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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089
|
// +build linux
package libcontainer
import (
"fmt"
"io"
"io/ioutil"
"os"
"os/exec"
"path"
"path/filepath"
"strings"
"time"
securejoin "github.com/cyphar/filepath-securejoin"
"github.com/moby/sys/mountinfo"
"github.com/mrunalp/fileutils"
"github.com/opencontainers/runc/libcontainer/cgroups"
"github.com/opencontainers/runc/libcontainer/cgroups/fs2"
"github.com/opencontainers/runc/libcontainer/configs"
"github.com/opencontainers/runc/libcontainer/devices"
"github.com/opencontainers/runc/libcontainer/userns"
"github.com/opencontainers/runc/libcontainer/utils"
"github.com/opencontainers/runtime-spec/specs-go"
"github.com/opencontainers/selinux/go-selinux/label"
"github.com/sirupsen/logrus"
"golang.org/x/sys/unix"
)
const defaultMountFlags = unix.MS_NOEXEC | unix.MS_NOSUID | unix.MS_NODEV
type mountConfig struct {
root string
label string
cgroup2Path string
rootlessCgroups bool
cgroupns bool
}
// needsSetupDev returns true if /dev needs to be set up.
func needsSetupDev(config *configs.Config) bool {
for _, m := range config.Mounts {
if m.Device == "bind" && utils.CleanPath(m.Destination) == "/dev" {
return false
}
}
return true
}
// prepareRootfs sets up the devices, mount points, and filesystems for use
// inside a new mount namespace. It doesn't set anything as ro. You must call
// finalizeRootfs after this function to finish setting up the rootfs.
func prepareRootfs(pipe io.ReadWriter, iConfig *initConfig) (err error) {
config := iConfig.Config
if err := prepareRoot(config); err != nil {
return newSystemErrorWithCause(err, "preparing rootfs")
}
mountConfig := &mountConfig{
root: config.Rootfs,
label: config.MountLabel,
cgroup2Path: iConfig.Cgroup2Path,
rootlessCgroups: iConfig.RootlessCgroups,
cgroupns: config.Namespaces.Contains(configs.NEWCGROUP),
}
setupDev := needsSetupDev(config)
for _, m := range config.Mounts {
for _, precmd := range m.PremountCmds {
if err := mountCmd(precmd); err != nil {
return newSystemErrorWithCause(err, "running premount command")
}
}
if err := mountToRootfs(m, mountConfig); err != nil {
return newSystemErrorWithCausef(err, "mounting %q to rootfs at %q", m.Source, m.Destination)
}
for _, postcmd := range m.PostmountCmds {
if err := mountCmd(postcmd); err != nil {
return newSystemErrorWithCause(err, "running postmount command")
}
}
}
if setupDev {
if err := createDevices(config); err != nil {
return newSystemErrorWithCause(err, "creating device nodes")
}
if err := setupPtmx(config); err != nil {
return newSystemErrorWithCause(err, "setting up ptmx")
}
if err := setupDevSymlinks(config.Rootfs); err != nil {
return newSystemErrorWithCause(err, "setting up /dev symlinks")
}
}
// Signal the parent to run the pre-start hooks.
// The hooks are run after the mounts are setup, but before we switch to the new
// root, so that the old root is still available in the hooks for any mount
// manipulations.
// Note that iConfig.Cwd is not guaranteed to exist here.
if err := syncParentHooks(pipe); err != nil {
return err
}
// The reason these operations are done here rather than in finalizeRootfs
// is because the console-handling code gets quite sticky if we have to set
// up the console before doing the pivot_root(2). This is because the
// Console API has to also work with the ExecIn case, which means that the
// API must be able to deal with being inside as well as outside the
// container. It's just cleaner to do this here (at the expense of the
// operation not being perfectly split).
if err := unix.Chdir(config.Rootfs); err != nil {
return newSystemErrorWithCausef(err, "changing dir to %q", config.Rootfs)
}
s := iConfig.SpecState
s.Pid = unix.Getpid()
s.Status = specs.StateCreating
if err := iConfig.Config.Hooks[configs.CreateContainer].RunHooks(s); err != nil {
return err
}
if config.NoPivotRoot {
err = msMoveRoot(config.Rootfs)
} else if config.Namespaces.Contains(configs.NEWNS) {
err = pivotRoot(config.Rootfs)
} else {
err = chroot()
}
if err != nil {
return newSystemErrorWithCause(err, "jailing process inside rootfs")
}
if setupDev {
if err := reOpenDevNull(); err != nil {
return newSystemErrorWithCause(err, "reopening /dev/null inside container")
}
}
if cwd := iConfig.Cwd; cwd != "" {
// Note that spec.Process.Cwd can contain unclean value like "../../../../foo/bar...".
