this can
happen.  I suspect client_alive_scheduled shouldn't be reset, but I've
yet to
confirm this, or even whether these remaining hangs are in fact related to the
problem at hand here or some other new problem.

I also agree adding a configurable delay doesn't solve anything.  After all,
what's a hang if not a delay?

BTW, I reassigned the bug to myself as a remainder that I need to spend time on
this.  But if don't you want me to do this re-assignment, that's fine
too.



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

http://bugzilla.mindrot.org/show_bug.cgi?id=52


djm at mindrot.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|914                         |
              nThis|                            |






------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.