Apple checks EUID==0 on ioctl(/dev/pf), whereas OpenBSD and FreeBSD only
check permissions on open(/dev/pf). This means that on OS X, it is not
possible to open /dev/pf, drop privileges, and send an ioctl to the file
descriptor opened earlier with EUID==0. It also means Apple broke the
Unix way of dealing with device nodes - why are there file permissions
on /dev/pf when they later enforce EUID==0 on use, thereby breaking
basic Unix mechanisms? Work around this by disallowing -u with pf
proxyspecs and by not automatically dropping to nobody on Mac OS X.
Issue: #65
Reported by: Vladimir Marteev
Fork into a monitor parent process and an actual proxy child process,
communicating over AF_UNIX sockets. Certain privileged operations are
performed through the privileged parent process, like opening log files
or listener sockets, while all other operations happen in the child
process, which can now drop its privileges without side-effects for
log file opening and other privileged operations. This is also a
preparation for -l/-L logfile reopening through SIGUSR1.
This means that -S and -F are no longer relative to chroot() if used
with -j. This is a deliberate POLA violation.
Local process lookup is independent of the NAT engine used, it depends
only on the operating system's process enumeration API. Moving the code
out of NAT lookup also makes it work for static and SNI proxyspecs.
This exposes the pid lookup code as a standard attribute
of NAT lookup -- if a matching process cannot be found,
or if pid lookup isn't supported by the NAT backend,
a pid of -1 is returned.
This also adds the local_pid to the pxyconn context; this
will be used to populate log strings.
