gh-95023: Added os.setns and os.unshare functions (#95046)

Added os.setns and os.unshare to easily switch between namespaces
on Linux.

Co-authored-by: Christian Heimes <christian@python.org>
Co-authored-by: CAM Gerlach <CAM.Gerlach@Gerlach.CAM>
Co-authored-by: Victor Stinner <vstinner@python.org>

Doc/library/os.rst

@@ -590,6 +590,44 @@ process and user.
       See the documentation for :func:`getgroups` for cases where it may not
       return the same group list set by calling setgroups().
 
+.. function:: setns(fd, nstype=0)
+
+   Reassociate the current thread with a Linux namespace.
+   See the :manpage:`setns(2)` and :manpage:`namespaces(7)` man pages for more
+   details.
+
+   If *fd* refers to a :file:`/proc/{pid}/ns/` link, ``setns()`` reassociates the
+   calling thread with the namespace associated with that link,
+   and *nstype* may be set to one of the
+   :ref:`CLONE_NEW* constants <os-unshare-clone-flags>`
+   to impose constraints on the operation
+   (``0`` means no constraints).
+
+   Since Linux 5.8, *fd* may refer to a PID file descriptor obtained from
+   :func:`~os.pidfd_open`. In this case, ``setns()`` reassociates the calling thread
+   into one or more of the same namespaces as the thread referred to by *fd*.
+   This is subject to any constraints imposed by *nstype*,
+   which is a bit mask combining one or more of the
+   :ref:`CLONE_NEW* constants <os-unshare-clone-flags>`,
+   e.g. ``setns(fd, os.CLONE_NEWUTS | os.CLONE_NEWPID)``.
+   The caller's memberships in unspecified namespaces are left unchanged.
+
+   *fd* can be any object with a :meth:`~io.IOBase.fileno` method, or a raw file descriptor.
+
+   This example reassociates the thread with the ``init`` process's network namespace::
+
+      fd = os.open("/proc/1/ns/net", os.O_RDONLY)
+      os.setns(fd, os.CLONE_NEWNET)
+      os.close(fd)
+
+   .. availability:: Linux >= 3.0 with glibc >= 2.14.
+
+   .. versionadded:: 3.12
+
+   .. seealso::
+
+      The :func:`~os.unshare` function.
+
 .. function:: setpgrp()
 
    Call the system call :c:func:`setpgrp` or ``setpgrp(0, 0)`` depending on
@@ -756,6 +794,49 @@ process and user.
       The function is now always available and is also available on Windows.
 
 
+.. function:: unshare(flags)
+
+   Disassociate parts of the process execution context, and move them into a
+   newly created namespace.
+   See the :manpage:`unshare(2)`
+   man page for more details.
+   The *flags* argument is a bit mask, combining zero or more of the
+   :ref:`CLONE_* constants <os-unshare-clone-flags>`,
+   that specifies which parts of the execution context should be
+   unshared from their existing associations and moved to a new namespace.
+   If the *flags* argument is ``0``, no changes are made to the calling process's
+   execution context.
+
+   .. availability:: Linux >= 2.6.16.
+
+   .. versionadded:: 3.12
+
+   .. seealso::
+
+      The :func:`~os.setns` function.
+
+.. _os-unshare-clone-flags:
+
+Flags to the :func:`unshare` function, if the implementation supports them.
+See :manpage:`unshare(2)` in the Linux manual
+for their exact effect and availability.
+
+.. data:: CLONE_FILES
+          CLONE_FS
+          CLONE_NEWCGROUP
+          CLONE_NEWIPC
+          CLONE_NEWNET
+          CLONE_NEWNS
+          CLONE_NEWPID
+          CLONE_NEWTIME
+          CLONE_NEWUSER
+          CLONE_NEWUTS
+          CLONE_SIGHAND
+          CLONE_SYSVSEM
+          CLONE_THREAD
+          CLONE_VM
+
+
 .. _os-newstreams:
 
 File Object Creation

Include/internal/pycore_global_strings.h

@@ -489,6 +489,7 @@ struct _Py_global_strings {
         STRUCT_FOR_ID(node_depth)
         STRUCT_FOR_ID(node_offset)
         STRUCT_FOR_ID(ns)
+        STRUCT_FOR_ID(nstype)
         STRUCT_FOR_ID(number)
         STRUCT_FOR_ID(obj)
         STRUCT_FOR_ID(object)

