| 1 | //go:build linux |
| 2 | |
| 3 | package instanceexec |
| 4 | |
| 5 | import ( |
| 6 | "encoding/json" |
| 7 | "errors" |
| 8 | "io" |
| 9 | "net" |
| 10 | "os" |
| 11 | "strings" |
| 12 | "time" |
| 13 | |
| 14 | "github.com/carlosframework/platform/internal/instancectl" |
| 15 | "golang.org/x/sys/unix" |
| 16 | ) |
| 17 | |
| 18 | func scopedClaimInit(argv0 string, argv, env []string, snapshot string) error { |
| 19 | return runInitChecked(argv0, argv, env, func() error { |
| 20 | if err := ScopeAbstractUnix(); err != nil { |
| 21 | return err |
| 22 | } |
| 23 | return ScopeClaimGeneration() |
| 24 | }, func() error { |
| 25 | return checkClaimExec(snapshot, instancectl.ClaimExecSocket, 0, claimBootTime) |
| 26 | }) |
| 27 | } |
| 28 | func claimBootTime() (string, int64, error) { |
| 29 | b, err := os.ReadFile(instancectl.ClaimExecBootID) |
| 30 | if err != nil { |
| 31 | return "", 0, err |
| 32 | } |
| 33 | var ts unix.Timespec |
| 34 | if err = unix.ClockGettime(unix.CLOCK_BOOTTIME, &ts); err != nil { |
| 35 | return "", 0, err |
| 36 | } |
| 37 | return strings.TrimSpace(string(b)), ts.Nano(), nil |
| 38 | } |
| 39 | |
| 40 | // The path is a root-controlled ExecStart argument, never tenant environment |
| 41 | // or a field of launch.run.json. Its read-only bind cannot be opted out of. |
| 42 | func checkClaimExec(snapshot, socket string, owner uint32, now func() (string, int64, error)) error { |
| 43 | fd, err := unix.Open(snapshot, unix.O_RDONLY|unix.O_CLOEXEC|unix.O_NOFOLLOW, 0) |
| 44 | if err != nil { |
| 45 | return err |
| 46 | } |
| 47 | f := os.NewFile(uintptr(fd), snapshot) |
| 48 | defer f.Close() |
| 49 | var st unix.Stat_t |
| 50 | if err = unix.Fstat(fd, &st); err != nil { |
| 51 | return err |
| 52 | } |
| 53 | if st.Uid != owner || st.Mode&unix.S_IFMT != unix.S_IFREG || st.Mode&0222 != 0 { |
| 54 | return instancectl.ErrClaimUnit |
| 55 | } |
| 56 | var want instancectl.ClaimExecPermit |
| 57 | if err = json.NewDecoder(io.LimitReader(f, 4096)).Decode(&want); err != nil { |
| 58 | return err |
| 59 | } |
| 60 | if !want.Valid() || want.Containment != 1 { |
| 61 | return instancectl.ErrClaimUnit |
| 62 | } |
| 63 | conn, err := net.DialTimeout("unix", socket, 5*time.Second) |
| 64 | if err != nil { |
| 65 | return err |
| 66 | } |
| 67 | defer conn.Close() |
| 68 | if err = conn.SetDeadline(time.Now().Add(5 * time.Second)); err != nil { |
| 69 | return err |
| 70 | } |
| 71 | raw, err := conn.(*net.UnixConn).SyscallConn() |
| 72 | if err != nil { |
| 73 | return err |
| 74 | } |
| 75 | var cred *unix.Ucred |
| 76 | var peerErr error |
| 77 | if err = raw.Control(func(fd uintptr) { cred, peerErr = unix.GetsockoptUcred(int(fd), unix.SOL_SOCKET, unix.SO_PEERCRED) }); err != nil { |
| 78 | return err |
| 79 | } |
| 80 | if peerErr != nil { |
| 81 | return peerErr |
| 82 | } |
| 83 | if cred.Uid != owner { |
| 84 | return instancectl.ErrClaimUnit |
| 85 | } |
| 86 | if err = json.NewEncoder(conn).Encode(want); err != nil { |
| 87 | return err |
| 88 | } |
| 89 | var got instancectl.ClaimExecPermit |
| 90 | if err = json.NewDecoder(io.LimitReader(conn, 4096)).Decode(&got); err != nil { |
| 91 | return err |
| 92 | } |
| 93 | if got != want { |
| 94 | return instancectl.ErrClaimUnit |
| 95 | } |
| 96 | // Explicit close precedes the final time sample. No IO or namespace setup |
| 97 | // follows this sample before runInitChecked calls ForkExec. |
| 98 | if err = conn.Close(); err != nil && !errors.Is(err, net.ErrClosed) { |
| 99 | return err |
| 100 | } |
| 101 | boot, n, err := now() |
| 102 | if err != nil { |
| 103 | return err |
| 104 | } |
| 105 | if boot != want.BootID || n >= want.SafeUntil { |
| 106 | return instancectl.ErrClaimUnit |
| 107 | } |
| 108 | return nil |
| 109 | } |
| 110 | |