[Kernel-packages] [Bug 1593657] Re: ControlPath unix socket is created in a weird way

2016-06-17 Thread Uqbar
I added a couple of extra debugging lines in mux.c, recompiled and ran it. debug3: link( "/tmp/sshcontrol-f1b458a85cd2aa9d5c115abc12afee9f485ad6b0.XSLUj2BsBSDnTkhZ","/tmp/sshcontrol-f1b458a85cd2aa9d5c115abc12afee9f485ad6b0" ): 0 debug3: unlink( "/tmp/sshcontrol-f1b458a85cd2aa9d5c115abc12afee9f4

[Kernel-packages] [Bug 1593657] Re: ControlPath unix socket is created in a weird way

2016-06-17 Thread Uqbar
I am not 100% sure it's not the OpenSSH client. These two debug lines make me think there's something more to check: debug3: muxserver_listen: temporary control path /tmp/sshcontrol-f1b458a85cd2aa9d5c115abc12afee9f485ad6b0.OqWT4bhQP7Hgit3A debug1: channel 1: new [/tmp/sshcontrol-f1b458a85cd2aa9d

[Kernel-packages] [Bug 1593657] Re: ControlPath unix socket is created in a weird way

2016-06-17 Thread Uqbar
OK. I did just a bare minimum checks. The file is called mux.c, the function is "void muxserver_listen(void)" at line 1241. Two comments seem quite important: /* * Use a temporary path before listen so we can pseudo-atomically * establish the listening socket in its fina

[Kernel-packages] [Bug 1593657] Re: ControlPath unix socket is created in a weird way

2016-06-17 Thread Uqbar
I am not really sure this is not OpenSSH. The logs above are quoting both socket names. And the name, afaik, is defined by the program (openssh-client) not by the kernel. But I could be badly wrong as I have no time to inspect the software source code. -- You received this bug notification beca

[Kernel-packages] [Bug 1593657] Re: ControlPath unix socket is created in a weird way

2016-06-17 Thread Uqbar
I am reporting here the debugging logs I've got with "-vvv" and also expunging all stuff not related to the bug: OpenSSH_7.2p2 Ubuntu-4ubuntu1, OpenSSL 1.0.2g-fips 1 Mar 2016 debug1: auto-mux: Trying existing master debug1: Control socket "/tmp/sshcontrol-f1b458a85cd2aa9d5c115abc12afee9f485ad6b0