Include/internal/pycore_runtime_init_generated.h

@@ -998,6 +998,7 @@ extern "C" {
                 INIT_ID(node_depth), \
                 INIT_ID(node_offset), \
                 INIT_ID(ns), \
+                INIT_ID(nstype), \
                 INIT_ID(number), \
                 INIT_ID(obj), \
                 INIT_ID(object), \
@@ -2307,6 +2308,8 @@ _PyUnicode_InitStaticStrings(void) {
     PyUnicode_InternInPlace(&string);
     string = &_Py_ID(ns);
     PyUnicode_InternInPlace(&string);
+    string = &_Py_ID(nstype);
+    PyUnicode_InternInPlace(&string);
     string = &_Py_ID(number);
     PyUnicode_InternInPlace(&string);
     string = &_Py_ID(obj);
@@ -6546,6 +6549,10 @@ _PyStaticObjects_CheckRefcnt(void) {
         _PyObject_Dump((PyObject *)&_Py_ID(ns));
         Py_FatalError("immortal object has less refcnt than expected _PyObject_IMMORTAL_REFCNT");
     };
+    if (Py_REFCNT((PyObject *)&_Py_ID(nstype)) < _PyObject_IMMORTAL_REFCNT) {
+        _PyObject_Dump((PyObject *)&_Py_ID(nstype));
+        Py_FatalError("immortal object has less refcnt than expected _PyObject_IMMORTAL_REFCNT");
+    };
     if (Py_REFCNT((PyObject *)&_Py_ID(number)) < _PyObject_IMMORTAL_REFCNT) {
         _PyObject_Dump((PyObject *)&_Py_ID(number));
         Py_FatalError("immortal object has less refcnt than expected _PyObject_IMMORTAL_REFCNT");

Lib/test/test_posix.py

@@ -2191,6 +2191,53 @@ def test_utime(self):
                 os.utime("path", dir_fd=0)
 
 
+class NamespacesTests(unittest.TestCase):
+    """Tests for os.unshare() and os.setns()."""
+
+    @unittest.skipUnless(hasattr(os, 'unshare'), 'needs os.unshare()')
+    @unittest.skipUnless(hasattr(os, 'setns'), 'needs os.setns()')
+    @unittest.skipUnless(os.path.exists('/proc/self/ns/uts'), 'need /proc/self/ns/uts')
+    @support.requires_linux_version(3, 0, 0)
+    def test_unshare_setns(self):
+        code = """if 1:
+            import errno
+            import os
+            import sys
+            fd = os.open('/proc/self/ns/uts', os.O_RDONLY)
+            try:
+                original = os.readlink('/proc/self/ns/uts')
+                try:
+                    os.unshare(os.CLONE_NEWUTS)
+                except OSError as e:
+                    if e.errno == errno.ENOSPC:
+                        # skip test if limit is exceeded
+                        sys.exit()
+                    raise
+                new = os.readlink('/proc/self/ns/uts')
+                if original == new:
+                    raise Exception('os.unshare failed')
+                os.setns(fd, os.CLONE_NEWUTS)
+                restored = os.readlink('/proc/self/ns/uts')
+                if original != restored:
+                    raise Exception('os.setns failed')
+            except PermissionError:
+                # The calling process did not have the required privileges
+                # for this operation
+                pass
+            except OSError as e:
+                # Skip the test on these errors:
+                # - ENOSYS: syscall not available
+                # - EINVAL: kernel was not configured with the CONFIG_UTS_NS option
+                # - ENOMEM: not enough memory
+                if e.errno not in (errno.ENOSYS, errno.EINVAL, errno.ENOMEM):
+                    raise
+            finally:
+                os.close(fd)
+            """
+
+        assert_python_ok("-c", code)
+
+
 def tearDownModule():
     support.reap_children()
 

Misc/ACKS

@@ -344,6 +344,7 @@ Hervé Coatanhay
 Riccardo Coccioli
 Nick Coghlan
 Josh Cogliati
+Noam Cohen
 Dave Cole
 Terrence Cole
 Benjamin Collar

Misc/NEWS.d/next/Core and Builtins/2022-07-20-09-04-55.gh-issue-95023.bs-xd7.rst

@@ -0,0 +1 @@
+Implement :func:`os.setns` and  :func:`os.unshare` for Linux. Patch by Noam Cohen.

