[2/2,v2] openpty: use TIOCGPTPEER to open slave side fd

  Newer kernels expose the ioctl TIOCGPTPEER [1] call to userspace which allows to
safely allocate a file descriptor for a pty slave based solely on the master
file descriptor. This allows us to avoid path-based operations and makes this
function a lot safer in the face of devpts mounts in different mount namespaces.

[1]: https://patchwork.kernel.org/patch/9760743/

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
Changelog 2017-08-28:
* Instead of #ifdefing the TIOCGPTPEER ioctl flag we now try the ioctl() first
  and if it fails we fallback to path-based allocation of the slave fd. This
  allows us retain backward compatibility with kernels that do not support this
  ioctl call.
* A note on the following codepath

   if (name != NULL)
     {
       if (*buf == '\0')
         if (pts_name (master, &buf, sizeof (_buf)))
           goto fail;

       strcpy (name, buf);
     }

  "buf" is guaranteed to be allocated in this case. If the pts_name() call above
  failed we would have never reached this code path. If it has been called
  succesfully it will either have handed us a valid buffer or "buf" will still
  point to the static char array "_buf" which is initialized to 0.
---
 ChangeLog       |  5 +++++
 login/openpty.c | 36 +++++++++++++++++++++++++-----------
 2 files changed, 30 insertions(+), 11 deletions(-)
  

On Mon, 28 Aug 2017, Christian Brauner wrote:

> +  /* Try to allocate slave fd solely based on master fd first. */
> +  slave = ioctl (master, TIOCGPTPEER, O_RDWR | O_NOCTTY);

You still need the #ifdef.  openpty is used on non-Linux systems, and the 
oldest kernel headers supported for building glibc are 3.2.
  

On Aug 28 2017, Christian Brauner <christian.brauner@ubuntu.com> wrote:

> diff --git a/login/openpty.c b/login/openpty.c
> index 8fbc66a3ef..43b34d7a9c 100644
> --- a/login/openpty.c
> +++ b/login/openpty.c
> @@ -87,9 +87,9 @@ openpty (int *amaster, int *aslave, char *name,
>  	 const struct termios *termp, const struct winsize *winp)
>  {
>  #ifdef PATH_MAX
> -  char _buf[PATH_MAX];
> +  char _buf[PATH_MAX] = {0};
>  #else
> -  char _buf[512];
> +  char _buf[512] = {0};
>  #endif

No need to clear the whole array, just the first byte.

Andreas.