// However, we are safe to call MkDirAll directly because we are in the jail here.
if err := os.MkdirAll(cwd, 0o755); err != nil {
return err
}
}
return nil
}
// finalizeRootfs sets anything to ro if necessary. You must call
// prepareRootfs first.
func finalizeRootfs(config *configs.Config) (err error) {
// All tmpfs mounts and /dev were previously mounted as rw
// by mountPropagate. Remount them read-only as requested.
for _, m := range config.Mounts {
if m.Flags&unix.MS_RDONLY != unix.MS_RDONLY {
continue
}
if m.Device == "tmpfs" || utils.CleanPath(m.Destination) == "/dev" {
if err := remountReadonly(m); err != nil {
return newSystemErrorWithCausef(err, "remounting %q as readonly", m.Destination)
}
}
}
// set rootfs ( / ) as readonly
if config.Readonlyfs {
if err := setReadonly(); err != nil {
return newSystemErrorWithCause(err, "setting rootfs as readonly")
}
}
if config.Umask != nil {
unix.Umask(int(*config.Umask))
} else {
unix.Umask(0o022)
}
return nil
}
// /tmp has to be mounted as private to allow MS_MOVE to work in all situations
func prepareTmp(topTmpDir string) (string, error) {
tmpdir, err := ioutil.TempDir(topTmpDir, "runctop")
if err != nil {
return "", err
}
if err := unix.Mount(tmpdir, tmpdir, "bind", unix.MS_BIND, ""); err != nil {
return "", err
}
if err := unix.Mount("", tmpdir, "", uintptr(unix.MS_PRIVATE), ""); err != nil {
return "", err
}
return tmpdir, nil
}
func cleanupTmp(tmpdir string) {
_ = unix.Unmount(tmpdir, 0)
_ = os.RemoveAll(tmpdir)
}
func mountCmd(cmd configs.Command) error {
command := exec.Command(cmd.Path, cmd.Args[:]...)
command.Env = cmd.Env
command.Dir = cmd.Dir
if out, err := command.CombinedOutput(); err != nil {
return fmt.Errorf("%#v failed: %s: %v", cmd, string(out), err)
}
return nil
}
func prepareBindMount(m *configs.Mount, rootfs string) error {
stat, err := os.Stat(m.Source)
if err != nil {
// error out if the source of a bind mount does not exist as we will be
// unable to bind anything to it.
return err
}
// ensure that the destination of the bind mount is resolved of symlinks at mount time because
// any previous mounts can invalidate the next mount's destination.
// this can happen when a user specifies mounts within other mounts to cause breakouts or other
// evil stuff to try to escape the container's rootfs.
var dest string
if dest, err = securejoin.SecureJoin(rootfs, m.Destination); err != nil {
return err
}
if err := checkProcMount(rootfs, dest, m.Source); err != nil {
return err
}
if err := createIfNotExists(dest, stat.IsDir()); err != nil {
return err
}
return nil
}
func mountCgroupV1(m *configs.Mount, c *mountConfig) error {
binds, err := getCgroupMounts(m)
if err != nil {
return err
}
var merged []string
for _, b := range binds {
ss := filepath.Base(b.Destination)
if strings.Contains(ss, ",") {
merged = append(merged, ss)
}
}
tmpfs := &configs.Mount{
Source: "tmpfs",
Device: "tmpfs",
Destination: m.Destination,
Flags: defaultMountFlags,
Data: "mode=755",
PropagationFlags: m.PropagationFlags,
}
if err := mountToRootfs(tmpfs, c); err != nil {
return err
}
for _, b := range binds {
if c.cgroupns {
subsystemPath := filepath.Join(c.root, b.Destination)
if err := os.MkdirAll(subsystemPath, 0o755); err != nil {
return err
}
if err := utils.WithProcfd(c.root, b.Destination, func(procfd string) error {
flags := defaultMountFlags
if m.Flags&unix.MS_RDONLY != 0 {
flags = flags | unix.MS_RDONLY
}
var (
source = "cgroup"
data = filepath.Base(subsystemPath)
)
if data == "systemd" {
data = cgroups.CgroupNamePrefix + data
source = "systemd"
}
return unix.Mount(source, procfd, "cgroup", uintptr(flags), data)
}); err != nil {
return err
}
} else {
if err := mountToRootfs(b, c); err != nil {
return err
}
}
}
for _, mc := range merged {
for _, ss := range strings.Split(mc, ",") {
// symlink(2) is very dumb, it will just shove the path into
// the link and doesn't do any checks or relative path
// conversion. Also, don't error out if the cgroup already exists.