Modules/clinic/posixmodule.c.h

@@ -5211,6 +5211,147 @@ exit:
 
 #endif /* (defined(__linux__) && defined(__NR_pidfd_open)) */
 
+#if defined(HAVE_SETNS)
+
+PyDoc_STRVAR(os_setns__doc__,
+"setns($module, /, fd, nstype=0)\n"
+"--\n"
+"\n"
+"Move the calling thread into different namespaces.\n"
+"\n"
+"  fd\n"
+"    A file descriptor to a namespace.\n"
+"  nstype\n"
+"    Type of namespace.");
+
+#define OS_SETNS_METHODDEF    \
+    {"setns", _PyCFunction_CAST(os_setns), METH_FASTCALL|METH_KEYWORDS, os_setns__doc__},
+
+static PyObject *
+os_setns_impl(PyObject *module, int fd, int nstype);
+
+static PyObject *
+os_setns(PyObject *module, PyObject *const *args, Py_ssize_t nargs, PyObject *kwnames)
+{
+    PyObject *return_value = NULL;
+    #if defined(Py_BUILD_CORE) && !defined(Py_BUILD_CORE_MODULE)
+
+    #define NUM_KEYWORDS 2
+    static struct {
+        PyGC_Head _this_is_not_used;
+        PyObject_VAR_HEAD
+        PyObject *ob_item[NUM_KEYWORDS];
+    } _kwtuple = {
+        .ob_base = PyVarObject_HEAD_INIT(&PyTuple_Type, NUM_KEYWORDS)
+        .ob_item = { &_Py_ID(fd), &_Py_ID(nstype), },
+    };
+    #undef NUM_KEYWORDS
+    #define KWTUPLE (&_kwtuple.ob_base.ob_base)
+
+    #else  // !Py_BUILD_CORE
+    #  define KWTUPLE NULL
+    #endif  // !Py_BUILD_CORE
+
+    static const char * const _keywords[] = {"fd", "nstype", NULL};
+    static _PyArg_Parser _parser = {
+        .keywords = _keywords,
+        .fname = "setns",
+        .kwtuple = KWTUPLE,
+    };
+    #undef KWTUPLE
+    PyObject *argsbuf[2];
+    Py_ssize_t noptargs = nargs + (kwnames ? PyTuple_GET_SIZE(kwnames) : 0) - 1;
+    int fd;
+    int nstype = 0;
+
+    args = _PyArg_UnpackKeywords(args, nargs, NULL, kwnames, &_parser, 1, 2, 0, argsbuf);
+    if (!args) {
+        goto exit;
+    }
+    if (!_PyLong_FileDescriptor_Converter(args[0], &fd)) {
+        goto exit;
+    }
+    if (!noptargs) {
+        goto skip_optional_pos;
+    }
+    nstype = _PyLong_AsInt(args[1]);
+    if (nstype == -1 && PyErr_Occurred()) {
+        goto exit;
+    }
+skip_optional_pos:
+    return_value = os_setns_impl(module, fd, nstype);
+
+exit:
+    return return_value;
+}
+
+#endif /* defined(HAVE_SETNS) */
+
+#if defined(HAVE_UNSHARE)
+
+PyDoc_STRVAR(os_unshare__doc__,
+"unshare($module, /, flags)\n"
+"--\n"
+"\n"
+"Disassociate parts of a process (or thread) execution context.\n"
+"\n"
+"  flags\n"
+"    Namespaces to be unshared.");
+
+#define OS_UNSHARE_METHODDEF    \
+    {"unshare", _PyCFunction_CAST(os_unshare), METH_FASTCALL|METH_KEYWORDS, os_unshare__doc__},
+
+static PyObject *
+os_unshare_impl(PyObject *module, int flags);
+
+static PyObject *
+os_unshare(PyObject *module, PyObject *const *args, Py_ssize_t nargs, PyObject *kwnames)
+{
+    PyObject *return_value = NULL;
+    #if defined(Py_BUILD_CORE) && !defined(Py_BUILD_CORE_MODULE)
+
+    #define NUM_KEYWORDS 1
+    static struct {
+        PyGC_Head _this_is_not_used;
+        PyObject_VAR_HEAD
+        PyObject *ob_item[NUM_KEYWORDS];
+    } _kwtuple = {
+        .ob_base = PyVarObject_HEAD_INIT(&PyTuple_Type, NUM_KEYWORDS)
+        .ob_item = { &_Py_ID(flags), },
+    };
+    #undef NUM_KEYWORDS
+    #define KWTUPLE (&_kwtuple.ob_base.ob_base)
+
+    #else  // !Py_BUILD_CORE
+    #  define KWTUPLE NULL
+    #endif  // !Py_BUILD_CORE
+
+    static const char * const _keywords[] = {"flags", NULL};
+    static _PyArg_Parser _parser = {
+        .keywords = _keywords,
+        .fname = "unshare",
+        .kwtuple = KWTUPLE,
+    };
+    #undef KWTUPLE
+    PyObject *argsbuf[1];
+    int flags;
+
+    args = _PyArg_UnpackKeywords(args, nargs, NULL, kwnames, &_parser, 1, 1, 0, argsbuf);
+    if (!args) {
+        goto exit;
+    }
+    flags = _PyLong_AsInt(args[0]);
+    if (flags == -1 && PyErr_Occurred()) {
+        goto exit;
+    }
+    return_value = os_unshare_impl(module, flags);
+
+exit:
+    return return_value;
+}
+
+#endif /* defined(HAVE_UNSHARE) */
+
 #if (defined(HAVE_READLINK) || defined(MS_WINDOWS))
 
 PyDoc_STRVAR(os_readlink__doc__,
@@ -11085,6 +11226,14 @@ exit:
     #define OS_PIDFD_OPEN_METHODDEF
 #endif /* !defined(OS_PIDFD_OPEN_METHODDEF) */
 
+#ifndef OS_SETNS_METHODDEF
+    #define OS_SETNS_METHODDEF
+#endif /* !defined(OS_SETNS_METHODDEF) */
+
+#ifndef OS_UNSHARE_METHODDEF
+    #define OS_UNSHARE_METHODDEF
+#endif /* !defined(OS_UNSHARE_METHODDEF) */
+
 #ifndef OS_READLINK_METHODDEF
     #define OS_READLINK_METHODDEF
 #endif /* !defined(OS_READLINK_METHODDEF) */
@@ -11368,4 +11517,4 @@ exit:
 #ifndef OS_WAITSTATUS_TO_EXITCODE_METHODDEF
     #define OS_WAITSTATUS_TO_EXITCODE_METHODDEF
 #endif /* !defined(OS_WAITSTATUS_TO_EXITCODE_METHODDEF) */
-/*[clinic end generated code: output=471ab8f2ad3d46b0 input=a9049054013a1b77]*/
+/*[clinic end generated code: output=90f5e6995114e5ca input=a9049054013a1b77]*/

Modules/posixmodule.c

@@ -8581,6 +8581,64 @@ os_pidfd_open_impl(PyObject *module, pid_t pid, unsigned int flags)
 #endif
 
 
+#ifdef HAVE_SETNS
+/*[clinic input]
+os.setns
+  fd: fildes
+    A file descriptor to a namespace.
+  nstype: int = 0
+    Type of namespace.
+
+Move the calling thread into different namespaces.
+[clinic start generated code]*/
+
+static PyObject *
+os_setns_impl(PyObject *module, int fd, int nstype)
+/*[clinic end generated code: output=5dbd055bfb66ecd0 input=42787871226bf3ee]*/
+{
+    int res;
+
+    Py_BEGIN_ALLOW_THREADS
+    res = setns(fd, nstype);
+    Py_END_ALLOW_THREADS
+
+    if (res != 0) {
+        return posix_error();
+    }
+
+    Py_RETURN_NONE;
+}
+#endif
+
+
+#ifdef HAVE_UNSHARE
+/*[clinic input]
+os.unshare
+  flags: int
+    Namespaces to be unshared.
+
+Disassociate parts of a process (or thread) execution context.
+[clinic start generated code]*/
+
+static PyObject *
+os_unshare_impl(PyObject *module, int flags)
+/*[clinic end generated code: output=1b3177906dd237ee input=9e065db3232b8b1b]*/
+{
+    int res;
+
+    Py_BEGIN_ALLOW_THREADS
+    res = unshare(flags);
+    Py_END_ALLOW_THREADS
+
+    if (res != 0) {
+        return posix_error();
+    }
+    
+    Py_RETURN_NONE;
+}
+#endif
+
+
 #if defined(HAVE_READLINK) || defined(MS_WINDOWS)
 /*[clinic input]
 os.readlink
@@ -14945,6 +15003,8 @@ static PyMethodDef posix_methods[] = {
     OS__ADD_DLL_DIRECTORY_METHODDEF
     OS__REMOVE_DLL_DIRECTORY_METHODDEF
     OS_WAITSTATUS_TO_EXITCODE_METHODDEF
+    OS_SETNS_METHODDEF
+    OS_UNSHARE_METHODDEF
     {NULL,              NULL}            /* Sentinel */
 };
 
