openssh bugs - Sep 2016 - [Bug 2612] New: Sporadically need to press <ENTER> for multiplexed connection to finish

https://bugzilla.mindrot.org/show_bug.cgi?id=2612

            Bug ID: 2612
           Summary: Sporadically need to press <ENTER> for multiplexed
                    connection to finish
           Product: Portable OpenSSH
           Version: 7.3p1
          Hardware: amd64
                OS: Linux
            Status: NEW
          Severity: minor
          Priority: P5
         Component: ssh
          Assignee: unassigned-bugs at mindrot.org
          Reporter: omar.estrada at gmail.com

Sometimes, when issuing a remote command via a multiplexed connection,
it doesn't finish until I hit <ENTER>. Most of the times it terminates
normally. This occurs approximately 1 out of 10 times. This affects the
usage of multiplexed connections in scripts.

Adding -vvv to the "slave" connection shows: 

    debug3: mux_client_request_session: entering
    debug3: mux_client_request_alive: entering
    debug3: mux_client_request_alive: done pid = 5485
    debug3: mux_client_request_session: session request sent
    debug1: mux_client_request_session: master session id: 3
    <Have to hit ENTER here>
    debug3: mux_client_read_packet: read header failed: Broken pipe
    debug2: Received exit status from master 0

I checked out the code (commit
775f8a23f2353f5869003c57a213d14b28e0736e) and, after adding some
additional log messages to mux.c, I notice the flow was blocking in
mux_client_read(), in the call to poll():

    1428 static int                                                     
    1429 mux_client_read(int fd, Buffer *b, u_int need)                 
    1430 {                                      
    ...
    1450                         case EAGAIN:                           
    1451                                 (void)poll(&pfd, 1, -1);
<<<<<<<<<<<
    1452                                 /* FALLTHROUGH */              

Steps to reproduce:
1. In a terminal, issue: 
   ssh -M -S $TMPDIR/master <host>
2. In another terminal, issue:
   for i in `seq 30`; do ssh -vvv -o ControlPath=$TMPDIR/master <host>
ls; done

Expected behavior: the command finishes running without user
interaction.

Actual behavior: approx. 1 out of 10 times it's necessary to press
<ENTER> (no other key works) to continue executing the command.

WORKAROUND:

Notice that redirecting STDIN to /dev/null works around the problem
(e.g., for i in `seq 30`; do ssh -vvv -o ControlPath=$TMPDIR/master
<host> ls < /dev/null; done

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

https://bugzilla.mindrot.org/show_bug.cgi?id=2612

Darren Tucker <dtucker at zip.com.au> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dtucker at zip.com.au

--- Comment #1 from Darren Tucker <dtucker at zip.com.au> ---
I can't reproduce this on either of:
 - Fedora 24 (i686, kernel 4.7.3-200.fc24.i686+PAE)
 - Ubuntu 16.04.1 (x86_64, kernel 4.4.0-43-generic)

Could you describe your system some more (eg distro, kernel)?  What
shell are you using?

-- 
You are receiving this mail because:
You are watching the assignee of the bug.
You are watching someone on the CC list of the bug.

https://bugzilla.mindrot.org/show_bug.cgi?id=2612

Damien Miller <djm at mindrot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |djm at mindrot.org
             Status|NEW                         |RESOLVED
         Resolution|---                         |WORKSFORME

--- Comment #2 from Damien Miller <djm at mindrot.org> ---
9 months with no followup == no bug

-- 
You are receiving this mail because:
You are watching the assignee of the bug.
You are watching someone on the CC list of the bug.

https://bugzilla.mindrot.org/show_bug.cgi?id=2612

Damien Miller <djm at mindrot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #3 from Damien Miller <djm at mindrot.org> ---
Close all resolved bugs after release of OpenSSH 7.7.

-- 
You are receiving this mail because:
You are watching someone on the CC list of the bug.
You are watching the assignee of the bug.