Hello, ludo@gnu.org (Ludovic Courtès) writes: [...] > I believe this is a bug in lshd fixed by something along the lines of > the attached patch. > > Niels, what do you think? > > (Strangely I can’t find equivalent code in OpenSSH.) > > Thanks, > Ludo’. > > --- /tmp/lsh-2.1/src/server_x11.c 2016-08-01 11:58:54.714647794 +0200 > +++ /tmp/lsh-2.1/src/server_x11.c.new 2016-08-01 11:58:46.606563478 +0200 > @@ -151,7 +151,7 @@ DEFINE_COMMAND(open_forwarded_x11) > #define X11_MIN_DISPLAY 10 > #define X11_MAX_DISPLAY 1000 > > -/* FIXME: Create the /tmp/.X11-unix directory, if needed. Figure out > +/* FIXME: Figure out > * if and how we should use /tmp/.X17-lock. Consider using display > * "unix:17" instead of just ":17". > */ > @@ -253,6 +253,7 @@ open_x11_socket(struct ssh_channel *chan > > old_umask = umask(0077); > > + mkdir(X11_SOCKET_DIR, S_IRWXU | S_IRWXG | S_IRWXO | S_ISVTX); > old_cd = lsh_pushd(X11_SOCKET_DIR, &dir, 0, 0); > if (old_cd < 0) > { I tried the above fix and ran a VM with the attached config (the custom kernel stuff was to try something else at the same time). It fixes the error about the directory, but it would still fail at X11 forwarding (there was an error message: "Can't find any xauth information for X11 display"). I ended up figuring out it needed libxau to work; combined with your patch, this fixes X11 forwarding. Fixed in commit 0ec195ff02. For the record, I've done the tests in a VM using the attached system config. The custom kernel stuff is unrelated. Thanks, Closing, Maxim