if err := os.Symlink(mc, filepath.Join(c.root, m.Destination, ss)); err != nil && !os.IsExist(err) {
return err
}
}
}
return nil
}
func mountCgroupV2(m *configs.Mount, c *mountConfig) error {
dest, err := securejoin.SecureJoin(c.root, m.Destination)
if err != nil {
return err
}
if err := os.MkdirAll(dest, 0o755); err != nil {
return err
}
return utils.WithProcfd(c.root, m.Destination, func(procfd string) error {
if err := unix.Mount(m.Source, procfd, "cgroup2", uintptr(m.Flags), m.Data); err != nil {
// when we are in UserNS but CgroupNS is not unshared, we cannot mount cgroup2 (#2158)
if err == unix.EPERM || err == unix.EBUSY {
src := fs2.UnifiedMountpoint
if c.cgroupns && c.cgroup2Path != "" {
// Emulate cgroupns by bind-mounting
// the container cgroup path rather than
// the whole /sys/fs/cgroup.
src = c.cgroup2Path
}
err = unix.Mount(src, procfd, "", uintptr(m.Flags)|unix.MS_BIND, "")
if err == unix.ENOENT && c.rootlessCgroups {
err = nil
}
}
return err
}
return nil
})
}
func doTmpfsCopyUp(m *configs.Mount, rootfs, mountLabel string) (Err error) {
// Set up a scratch dir for the tmpfs on the host.
tmpdir, err := prepareTmp("/tmp")
if err != nil {
return newSystemErrorWithCause(err, "tmpcopyup: failed to setup tmpdir")
}
defer cleanupTmp(tmpdir)
tmpDir, err := ioutil.TempDir(tmpdir, "runctmpdir")
if err != nil {
return newSystemErrorWithCause(err, "tmpcopyup: failed to create tmpdir")
}
defer os.RemoveAll(tmpDir)
// Configure the *host* tmpdir as if it's the container mount. We change
// m.Destination since we are going to mount *on the host*.
oldDest := m.Destination
m.Destination = tmpDir
err = mountPropagate(m, "/", mountLabel)
m.Destination = oldDest
if err != nil {
return err
}
defer func() {
if Err != nil {
if err := unix.Unmount(tmpDir, unix.MNT_DETACH); err != nil {
logrus.Warnf("tmpcopyup: failed to unmount tmpdir on error: %v", err)
}
}
}()
return utils.WithProcfd(rootfs, m.Destination, func(procfd string) (Err error) {
// Copy the container data to the host tmpdir. We append "/" to force
// CopyDirectory to resolve the symlink rather than trying to copy the
// symlink itself.
if err := fileutils.CopyDirectory(procfd+"/", tmpDir); err != nil {
return fmt.Errorf("tmpcopyup: failed to copy %s to %s (%s): %w", m.Destination, procfd, tmpDir, err)
}
// Now move the mount into the container.
if err := unix.Mount(tmpDir, procfd, "", unix.MS_MOVE, ""); err != nil {
return fmt.Errorf("tmpcopyup: failed to move mount %s to %s (%s): %w", tmpDir, procfd, m.Destination, err)
}
return nil
})
}
func mountToRootfs(m *configs.Mount, c *mountConfig) error {
rootfs := c.root
mountLabel := c.label
dest, err := securejoin.SecureJoin(rootfs, m.Destination)
if err != nil {
return err
}
switch m.Device {
case "proc", "sysfs":
// If the destination already exists and is not a directory, we bail
// out This is to avoid mounting through a symlink or similar -- which
// has been a "fun" attack scenario in the past.
