Re: [PATCH] please review. file descriptor passing for libc_r.

* Daniel Eischen <eischen_at_vigrid.com> [031122 20:46] wrote:
> On Sat, 22 Nov 2003, Alfred Perlstein wrote:
> 
> > This should make things work properly for apps that are linked
> > against libc_r and use filedescriptor passing.
> > 
> > Can someone review and approve it please?
> 
> This isn't needed.  Any time a file descriptor is used, a call
> should first be made to _FD_LOCK() which ends up calling
> _thread_fd_lock() which then calls _thread_fd_table_init()
> for the file descriptor.
> 
> If there's a reason that this is necessary, then I think it
> is just covering up another problem.  I think I made this
> same comment a couple of years ago ;-)

Sorry for the long delay here, I was in Vegas for my sister's
birthday.

Actually there is a problem here. :(

Descriptor's passed can't be closed because the uthread kernel does this:

int
_close(int fd)
{
        int             flags;
        int             ret;
        struct stat     sb;
        struct fd_table_entry   *entry;

        if ((fd < 0) || (fd >= _thread_dtablesize) ||
            (fd == _thread_kern_pipe[0]) || (fd == _thread_kern_pipe[1]) ||
            (_thread_fd_table[fd] == NULL)) {
                /*
                 * Don't allow silly programs to close the kernel pipe
                 * and non-active descriptors.
                 */
                errno = EBADF;
                ret = -1;
        }
...

So basically, if the entry is not initialized we can't close descriptors.

What do you suggest we do?

-- 
- Alfred Perlstein
- Research Engineering Development Inc.
- email: bright_at_mu.org cell: 408-480-4684