[Bug 2953] Race during daemon reload may cause to fail to listen on configured ports

[bugzilla-daemon at bugzilla.mindrot.org]

https://bugzilla.mindrot.org/show_bug.cgi?id=2953

Damien Miller <djm at mindrot.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Attachment #3222|0                           |1
        is obsolete|                            |
             Status|NEW                         |ASSIGNED
                 CC|                            |djm at mindrot.org,
                   |                            |dtucker at dtucker.net
           Assignee|unassigned-bugs at mindrot.org |djm at mindrot.org
   Attachment #3245|                            |ok?(dtucker at dtucker.net)
              Flags|                            |

--- Comment #3 from Damien Miller <djm at mindrot.org> ---
Created attachment 3245
  --> https://bugzilla.mindrot.org/attachment.cgi?id=3245&action=edit
reuse startup_pipes; fix race with rexeced children too

I don't like the complexity of having another set of fds just for
handling the listen_socks. Instead, this reuses the existing
startup_pipes for signalling, but writing a char over them when the
child process has finished "early startup".

"early startup" includes closing listen_socks but also, for the rexec
case, receiving the re-executed child's state from sshd. This avoids
another race condition: at SIGHUP the listener sshd was exiting before
the child received its configuration.

-- 
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.