[Libguestfs] [libnbd PATCH v3 06/19] socket activation: plug AF_UNIX socket address (filesystem) leak on error

A prior patch highlighted the following leak: whenever any construction step after bind() fails, the UNIX domain socket address (i.e., the pathname in the filesystem) is leaked. Plug the leak by inserting a new error handling section that unlinks the pathname. Signed-off-by: Laszlo Ersek <lersek(a)redhat.com&gt; Reviewed-by: Richard W.M. Jones <rjones(a)redhat.com&gt; --- Notes: v4: - pick up Rich's R-b - resolve rebase conflict near the prepare_socket_activation_environment() call site, due to keeping set_error() internal to that callee, in patch "socket activation: clean up responsibilities of prep.sock.act.env.()" context:-U12 generator/states-connect-socket-activation.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/generator/states-connect-socket-activation.c b/generator/states-connect-socket-activation.c index b20c3f1f10bb..752ee73bb62f 100644 --- a/generator/states-connect-socket-activation.c +++ b/generator/states-connect-socket-activation.c @@ -137,30 +137,30 @@ CONNECT_SA.START: goto free_sockpath; } addr.sun_family = AF_UNIX; memcpy (addr.sun_path, sockpath, strlen (sockpath) + 1); if (bind (s, (struct sockaddr *) &addr, sizeof addr) == -1) { set_error (errno, "bind: %s", sockpath); goto close_socket; } if (listen (s, SOMAXCONN) == -1) { set_error (errno, "listen"); - goto close_socket; + goto unlink_sockpath; } if (prepare_socket_activation_environment (&env) == -1) /* prepare_socket_activation_environment() calls set_error() internally */ - goto close_socket; + goto unlink_sockpath; pid = fork (); if (pid == -1) { set_error (errno, "fork"); goto empty_env; } if (pid == 0) { /* child - run command */ if (s != FIRST_SOCKET_ACTIVATION_FD) { if (dup2 (s, FIRST_SOCKET_ACTIVATION_FD) == -1) { nbd_internal_fork_safe_perror ("dup2"); _exit (126); @@ -211,24 +211,28 @@ CONNECT_SA.START: h->sact_sockpath = sockpath; h->pid = pid; h->connaddrlen = sizeof addr; memcpy (&h->connaddr, &addr, h->connaddrlen); next = %^CONNECT.START; /* fall through, for releasing the temporaries */ empty_env: string_vector_empty (&env); +unlink_sockpath: + if (next == %.DEAD) + unlink (sockpath); + close_socket: close (s); free_sockpath: if (next == %.DEAD) free (sockpath); rmdir_tmpdir: if (next == %.DEAD) rmdir (tmpdir); free_tmpdir: