summaryrefslogtreecommitdiff
path: root/linuxnamespaces/__init__.py
blob: 8c1def3ba712c9d723028a1a10ab1b3f05d6da79 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
# Copyright 2024 Helmut Grohne <helmut@subdivi.de>
# SPDX-License-Identifier: GPL-3

"""Provide plumbing-layer functionality for working with Linux namespaces in
Python.
"""

import asyncio
import bisect
import contextlib
import dataclasses
import errno
import fcntl
import os
import pathlib
import socket
import stat
import struct
import subprocess
import typing

from .filedescriptor import *
from .atlocation import *
from .syscalls import *


def subidranges(
    kind: typing.Literal["uid", "gid"], login: str | None = None
) -> typing.Iterator[tuple[int, int]]:
    """Parse a `/etc/sub?id` file for ranges allocated to the given or current
    user. Return all ranges as (start, count) pairs.
    """
    if login is None:
        login = os.getlogin()
    with open(f"/etc/sub{kind}") as filelike:
        for line in filelike:
            parts = line.strip().split(":")
            if parts[0] == login:
                yield (int(parts[1]), int(parts[2]))


@dataclasses.dataclass(frozen=True)
class IDMapping:
    """Represent one range in a user or group id mapping."""

    innerstart: int
    outerstart: int
    count: int

    def __post_init__(self) -> None:
        if self.outerstart < 0:
            raise ValueError("outerstart must not be negative")
        if self.innerstart < 0:
            raise ValueError("innerstart must not be negative")
        if self.count <= 0:
            raise ValueError("count must be positive")
        if self.outerstart + self.count >= 1 << 64:
            raise ValueError("outerstart + count exceed 64bits")
        if self.innerstart + self.count >= 1 << 64:
            raise ValueError("innerstart + count exceed 64bits")


class IDAllocation:
    """This represents a subset of IDs (user or group). It can be used to
    allocate a contiguous range for use with a user namespace.
    """

    def __init__(self) -> None:
        self.ranges: list[tuple[int, int]] = []

    def add_range(self, start: int, count: int) -> None:
        """Add count ids starting from start to this allocation."""
        if start < 0 or count <= 0:
            raise ValueError("invalid range")
        index = bisect.bisect_right(self.ranges, (start, 0))
        prevrange = None
        if index > 0:
            prevrange = self.ranges[index - 1]
            if prevrange[0] + prevrange[1] > start:
                raise ValueError("attempt to add overlapping range")
        nextrange = None
        if index < len(self.ranges):
            nextrange = self.ranges[index]
            if nextrange[0] < start + count:
                raise ValueError("attempt to add overlapping range")
        if prevrange and prevrange[0] + prevrange[1] == start:
            if nextrange and nextrange[0] == start + count:
                self.ranges[index - 1] = (
                    prevrange[0],
                    prevrange[1] + count + nextrange[1],
                )
                del self.ranges[index]
            else:
                self.ranges[index - 1] = (prevrange[0], prevrange[1] + count)
        elif nextrange and nextrange[0] == start + count:
            self.ranges[index] = (start, count + nextrange[1])
        else:
            self.ranges.insert(index, (start, count))

    @classmethod
    def loadsubid(
        cls, kind: typing.Literal["uid", "gid"], login: str | None = None,
    ) -> "IDAllocation":
        """Load a `/etc/sub?id` file and return ids allocated to the given
        login or current user.
        """
        self = cls()
        for start, count in subidranges(kind, login):
            self.add_range(start, count)
        return self

    def find(self, count: int) -> int:
        """Locate count contiguous ids from this allocation. The start of
        the allocation is returned. The allocation object is left unchanged.
        """
        for start, available in self.ranges:
            if available >= count:
                return start
        raise ValueError("could not satisfy allocation request")

    def allocate(self, count: int) -> int:
        """Allocate count contiguous ids from this allocation. The start of
        the allocation is returned and the ids are removed from this
        IDAllocation object.
        """
        for index, (start, available) in enumerate(self.ranges):
            if available > count:
                self.ranges[index] = (start + count, available - count)
                return start
            if available == count:
                del self.ranges[index]
                return start
        raise ValueError("could not satisfy allocation request")

    def allocatemap(self, count: int, target: int = 0) -> IDMapping:
        """Allocate count contiguous ids from this allocation. An IDMapping
        with its innerstart set to target is returned. The allocation is
        removed from this IDAllocation object.
        """
        return IDMapping(target, self.allocate(count), count)

    def reserve(self, start: int, count: int) -> None:
        """Reserve (and remove) the given range from this allocation. If the
        range is not fully contained in this allocation, a ValueError is
        raised.
        """
        if count < 0:
            raise ValueError("negative count")
        index = bisect.bisect_right(self.ranges, (start, float("inf"))) - 1
        if index < 0:
            raise ValueError("range to reserve not found")
        cur_start, cur_count = self.ranges[index]
        assert cur_start <= start
        if cur_start == start:
            # Requested range starts at range boundary
            if cur_count < count:
                raise ValueError("range to reserve not found")
            if cur_count == count:
                # Requested range matches a range exactly
                del self.ranges[index]
            else:
                # Requested range is a head of the matched range
                self.ranges[index] = (start + count, cur_count - count)
        elif cur_start + cur_count >= start + count:
            # Requested range fits into a matched range
            self.ranges[index] = (cur_start, start - cur_start)
            if cur_start + cur_count > start + count:
                # Requested range punches a hole into a matched range
                self.ranges.insert(
                    index + 1,
                    (start + count, cur_start + cur_count - (start + count)),
                )
            # else: Requested range is a tail of a matched range
        else:
            raise ValueError("range to reserve not found")


def newidmap(
    kind: typing.Literal["uid", "gid"],
    pid: int,
    mapping: list[IDMapping],
    helper: bool | None = None,
) -> None:
    """Apply the given uid or gid mapping to the given process. A positive pid
    identifies a process, other values identify the currently running process.
    Whether setuid binaries newuidmap and newgidmap are used is determined via
    the helper argument. A None value indicate automatic detection of whether
    a helper is required for setting up the given mapping.
    """

    assert kind in ("uid", "gid")
    if pid <= 0:
        pid = os.getpid()
    if helper is None:
        # We cannot reliably test whether we have the right EUID and we don't
        # implement checking whether setgroups has been denied either. Please
        # be explicit about the helper choice in such cases.
        helper = len(mapping) > 1 or mapping[0].count > 1
    if helper:
        argv = [f"new{kind}map", str(pid)]
        for idblock in mapping:
            argv.extend(map(str, dataclasses.astuple(idblock)))
        subprocess.check_call(argv)
    else:
        pathlib.Path(f"/proc/{pid}/{kind}_map").write_text(
            "".join(
                "%d %d %d\n" % dataclasses.astuple(idblock)
                for idblock in mapping
            ),
            encoding="ascii",
        )


def newuidmap(pid: int, mapping: list[IDMapping], helper: bool = True) -> None:
    """Apply a given uid mapping to the given process. Refer to newidmap for
    details.
    """
    newidmap("uid", pid, mapping, helper)


def newgidmap(pid: int, mapping: list[IDMapping], helper: bool = True) -> None:
    """Apply a given gid mapping to the given process. Refer to newidmap for
    details.
    """
    newidmap("gid", pid, mapping, helper)


def newidmaps(
    pid: int,
    uidmapping: list[IDMapping],
    gidmapping: list[IDMapping],
    helper: bool = True,
) -> None:
    """Apply a given uid and gid mapping to the given process. Refer to
    newidmap for details.
    """
    newgidmap(pid, gidmapping, helper)
    newuidmap(pid, uidmapping, helper)


class run_in_fork:
    """Decorator for running the decorated function once in a separate process.
    """

    def __init__(self, function: typing.Callable[[], None]):
        """Fork a new process that will eventually run the given function and
        then exit.
        """
        self.efd = EventFD()
        self.pid = os.fork()
        if self.pid == 0:
            self.efd.read()
            self.efd.close()
            function()
            os._exit(0)

    def start(self) -> None:
        """Start the decorated function. It can only be started once."""
        if not self.efd:
            raise ValueError("this function can only be called once")
        self.efd.write(1)
        self.efd.close()

    def wait(self) -> None:
        """Wait for the process running the decorated function to finish."""
        if self.efd:
            raise ValueError("start must be called before wait")
        ret = os.waitpid(self.pid, 0)
        if ret != (self.pid, 0):
            raise ValueError("something failed")

    def __call__(self) -> None:
        """Start the decorated function and wait for its process to finish."""
        self.start()
        self.wait()


class async_run_in_fork:
    """Decorator for running the decorated function once in a separate process.
    Note that the decorator can only be used inside asynchronous code as it
    uses the running event loop. The decorated function instead must be
    synchronous and it must not access the event loop of the main process.
    """

    def __init__(self, function: typing.Callable[[], None]):
        """Fork a new process that will eventually run the given function and
        then exit.
        """
        loop = asyncio.get_running_loop()
        with asyncio.get_child_watcher() as watcher:
            if not watcher.is_active():
                raise RuntimeError(
                    "active child watcher required for creating a process"
                )
            self.future = loop.create_future()
            self.efd = EventFD()
            self.pid = os.fork()
            if self.pid == 0:
                self.efd.read()
                self.efd.close()
                function()
                os._exit(0)
            watcher.add_child_handler(self.pid, self._child_callback)

    def _child_callback(self, pid: int, returncode: int) -> None:
        if self.pid != pid:
            return
        self.future.set_result(returncode)

    def start(self) -> None:
        """Start the decorated function. It can only be started once."""
        if not self.efd:
            raise ValueError("this function can only be called once")
        self.efd.write(1)
        self.efd.close()

    async def wait(self) -> None:
        """Wait for the process running the decorated function to finish."""
        if self.efd:
            raise ValueError("start must be called before wait")
        ret = await self.future
        if ret != 0:
            raise ValueError("something failed")

    async def __call__(self) -> None:
        """Start the decorated function and wait for its process to finish."""
        self.start()
        await self.wait()


def bind_mount(
    source: AtLocationLike,
    target: AtLocationLike,
    recursive: bool = False,
    readonly: bool = False,
) -> None:
    """Create a bind mount from source to target. Depending on whether one of
    the locations involves a file descriptor or not, the new or old mount API
    will be used.
    """
    source = AtLocation(source)
    target = AtLocation(target)
    try:
        if readonly:
            # We would have to remount to apply the readonly flag, see
            # https://git.kernel.org/pub/scm/utils/util-linux/util-linux.git/commit/?id=9ac77b8a78452eab0612523d27fee52159f5016a
            raise ValueError()
        srcloc = os.fspath(source)
        tgtloc = os.fspath(target)
    except ValueError:
        otflags = OpenTreeFlags.OPEN_TREE_CLONE
        if recursive:
            otflags |= OpenTreeFlags.AT_RECURSIVE
        with open_tree(source, otflags) as srcfd:
            if readonly:
                mount_setattr(srcfd, recursive, MountAttrFlags.RDONLY)
            move_mount(srcfd, target)
    else:
        mflags = MountFlags.BIND
        if recursive:
            mflags |= MountFlags.REC
        mount(srcloc, tgtloc, None, mflags)


_P = typing.ParamSpec("_P")

class _ExceptionExitCallback:
    """Helper class that invokes a callback when a context manager exists with
    a failure.
    """

    def __init__(
        self,
        callback: typing.Callable[_P, typing.Any],
        *args: _P.args,
        **kwargs: _P.kwargs,
    ) -> None:
        self.callback = callback
        self.args = args
        self.kwargs = kwargs

    def __enter__(self) -> None:
        pass

    def __exit__(
        self,
        exc_type: type[BaseException] | None,
        exc_value: BaseException | None,
        traceback: typing.Any,
    ) -> None:
        if exc_type is not None:
            self.callback(*self.args, **self.kwargs)


def populate_dev(
    origroot: AtLocationLike,
    newroot: PathConvertible,
    *,
    fuse: bool = True,
    pidns: bool = True,
    tun: bool = True,
) -> None:
    """Mount a tmpfs to the dev directory beneath newroot and populate it with
    basic devices by bind mounting them from the dev directory beneath
    origroot. Also mount a new pts instance.

    Even though a CAP_SYS_ADMIN-enabled process can umount components of the
    /dev hierarchy, they they cannot gain privileges in doing so as no
    hierarchies are restricted via tmpfs mounts or read-only bind mounts.
    """
    origdev = AtLocation(origroot) / "dev"
    newdev = AtLocation(newroot) / "dev"
    directories = {"pts", "shm"}
    files = set()
    symlinks = {
        "fd": "/proc/self/fd",
        "stdin": "/proc/self/fd/0",
        "stdout": "/proc/self/fd/1",
        "stderr": "/proc/self/fd/2",
    }
    bind_mounts: dict[str, AtLocation] = {}
    with contextlib.ExitStack() as exitstack:
        for fn in "null zero full random urandom tty".split():
            files.add(fn)
            bind_mounts[fn] = exitstack.enter_context(
                open_tree(origdev / fn, OpenTreeFlags.OPEN_TREE_CLONE)
            )
        if fuse:
            files.add("fuse")
            bind_mounts["fuse"] = exitstack.enter_context(
                open_tree(origdev / "fuse", OpenTreeFlags.OPEN_TREE_CLONE)
            )
        if pidns:
            symlinks["ptmx"] = "pts/ptmx"
        else:
            bind_mounts["pts"] = exitstack.enter_context(
                open_tree(
                    origdev / "pts",
                    OpenTreeFlags.AT_RECURSIVE | OpenTreeFlags.OPEN_TREE_CLONE,
                )
            )
            files.add("ptmx")
            bind_mounts["ptmx"] = exitstack.enter_context(
                open_tree(origdev / "ptmx", OpenTreeFlags.OPEN_TREE_CLONE)
            )
        if tun:
            directories.add("net")
            files.add("net/tun")
            bind_mounts["net/tun"] = exitstack.enter_context(
                open_tree(origdev / "net/tun", OpenTreeFlags.OPEN_TREE_CLONE)
            )
        mount(
            "devtmpfs",
            newdev,
            "tmpfs",
            MountFlags.NOSUID | MountFlags.NOEXEC,
            "mode=0755",
        )
        exitstack.enter_context(
            _ExceptionExitCallback(umount, newdev, UmountFlags.DETACH)
        )
        for fn in directories:
            (newdev / fn).mkdir()
            (newdev / fn).chmod(0o755)
        mount(
            "tmpfs",
            newdev / "shm",
            "tmpfs",
            MountFlags.NOSUID | MountFlags.NODEV,
        )
        for fn in files:
            (newdev / fn).mknod(stat.S_IFREG)
        for fn, target in symlinks.items():
            (newdev / fn).symlink_to(target)
        if pidns:
            mount(
                "devpts",
                newdev / "pts",
                "devpts",
                MountFlags.NOSUID | MountFlags.NOEXEC,
                "gid=5,mode=620,ptmxmode=666",
            )
        for fn, fd in bind_mounts.items():
            move_mount(fd, newdev / fn)


def populate_proc(
    origroot: AtLocationLike,
    newroot: PathConvertible,
    namespaces: CloneFlags,
) -> None:
    """Mount a /proc hierarchy.

    Note that a user with CAP_SYS_ADMIN can change read-only bind mounts.
    Still those bind mounts provide guidance to the container as to which
    aspects it should manage.
    """
    assert namespaces & CloneFlags.NEWNS == CloneFlags.NEWNS
    newproc = AtLocation(newroot) / "proc"
    rwns = CloneFlags.NEWPID
    if namespaces & rwns == rwns:
        mount(
            "proc",
            newproc,
            "proc",
            MountFlags.NOSUID | MountFlags.NODEV | MountFlags.NOEXEC,
        )
    else:
        bind_mount(AtLocation(origroot) / "proc", newproc, True)
    with _ExceptionExitCallback(umount, newproc, UmountFlags.DETACH):
        rwns |= CloneFlags.NEWUSER | CloneFlags.NEWIPC | CloneFlags.NEWUTS
        if namespaces & rwns != rwns:
            psn: AtLocation | None = None
            if namespaces & CloneFlags.NEWNET == CloneFlags.NEWNET:
                psn = open_tree(
                    newproc / "sys/net",
                    OpenTreeFlags.OPEN_TREE_CLONE | OpenTreeFlags.AT_RECURSIVE,
                )
            bind_mount(newproc / "sys", newproc / "sys", True, True)
            if psn is not None:
                move_mount(psn, newproc / "sys/net")
        elif namespaces & CloneFlags.NEWNET != CloneFlags.NEWNET:
            bind_mount(newproc / "sys/net", newproc / "sys/net", True, True)


def populate_sys(
    origroot: AtLocationLike,
    newroot: PathConvertible,
    namespaces: CloneFlags,
    rootcgroup: PathConvertible | None = None,
    module: bool = True,
    devices: bool = False,
) -> None:
    """Create a /sys hierarchy below newroot. Bind the cgroup hierarchy. The
    cgroup hierarchy will be mounted read-only if mounting the root group.

    The module parameter indicates whether the /sys/module should be made
    available read-only (True) or not at all (False). The devices parameter
    indicates whether the devices hierarchy should be made available. If the
    given namespaces happen to include a network namespace, virtual network
    devices will be modifiable.

    A process with CAP_SYS_ADMIN can remount the created bind mounts read-write
    or umount hiding mounts and thus elevate their privileges.
    """
    newsys = AtLocation(newroot) / "sys"
    mflags = MountFlags.NOSUID | MountFlags.NOEXEC | MountFlags.NODEV
    bind_mounts: dict[
        str, tuple[typing.Union[str, pathlib.PurePath], bool]
    ] = {}
    if rootcgroup is None:
        bind_mounts["fs/cgroup"] = ("fs/cgroup", True)
    else:
        bind_mounts["fs/cgroup"] = (
            "fs/cgroup" / pathlib.PurePath(rootcgroup).relative_to("/"), False
        )
    if module:
        bind_mounts["module"] = ("module", True)
    if devices:
        for subdir in ("bus", "class", "dev", "devices"):
            bind_mounts[subdir] = (subdir, True)
    if namespaces & CloneFlags.NEWNET:
        if not devices:
            bind_mounts["class/net"] = ("class/net", True)
        bind_mounts["devices/virtual/net"] = ("devices/virtual/net", False)

    bind_fds: dict[str, AtLocation] = {}
    with contextlib.ExitStack() as exitstack:
        for target, (source, rdonly) in bind_mounts.items():
            bindfd = exitstack.enter_context(
                open_tree(
                    AtLocation(origroot) / "sys" / source,
                    OpenTreeFlags.OPEN_TREE_CLONE | OpenTreeFlags.AT_RECURSIVE,
                ),
            )
            if rdonly:
                mount_setattr(bindfd, True, MountAttrFlags.RDONLY)
            bind_fds[target] = bindfd

        mount("sysfs", newsys, "tmpfs", mflags, "mode=0755")

        exitstack.enter_context(
            _ExceptionExitCallback(umount, newsys, UmountFlags.DETACH)
        )
        dirs = set()
        for subdir in bind_fds:
            dirs.add(subdir)
            while "/" in subdir:
                subdir = subdir.rsplit("/", 1)[0]
                dirs.add(subdir)
        for subdir in sorted(dirs):
            (newsys / subdir).mkdir()
            (newsys / subdir).chmod(0o755)
        mflags |= MountFlags.REMOUNT | MountFlags.RDONLY
        mount("sysfs", newsys, "tmpfs", mflags, "mode=0755")
        for subdir, bindfd in sorted(bind_fds.items()):
            move_mount(bindfd, newsys / subdir)


def unshare_user_idmap(
    uidmap: list[IDMapping],
    gidmap: list[IDMapping],
    flags: CloneFlags = CloneFlags.NEWUSER,
) -> None:
    """Unshare the given namespaces (must include user) and set up the given
    id mappings.
    """
    pid = os.getpid()
    @run_in_fork
    def setup_idmaps() -> None:
        newidmaps(pid, uidmap, gidmap)
    unshare(flags)
    setup_idmaps()

def unshare_user_idmap_nohelper(
    uid: int, gid: int, flags: CloneFlags = CloneFlags.NEWUSER
) -> None:
    """Unshare the given namespaces (must include user) and
    map the current user and group to the given uid and gid
    without using the setuid helpers.
    """
    uidmap = IDMapping(uid, os.getuid(), 1)
    gidmap = IDMapping(gid, os.getgid(), 1)
    unshare(flags)
    pathlib.Path("/proc/self/setgroups").write_bytes(b"deny")
    newidmaps(-1, [uidmap], [gidmap], False)


class _AsyncFilesender:
    bs = 65536

    def __init__(self, from_fd: int, to_fd: int, count: int | None = None):
        self.from_fd = from_fd
        self.to_fd = to_fd
        self.copied = 0
        self.remain = count
        self.loop = asyncio.get_running_loop()
        self.fut: asyncio.Future[int] = self.loop.create_future()
        self.loop.add_writer(self.to_fd, self.handle_write)

    def handle_write(self) -> None:
        try:
            ret = os.sendfile(
                self.to_fd,
                self.from_fd,
                None,
                self.bs if self.remain is None else min(self.bs, self.remain),
            )
        except OSError as err:
            if err.errno != errno.EAGAIN:
                self.loop.remove_writer(self.to_fd)
                self.fut.set_exception(err)
        else:
            self.copied += ret
            if self.remain is not None:
                self.remain -= ret
            if ret == 0 or self.remain == 0:
                self.loop.remove_writer(self.to_fd)
                self.fut.set_result(self.copied)


class _AsyncSplicer:
    bs = 65536

    def __init__(self, from_fd: int, to_fd: int, count: int | None = None):
        self.from_fd = from_fd
        self.to_fd = to_fd
        self.copied = 0
        self.remain = count
        self.wait_read = True
        self.loop = asyncio.get_running_loop()
        self.fut: asyncio.Future[int] = self.loop.create_future()
        self.loop.add_reader(self.from_fd, self.handle_io)

    def handle_io(self) -> None:
        try:
            ret = os.splice(
                self.from_fd,
                self.to_fd,
                self.bs if self.remain is None else min(self.bs, self.remain),
                flags=os.SPLICE_F_NONBLOCK,
            )
        except OSError as err:
            if err.errno == errno.EAGAIN:
                self.wait_read = not self.wait_read
                if self.wait_read:
                    self.loop.remove_writer(self.to_fd)
                    self.loop.add_reader(self.from_fd, self.handle_io)
                else:
                    self.loop.remove_reader(self.from_fd)
                    self.loop.add_writer(self.to_fd, self.handle_io)
            else:
                self.loop.remove_reader(self.from_fd)
                self.loop.remove_writer(self.to_fd)
                self.fut.set_exception(err)
        else:
            self.copied += ret
            if self.remain is not None:
                self.remain -= ret
            if ret == 0 or self.remain == 0:
                self.loop.remove_reader(self.from_fd)
                self.loop.remove_writer(self.to_fd)
                self.fut.set_result(self.copied)


class _AsyncCopier:
    bs = 65536

    def __init__(self, from_fd: int, to_fd: int, count: int | None = None):
        self.from_fd = from_fd
        self.to_fd = to_fd
        self.buffer = b""
        self.copied = 0  # bytes read and written
        self.remain = count  # remaining bytes not yet read
        # eof can be an exception when a read failed and otherwise indicates
        # whether a read returned 0.
        self.eof: bool | OSError = False
        self.loop = asyncio.get_running_loop()
        self.fut: asyncio.Future[int] = self.loop.create_future()
        self.loop.add_reader(self.from_fd, self.handle_readable)

    def handle_readable(self) -> None:
        try:
            data = os.read(
                self.from_fd,
                self.bs if self.remain is None else min(self.bs, self.remain),
            )
        except OSError as err:
            if err.errno != errno.EAGAIN:
                self.loop.remove_reader(self.from_fd)
                if self.buffer:
                    self.eof = err
                else:
                    self.fut.set_exception(err)
        else:
            if data:
                if self.remain is not None:
                    self.remain -= len(data)
                self.buffer += data
                if len(self.buffer) == len(data):
                    self.loop.add_writer(self.to_fd, self.handle_writeable)
                if self.remain == 0 or len(self.buffer) >= self.bs:
                    self.loop.remove_reader(self.from_fd)
            else:
                self.eof = True
                self.loop.remove_reader(self.from_fd)

    def handle_writeable(self) -> None:
        try:
            written = os.write(self.to_fd, self.buffer)
        except OSError as err:
            if err.errno != errno.EAGAIN:
                self.loop.remove_writer(self.to_fd)
                if isinstance(self.eof, OSError):
                    self.fut.set_exception(self.eof)
                else:
                    self.loop.remove_reader(self.from_fd)
                    self.fut.set_exception(err)
        else:
            self.buffer = self.buffer[written:]
            self.copied += written
            if not self.buffer:
                self.loop.remove_writer(self.to_fd)
                if self.eof is True or self.remain == 0:
                    self.fut.set_result(self.copied)
                elif isinstance(self.eof, OSError):
                    self.fut.set_exception(self.eof)
            elif not self.eof and self.remain and len(self.buffer) < self.bs:
                self.loop.add_reader(self.from_fd, self.handle_readable)


def async_copyfd(
    from_fd: int, to_fd: int, count: int | None = None
) -> asyncio.Future[int]:
    """Copy the given number of bytes from the first file descriptor to the
    second file descriptor in an asyncio context. Both copies are performed
    binary. An efficient implementation is chosen depending on the file type
    of file descriptors.
    """
    from_mode = os.fstat(from_fd).st_mode
    if stat.S_ISREG(from_mode):
        return _AsyncFilesender(from_fd, to_fd, count).fut
    if stat.S_ISFIFO(from_mode) or stat.S_ISFIFO(os.fstat(to_fd).st_mode):
        return _AsyncSplicer(from_fd, to_fd, count).fut
    return _AsyncCopier(from_fd, to_fd, count).fut


class _AsyncPidfdWaiter:
    def __init__(self, pidfd: int, flags: int):
        self.pidfd = pidfd
        self.flags = flags
        self.loop = asyncio.get_running_loop()
        self.fut: asyncio.Future[
            os.waitid_result | None
        ] = self.loop.create_future()
        self.loop.add_reader(pidfd, self.handle_readable)

    def handle_readable(self) -> None:
        try:
            result = os.waitid(os.P_PIDFD, self.pidfd, self.flags)
        except OSError as err:
            if err.errno != errno.EAGAIN:
                self.loop.remove_reader(self.pidfd)
                self.fut.set_exception(err)
        except Exception as err:
            self.loop.remove_reader(self.pidfd)
            self.fut.set_exception(err)
        else:
            self.loop.remove_reader(self.pidfd)
            self.fut.set_result(result)


def async_waitpidfd(
    pidfd: int, flags: int
) -> asyncio.Future[os.waitid_result | None]:
    """Asynchronously wait for a process represented as a pidfd. This is an
    async variant of waitid(P_PIDFD, pidfd, flags).
    """
    return _AsyncPidfdWaiter(pidfd, flags).fut


def enable_loopback_if() -> None:
    """Enable the loopback network interface that is initially down in a new
    network namespace.
    """
    # We us the old and deprecated ioctl API rather than netlink, because it
    # is way simpler and good enough for our purpose. The interface is always
    # created as "lo" by the kernel and it'll have loopback addresses
    # configured automatically. All that we have to do is "up" it.
    SIOCSIFFLAGS = 0x8914
    IFF_UP = 1
    with socket.socket(socket.AF_INET, socket.SOCK_DGRAM, 0) as sock:
        fcntl.ioctl(sock, SIOCSIFFLAGS, struct.pack("@16sH", b"lo", IFF_UP))