// TODO: This won't be necessary once we switch to libpathrs and we can
// stop all of these symlink-exchange attacks.
if fi, err := os.Lstat(dest); err != nil {
if !os.IsNotExist(err) {
return err
}
} else if fi.Mode()&os.ModeDir == 0 {
return fmt.Errorf("filesystem %q must be mounted on ordinary directory", m.Device)
}
if err := os.MkdirAll(dest, 0o755); err != nil {
return err
}
// Selinux kernels do not support labeling of /proc or /sys
return mountPropagate(m, rootfs, "")
case "mqueue":
if err := os.MkdirAll(dest, 0o755); err != nil {
return err
}
if err := mountPropagate(m, rootfs, ""); err != nil {
return err
}
return label.SetFileLabel(dest, mountLabel)
case "tmpfs":
stat, err := os.Stat(dest)
if err != nil {
if err := os.MkdirAll(dest, 0o755); err != nil {
return err
}
}
if m.Extensions&configs.EXT_COPYUP == configs.EXT_COPYUP {
err = doTmpfsCopyUp(m, rootfs, mountLabel)
} else {
err = mountPropagate(m, rootfs, mountLabel)
}
if err != nil {
return err
}
if stat != nil {
if err = os.Chmod(dest, stat.Mode()); err != nil {
return err
}
}
return nil
case "bind":
if err := prepareBindMount(m, rootfs); err != nil {
return err
}
if err := mountPropagate(m, rootfs, mountLabel); err != nil {
return err
}
// bind mount won't change mount options, we need remount to make mount options effective.
// first check that we have non-default options required before attempting a remount
if m.Flags&^(unix.MS_REC|unix.MS_REMOUNT|unix.MS_BIND) != 0 {
// only remount if unique mount options are set
if err := remount(m, rootfs); err != nil {
return err
}
}
if m.Relabel != "" {
if err := label.Validate(m.Relabel); err != nil {
return err
}
shared := label.IsShared(m.Relabel)
if err := label.Relabel(m.Source, mountLabel, shared); err != nil {
return err
}
}
case "cgroup":
if cgroups.IsCgroup2UnifiedMode() {
return mountCgroupV2(m, c)
}
return mountCgroupV1(m, c)
default:
if err := checkProcMount(rootfs, dest, m.Source); err != nil {
return err
}
if err := os.MkdirAll(dest, 0o755); err != nil {
return err
}
return mountPropagate(m, rootfs, mountLabel)
}
return nil
}
func getCgroupMounts(m *configs.Mount) ([]*configs.Mount, error) {
mounts, err := cgroups.GetCgroupMounts(false)
if err != nil {
return nil, err
}
cgroupPaths, err := cgroups.ParseCgroupFile("/proc/self/cgroup")
if err != nil {
return nil, err
}
var binds []*configs.Mount
for _, mm := range mounts {
dir, err := mm.GetOwnCgroup(cgroupPaths)
if err != nil {
return nil, err
}
relDir, err := filepath.Rel(mm.Root, dir)
if err != nil {
return nil, err
}
binds = append(binds, &configs.Mount{
Device: "bind",
Source: filepath.Join(mm.Mountpoint, relDir),
Destination: filepath.Join(m.Destination, filepath.Base(mm.Mountpoint)),
Flags: unix.MS_BIND | unix.MS_REC | m.Flags,
PropagationFlags: m.PropagationFlags,
})
}
return binds, nil
}
// checkProcMount checks to ensure that the mount destination is not over the top of /proc.
// dest is required to be an abs path and have any symlinks resolved before calling this function.
//
// if source is nil, don't stat the filesystem. This is used for restore of a checkpoint.
func checkProcMount(rootfs, dest, source string) error {
const procPath = "/proc"
path, err := filepath.Rel(filepath.Join(rootfs, procPath), dest)
if err != nil {
return err
}
// pass if the mount path is located outside of /proc
if strings.HasPrefix(path, "..") {
return nil
}
if path == "." {
// an empty source is pasted on restore
if source == "" {
return nil
}
// only allow a mount on-top of proc if it's source is "proc"
isproc, err := isProc(source)
if err != nil {
return err
}
// pass if the mount is happening on top of /proc and the source of
// the mount is a proc filesystem
if isproc {
return nil
}
return fmt.Errorf("%q cannot be mounted because it is not of type proc", dest)
}
// Here dest is definitely under /proc. Do not allow those,
// except for a few specific entries emulated by lxcfs.
validProcMounts := []string{
"/proc/cpuinfo",
"/proc/diskstats",
"/proc/meminfo",
"/proc/stat",
"/proc/swaps",
"/proc/uptime",
"/proc/loadavg",
"/proc/slabinfo",
"/proc/net/dev",
}
for _, valid := range validProcMounts {
path, err := filepath.Rel(filepath.Join(rootfs, valid), dest)
if err != nil {
return err
}
if path == "." {
return nil
}
}
return fmt.Errorf("%q cannot be mounted because it is inside /proc", dest)
}
func isProc(path string) (bool, error) {
var s unix.Statfs_t
if err := unix.Statfs(path, &s); err != nil {
return false, err
}
return s.Type == unix.PROC_SUPER_MAGIC, nil
}
func setupDevSymlinks(rootfs string) error {
links := [][2]string{
{"/proc/self/fd", "/dev/fd"},
{"/proc/self/fd/0", "/dev/stdin"},
{"/proc/self/fd/1", "/dev/stdout"},
{"/proc/self/fd/2", "/dev/stderr"},
}
// kcore support can be toggled with CONFIG_PROC_KCORE; only create a symlink
// in /dev if it exists in /proc.
if _, err := os.Stat("/proc/kcore"); err == nil {
links = append(links, [2]string{"/proc/kcore", "/dev/core"})
}
for _, link := range links {
var (
src = link[0]
dst = filepath.Join(rootfs, link[1])
)
if err := os.Symlink(src, dst); err != nil && !os.IsExist(err) {
return fmt.Errorf("symlink %s %s %s", src, dst, err)
}
}
return nil
}
// If stdin, stdout, and/or stderr are pointing to `/dev/null` in the parent's rootfs
// this method will make them point to `/dev/null` in this container's rootfs. This
// needs to be called after we chroot/pivot into the container's rootfs so that any
// symlinks are resolved locally.
func reOpenDevNull() error {
var stat, devNullStat unix.Stat_t
file, err := os.OpenFile("/dev/null", os.O_RDWR, 0)
if err != nil {
return fmt.Errorf("Failed to open /dev/null - %s", err)
}
defer file.Close() //nolint: errcheck
if err := unix.Fstat(int(file.Fd()), &devNullStat); err != nil {
return err
}
for fd := 0; fd < 3; fd++ {
if err := unix.Fstat(fd, &stat); err != nil {
return err
}
if stat.Rdev == devNullStat.Rdev {
// Close and re-open the fd.
if err := unix.Dup3(int(file.Fd()), fd, 0); err != nil {
return err
}
}
}
return nil
}
// Create the device nodes in the container.
func createDevices(config *configs.Config) error {
useBindMount := userns.RunningInUserNS() || config.Namespaces.Contains(configs.NEWUSER)
oldMask := unix.Umask(0o000)
for _, node := range config.Devices {
// The /dev/ptmx device is setup by setupPtmx()
if utils.CleanPath(node.Path) == "/dev/ptmx" {
continue
}
// containers running in a user namespace are not allowed to mknod
// devices so we can just bind mount it from the host.
if err := createDeviceNode(config.Rootfs, node, useBindMount); err != nil {
unix.Umask(oldMask)
return err
}
}
unix.Umask(oldMask)
return nil
}
func bindMountDeviceNode(rootfs, dest string, node *devices.Device) error {
f, err := os.Create(dest)
if err != nil && !os.IsExist(err) {
return err
}
if f != nil {
_ = f.Close()
}
return utils.WithProcfd(rootfs, dest, func(procfd string) error {
return unix.Mount(node.Path, procfd, "bind", unix.MS_BIND, "")
})
}
// Creates the device node in the rootfs of the container.
func createDeviceNode(rootfs string, node *devices.Device, bind bool) error {
if node.Path == "" {
// The node only exists for cgroup reasons, ignore it here.
return nil
}
dest, err := securejoin.SecureJoin(rootfs, node.Path)
if err != nil {
return err
}
if err := os.MkdirAll(filepath.Dir(dest), 0o755); err != nil {
return err
}
if bind {
return bindMountDeviceNode(rootfs, dest, node)
}
if err := mknodDevice(dest, node); err != nil {
if os.IsExist(err) {
return nil
} else if os.IsPermission(err) {
return bindMountDeviceNode(rootfs, dest, node)
}
return err
}
return nil
}
func mknodDevice(dest string, node *devices.Device) error {
fileMode := node.FileMode
switch node.Type {
case devices.BlockDevice:
fileMode |= unix.S_IFBLK
case devices.CharDevice:
fileMode |= unix.S_IFCHR
case devices.FifoDevice:
fileMode |= unix.S_IFIFO
default:
return fmt.Errorf("%c is not a valid device type for device %s", node.Type, node.Path)
}
dev, err := node.Mkdev()
if err != nil {
return err
}
if err := unix.Mknod(dest, uint32(fileMode), int(dev)); err != nil {
return err
}
return unix.Chown(dest, int(node.Uid), int(node.Gid))
}
// Get the parent mount point of directory passed in as argument. Also return
// optional fields.
func getParentMount(rootfs string) (string, string, error) {
mi, err := mountinfo.GetMounts(mountinfo.ParentsFilter(rootfs))
if err != nil {
return "", "", err
}
if len(mi) < 1 {
return "", "", fmt.Errorf("could not find parent mount of %s", rootfs)
}
// find the longest mount point
var idx, maxlen int
for i := range mi {
if len(mi[i].Mountpoint) > maxlen {
maxlen = len(mi[i].Mountpoint)
idx = i
}
}
return mi[idx].Mountpoint, mi[idx].Optional, nil
}
// Make parent mount private if it was shared
func rootfsParentMountPrivate(rootfs string) error {
sharedMount := false
parentMount, optionalOpts, err := getParentMount(rootfs)
if err != nil {
return err
}
optsSplit := strings.Split(optionalOpts, " ")
for _, opt := range optsSplit {
if strings.HasPrefix(opt, "shared:") {
sharedMount = true
break
}
}
// Make parent mount PRIVATE if it was shared. It is needed for two
// reasons. First of all pivot_root() will fail if parent mount is
// shared. Secondly when we bind mount rootfs it will propagate to
// parent namespace and we don't want that to happen.
if sharedMount {
return unix.Mount("", parentMount, "", unix.MS_PRIVATE, "")
}
return nil
}
func prepareRoot(config *configs.Config) error {
flag := unix.MS_SLAVE | unix.MS_REC
if config.RootPropagation != 0 {
flag = config.RootPropagation
}
if err := unix.Mount("", "/", "", uintptr(flag), ""); err != nil {
return err
}
// Make parent mount private to make sure following bind mount does
// not propagate in other namespaces. Also it will help with kernel
// check pass in pivot_root. (IS_SHARED(new_mnt->mnt_parent))
if err := rootfsParentMountPrivate(config.Rootfs); err != nil {
return err
}
return unix.Mount(config.Rootfs, config.Rootfs, "bind", unix.MS_BIND|unix.MS_REC, "")
}
func setReadonly() error {
flags := uintptr(unix.MS_BIND | unix.MS_REMOUNT | unix.MS_RDONLY)
err := unix.Mount("", "/", "", flags, "")
if err == nil {
return nil
}
var s unix.Statfs_t
if err := unix.Statfs("/", &s); err != nil {
return &os.PathError{Op: "statfs", Path: "/", Err: err}
}
flags |= uintptr(s.Flags)
return unix.Mount("", "/", "", flags, "")
}
func setupPtmx(config *configs.Config) error {
ptmx := filepath.Join(config.Rootfs, "dev/ptmx")
if err := os.Remove(ptmx); err != nil && !os.IsNotExist(err) {
return err
}
if err := os.Symlink("pts/ptmx", ptmx); err != nil {
return fmt.Errorf("symlink dev ptmx %s", err)
}
return nil
}
// pivotRoot will call pivot_root such that rootfs becomes the new root
// filesystem, and everything else is cleaned up.
func pivotRoot(rootfs string) error {
// While the documentation may claim otherwise, pivot_root(".", ".") is
// actually valid. What this results in is / being the new root but
// /proc/self/cwd being the old root. Since we can play around with the cwd
// with pivot_root this allows us to pivot without creating directories in
// the rootfs. Shout-outs to the LXC developers for giving us this idea.
oldroot, err := unix.Open("/", unix.O_DIRECTORY|unix.O_RDONLY, 0)
if err != nil {
return err
}
defer unix.Close(oldroot) //nolint: errcheck
newroot, err := unix.Open(rootfs, unix.O_DIRECTORY|unix.O_RDONLY, 0)
if err != nil {
return err
}
defer unix.Close(newroot) //nolint: errcheck
// Change to the new root so that the pivot_root actually acts on it.
if err := unix.Fchdir(newroot); err != nil {
return err
}
if err := unix.PivotRoot(".", "."); err != nil {
return fmt.Errorf("pivot_root %s", err)
}
// Currently our "." is oldroot (according to the current kernel code).