@@ -15390,6 +15450,53 @@ all_ins(PyObject *m)
 #ifdef SCHED_FX
     if (PyModule_AddIntConstant(m, "SCHED_FX", SCHED_FSS)) return -1;
 #endif
+
+/* constants for namespaces */
+#if defined(HAVE_SETNS) || defined(HAVE_UNSHARE)
+#ifdef CLONE_FS
+    if (PyModule_AddIntMacro(m, CLONE_FS)) return -1;
+#endif
+#ifdef CLONE_FILES
+    if (PyModule_AddIntMacro(m, CLONE_FILES)) return -1;
+#endif
+#ifdef CLONE_NEWNS
+    if (PyModule_AddIntMacro(m, CLONE_NEWNS)) return -1;
+#endif
+#ifdef CLONE_NEWCGROUP
+    if (PyModule_AddIntMacro(m, CLONE_NEWCGROUP)) return -1;
+#endif
+#ifdef CLONE_NEWUTS
+    if (PyModule_AddIntMacro(m, CLONE_NEWUTS)) return -1;
+#endif
+#ifdef CLONE_NEWIPC
+    if (PyModule_AddIntMacro(m, CLONE_NEWIPC)) return -1;
+#endif
+#ifdef CLONE_NEWUSER
+    if (PyModule_AddIntMacro(m, CLONE_NEWUSER)) return -1;
+#endif
+#ifdef CLONE_NEWPID
+    if (PyModule_AddIntMacro(m, CLONE_NEWPID)) return -1;
+#endif
+#ifdef CLONE_NEWNET
+    if (PyModule_AddIntMacro(m, CLONE_NEWNET)) return -1;
+#endif
+#ifdef CLONE_NEWTIME
+    if (PyModule_AddIntMacro(m, CLONE_NEWTIME)) return -1;
+#endif
+#ifdef CLONE_SYSVSEM
+    if (PyModule_AddIntMacro(m, CLONE_SYSVSEM)) return -1;
+#endif
+#ifdef CLONE_THREAD
+    if (PyModule_AddIntMacro(m, CLONE_THREAD)) return -1;
+#endif
+#ifdef CLONE_SIGHAND
+    if (PyModule_AddIntMacro(m, CLONE_SIGHAND)) return -1;
+#endif
+#ifdef CLONE_VM
+    if (PyModule_AddIntMacro(m, CLONE_VM)) return -1;
+#endif
+#endif
+
 #endif
 
 #ifdef USE_XATTRS

configure

@@ -19111,6 +19111,20 @@ fi
 done
 
 
+# check for namespace functions
+for ac_func in setns unshare
+do :
+  as_ac_var=`$as_echo "ac_cv_func_$ac_func" | $as_tr_sh`
+ac_fn_c_check_func "$LINENO" "$ac_func" "$as_ac_var"
+if eval test \"x\$"$as_ac_var"\" = x"yes"; then :
+  cat >>confdefs.h <<_ACEOF
+#define `$as_echo "HAVE_$ac_func" | $as_tr_cpp` 1
+_ACEOF
+
+fi
+done
+
+
 
 
 

configure.ac

@@ -5083,6 +5083,9 @@ AC_CHECK_FUNCS(setpgrp,
     [])
 )
 
+# check for namespace functions
+AC_CHECK_FUNCS([setns unshare])
+
 dnl We search for both crypt and crypt_r as one or the other may be defined
 dnl libxcrypt provides <crypt.h> and libcrypt with crypt_r() since
 dnl at least 3.1.1 from 2015.

pyconfig.h.in

@@ -1067,6 +1067,9 @@
 /* Define to 1 if you have the `setlocale' function. */
 #undef HAVE_SETLOCALE
 
+/* Define to 1 if you have the `setns' function. */
+#undef HAVE_SETNS
+
 /* Define to 1 if you have the `setpgid' function. */
 #undef HAVE_SETPGID
 
@@ -1437,6 +1440,9 @@
 /* Define to 1 if you have the `unlinkat' function. */
 #undef HAVE_UNLINKAT
 
+/* Define to 1 if you have the `unshare' function. */
+#undef HAVE_UNSHARE
+
 /* Define if you have a useable wchar_t type defined in wchar.h; useable means
    wchar_t must be an unsigned type with at least 16 bits. (see
    Include/unicodeobject.h). */