cpython/Doc/librestricted.tex

\chapter{Restricted Execution}
\label{restricted}

In general, Python programs have complete access to the underlying
operating system throug the various functions and classes, For
example, a Python program can open any file for reading and writing by
using the \code{open()} built-in function (provided the underlying OS
gives you permission!).  This is exactly what you want for most
applications.

There exists a class of applications for which this ``openness'' is
inappropriate.  Take Grail: a web browser that accepts ``applets'',
snippets of Python code, from anywhere on the Internet for execution
on the local system.  This can be used to improve the user interface
of forms, for instance.  Since the originator of the code is unknown,
it is obvious that it cannot be trusted with the full resources of the
local machine.

\emph{Restricted execution} is the basic framework in Python that allows
for the segregation of trusted and untrusted code.  It is based on the
notion that trusted Python code (a \emph{supervisor}) can create a
``padded cell' (or environment) with limited permissions, and run the
untrusted code within this cell.  The untrusted code cannot break out
of its cell, and can only interact with sensitive system resources
through interfaces defined and managed by the trusted code.  The term
``restricted execution'' is favored over ``safe-Python''
since true safety is hard to define, and is determined by the way the
restricted environment is created.  Note that the restricted
environments can be nested, with inner cells creating subcells of
lesser, but never greater, privilege.

An interesting aspect of Python's restricted execution model is that
the interfaces presented to untrusted code usually have the same names
as those presented to trusted code.  Therefore no special interfaces
need to be learned to write code designed to run in a restricted
environment.  And because the exact nature of the padded cell is
determined by the supervisor, different restrictions can be imposed,
depending on the application.  For example, it might be deemed
``safe'' for untrusted code to read any file within a specified
directory, but never to write a file.  In this case, the supervisor
may redefine the built-in
\code{open()} function so that it raises an exception whenever the
\var{mode} parameter is \code{'w'}.  It might also perform a
\code{chroot()}-like operation on the \var{filename} parameter, such
that root is always relative to some safe ``sandbox'' area of the
filesystem.  In this case, the untrusted code would still see an
built-in \code{open()} function in its environment, with the same
calling interface.  The semantics would be identical too, with
\code{IOError}s being raised when the supervisor determined that an
unallowable parameter is being used.

The Python run-time determines whether a particular code block is
executing in restricted execution mode based on the identity of the
\code{__builtins__} object in its global variables: if this is (the
dictionary of) the standard \code{__builtin__} module, the code is
deemed to be unrestricted, else it is deemed to be restricted.

Python code executing in restricted mode faces a number of limitations
that are designed to prevent it from escaping from the padded cell.
For instance, the function object attribute \code{func_globals} and the
class and instance object attribute \code{__dict__} are unavailable.

Two modules provide the framework for setting up restricted execution
environments:

\begin{description}

\item[rexec]
--- Basic restricted execution framework.

\item[Bastion]
--- Providing restricted access to objects.

\end{description}