// However, purely for safety, we will fchdir(oldroot) since there isn't
// really any guarantee from the kernel what /proc/self/cwd will be after a
// pivot_root(2).
if err := unix.Fchdir(oldroot); err != nil {
return err
}
// Make oldroot rslave to make sure our unmounts don't propagate to the
// host (and thus bork the machine). We don't use rprivate because this is
// known to cause issues due to races where we still have a reference to a
// mount while a process in the host namespace are trying to operate on
// something they think has no mounts (devicemapper in particular).
if err := unix.Mount("", ".", "", unix.MS_SLAVE|unix.MS_REC, ""); err != nil {
return err
}
// Preform the unmount. MNT_DETACH allows us to unmount /proc/self/cwd.
if err := unix.Unmount(".", unix.MNT_DETACH); err != nil {
return err
}
// Switch back to our shiny new root.
if err := unix.Chdir("/"); err != nil {
return fmt.Errorf("chdir / %s", err)
}
return nil
}
func msMoveRoot(rootfs string) error {
// Before we move the root and chroot we have to mask all "full" sysfs and
// procfs mounts which exist on the host. This is because while the kernel
// has protections against mounting procfs if it has masks, when using
// chroot(2) the *host* procfs mount is still reachable in the mount
// namespace and the kernel permits procfs mounts inside --no-pivot
// containers.
//
// Users shouldn't be using --no-pivot except in exceptional circumstances,
// but to avoid such a trivial security flaw we apply a best-effort
// protection here. The kernel only allows a mount of a pseudo-filesystem
// like procfs or sysfs if there is a *full* mount (the root of the
// filesystem is mounted) without any other locked mount points covering a
// subtree of the mount.
//
// So we try to unmount (or mount tmpfs on top of) any mountpoint which is
// a full mount of either sysfs or procfs (since those are the most
// concerning filesystems to us).
mountinfos, err := mountinfo.GetMounts(func(info *mountinfo.Info) (skip, stop bool) {
// Collect every sysfs and procfs filesystem, except for those which
// are non-full mounts or are inside the rootfs of the container.
if info.Root != "/" ||
(info.FSType != "proc" && info.FSType != "sysfs") ||
strings.HasPrefix(info.Mountpoint, rootfs) {
skip = true
}
return
})
if err != nil {
return err
}
for _, info := range mountinfos {
p := info.Mountpoint
// Be sure umount events are not propagated to the host.
if err := unix.Mount("", p, "", unix.MS_SLAVE|unix.MS_REC, ""); err != nil {
if err == unix.ENOENT {
// If the mountpoint doesn't exist that means that we've
// already blasted away some parent directory of the mountpoint
// and so we don't care about this error.
continue
}
return err
}
if err := unix.Unmount(p, unix.MNT_DETACH); err != nil {
if err != unix.EINVAL && err != unix.EPERM {
return err
} else {
// If we have not privileges for umounting (e.g. rootless), then
// cover the path.
if err := unix.Mount("tmpfs", p, "tmpfs", 0, ""); err != nil {
return err
}
}
}
}
// Move the rootfs on top of "/" in our mount namespace.
if err := unix.Mount(rootfs, "/", "", unix.MS_MOVE, ""); err != nil {
return err
}
return chroot()
}
func chroot() error {
if err := unix.Chroot("."); err != nil {
return err
}
return unix.Chdir("/")
}
// createIfNotExists creates a file or a directory only if it does not already exist.
func createIfNotExists(path string, isDir bool) error {
if _, err := os.Stat(path); err != nil {
if os.IsNotExist(err) {
if isDir {
return os.MkdirAll(path, 0o755)
}
if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil {
return err
}
f, err := os.OpenFile(path, os.O_CREATE, 0o755)
if err != nil {
return err
}
_ = f.Close()
}
}
return nil
}
// readonlyPath will make a path read only.
func readonlyPath(path string) error {
if err := unix.Mount(path, path, "", unix.MS_BIND|unix.MS_REC, ""); err != nil {
if os.IsNotExist(err) {
return nil
}
return &os.PathError{Op: "bind-mount", Path: path, Err: err}
}
var s unix.Statfs_t
if err := unix.Statfs(path, &s); err != nil {
return &os.PathError{Op: "statfs", Path: path, Err: err}
}
flags := uintptr(s.Flags) & (unix.MS_NOSUID | unix.MS_NODEV | unix.MS_NOEXEC)
if err := unix.Mount(path, path, "", flags|unix.MS_BIND|unix.MS_REMOUNT|unix.MS_RDONLY, ""); err != nil {
return &os.PathError{Op: "bind-mount-ro", Path: path, Err: err}
}
return nil
}
// remountReadonly will remount an existing mount point and ensure that it is read-only.
func remountReadonly(m *configs.Mount) error {
var (
dest = m.Destination
flags = m.Flags
)
for i := 0; i < 5; i++ {
// There is a special case in the kernel for
// MS_REMOUNT | MS_BIND, which allows us to change only the
// flags even as an unprivileged user (i.e. user namespace)
// assuming we don't drop any security related flags (nodev,
// nosuid, etc.). So, let's use that case so that we can do
// this re-mount without failing in a userns.
flags |= unix.MS_REMOUNT | unix.MS_BIND | unix.MS_RDONLY
if err := unix.Mount("", dest, "", uintptr(flags), ""); err != nil {
switch err {
case unix.EBUSY:
time.Sleep(100 * time.Millisecond)
continue
default:
return err
}
}
return nil
}
return fmt.Errorf("unable to mount %s as readonly max retries reached", dest)
}
// maskPath masks the top of the specified path inside a container to avoid
// security issues from processes reading information from non-namespace aware
// mounts ( proc/kcore ).
// For files, maskPath bind mounts /dev/null over the top of the specified path.
// For directories, maskPath mounts read-only tmpfs over the top of the specified path.
func maskPath(path string, mountLabel string) error {
if err := unix.Mount("/dev/null", path, "", unix.MS_BIND, ""); err != nil && !os.IsNotExist(err) {
if err == unix.ENOTDIR {
return unix.Mount("tmpfs", path, "tmpfs", unix.MS_RDONLY, label.FormatMountLabel("", mountLabel))
}
return err
}
return nil
}
// writeSystemProperty writes the value to a path under /proc/sys as determined from the key.
// For e.g. net.ipv4.ip_forward translated to /proc/sys/net/ipv4/ip_forward.
func writeSystemProperty(key, value string) error {
keyPath := strings.Replace(key, ".", "/", -1)
return ioutil.WriteFile(path.Join("/proc/sys", keyPath), []byte(value), 0o644)
}
func remount(m *configs.Mount, rootfs string) error {
return utils.WithProcfd(rootfs, m.Destination, func(procfd string) error {
flags := uintptr(m.Flags | unix.MS_REMOUNT)
err := unix.Mount(m.Source, procfd, m.Device, flags, "")
if err == nil {
return nil
}
// Check if the source has ro flag...
var s unix.Statfs_t
if err := unix.Statfs(m.Source, &s); err != nil {
return &os.PathError{Op: "statfs", Path: m.Source, Err: err}
}
if s.Flags&unix.MS_RDONLY != unix.MS_RDONLY {
return err
}
// ... and retry the mount with ro flag set.
flags |= unix.MS_RDONLY
return unix.Mount(m.Source, procfd, m.Device, flags, "")
})
}
// Do the mount operation followed by additional mounts required to take care
// of propagation flags. This will always be scoped inside the container rootfs.
func mountPropagate(m *configs.Mount, rootfs string, mountLabel string) error {
var (
data = label.FormatMountLabel(m.Data, mountLabel)
flags = m.Flags
)
// Delay mounting the filesystem read-only if we need to do further
// operations on it. We need to set up files in "/dev", and other tmpfs
// mounts may need to be chmod-ed after mounting. These mounts will be
// remounted ro later in finalizeRootfs(), if necessary.
if m.Device == "tmpfs" || utils.CleanPath(m.Destination) == "/dev" {
flags &= ^unix.MS_RDONLY
}
// Because the destination is inside a container path which might be
// mutating underneath us, we verify that we are actually going to mount
// inside the container with WithProcfd() -- mounting through a procfd
// mounts on the target.
if err := utils.WithProcfd(rootfs, m.Destination, func(procfd string) error {
return unix.Mount(m.Source, procfd, m.Device, uintptr(flags), data)
}); err != nil {
return fmt.Errorf("mount through procfd: %w", err)
}
// We have to apply mount propagation flags in a separate WithProcfd() call
// because the previous call invalidates the passed procfd -- the mount
// target needs to be re-opened.
if err := utils.WithProcfd(rootfs, m.Destination, func(procfd string) error {
for _, pflag := range m.PropagationFlags {
if err := unix.Mount("", procfd, "", uintptr(pflag), ""); err != nil {
return err
}
}
return nil
}); err != nil {
return fmt.Errorf("change mount propagation through procfd: %w", err)
}
return nil
}
|