diff options
Diffstat (limited to 'pkg')
25 files changed, 143 insertions, 142 deletions
diff --git a/pkg/sentry/fs/host/tty.go b/pkg/sentry/fs/host/tty.go index b5229098c..e29ae00f2 100644 --- a/pkg/sentry/fs/host/tty.go +++ b/pkg/sentry/fs/host/tty.go @@ -358,7 +358,7 @@ func (t *TTYFileOperations) checkChange(ctx context.Context, sig linux.Signal) e // // Linux ignores the result of kill_pgrp(). _ = pg.SendSignal(kernel.SignalInfoPriv(sig)) - return kernel.ERESTARTSYS + return syserror.ERESTARTSYS } // LINT.ThenChange(../../fsimpl/host/tty.go) diff --git a/pkg/sentry/fsimpl/host/tty.go b/pkg/sentry/fsimpl/host/tty.go index d372c60cb..27cbd3059 100644 --- a/pkg/sentry/fsimpl/host/tty.go +++ b/pkg/sentry/fsimpl/host/tty.go @@ -376,7 +376,7 @@ func (t *TTYFileDescription) checkChange(ctx context.Context, sig linux.Signal) // // Linux ignores the result of kill_pgrp(). _ = pg.SendSignal(kernel.SignalInfoPriv(sig)) - return kernel.ERESTARTSYS + return syserror.ERESTARTSYS } // LockPOSIX implements vfs.FileDescriptionImpl.LockPOSIX. diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index abaf29216..aa3a573c0 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -26,6 +26,7 @@ import ( ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" "gvisor.dev/gvisor/pkg/sentry/memmap" "gvisor.dev/gvisor/pkg/sentry/platform" + "gvisor.dev/gvisor/pkg/syserror" "gvisor.dev/gvisor/pkg/usermem" ) @@ -189,8 +190,8 @@ func (app *runApp) execute(t *Task) taskRunState { // a pending signal, causing another interruption, but that signal should // not interact with the interrupted syscall.) if t.haveSyscallReturn { - if sre, ok := SyscallRestartErrnoFromReturn(t.Arch().Return()); ok { - if sre == ERESTART_RESTARTBLOCK { + if sre, ok := syserror.SyscallRestartErrnoFromReturn(t.Arch().Return()); ok { + if sre == syserror.ERESTART_RESTARTBLOCK { t.Debugf("Restarting syscall %d with restart block after errno %d: not interrupted by handled signal", t.Arch().SyscallNo(), sre) t.Arch().RestartSyscallWithRestartBlock() } else { diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index cff2a8365..d6a2040bc 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -159,7 +159,7 @@ func (t *Task) deliverSignal(info *arch.SignalInfo, act arch.SignalAct) taskRunS sigact := computeAction(linux.Signal(info.Signo), act) if t.haveSyscallReturn { - if sre, ok := SyscallRestartErrnoFromReturn(t.Arch().Return()); ok { + if sre, ok := syserror.SyscallRestartErrnoFromReturn(t.Arch().Return()); ok { // Signals that are ignored, cause a thread group stop, or // terminate the thread group do not interact with interrupted // syscalls; in Linux terms, they are never returned to the signal @@ -168,11 +168,11 @@ func (t *Task) deliverSignal(info *arch.SignalInfo, act arch.SignalAct) taskRunS // signal that is actually handled (by userspace). if sigact == SignalActionHandler { switch { - case sre == ERESTARTNOHAND: + case sre == syserror.ERESTARTNOHAND: fallthrough - case sre == ERESTART_RESTARTBLOCK: + case sre == syserror.ERESTART_RESTARTBLOCK: fallthrough - case (sre == ERESTARTSYS && !act.IsRestart()): + case (sre == syserror.ERESTARTSYS && !act.IsRestart()): t.Debugf("Not restarting syscall %d after errno %d: interrupted by signal %d", t.Arch().SyscallNo(), sre, info.Signo) t.Arch().SetReturn(uintptr(-ExtractErrno(syserror.EINTR, -1))) default: diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go index a5903b0b5..2dbf86547 100644 --- a/pkg/sentry/kernel/task_syscall.go +++ b/pkg/sentry/kernel/task_syscall.go @@ -29,75 +29,8 @@ import ( "gvisor.dev/gvisor/pkg/usermem" ) -// SyscallRestartErrno represents a ERESTART* errno defined in the Linux's kernel -// include/linux/errno.h. These errnos are never returned to userspace -// directly, but are used to communicate the expected behavior of an -// interrupted syscall from the syscall to signal handling. -type SyscallRestartErrno int - -// These numeric values are significant because ptrace syscall exit tracing can -// observe them. -// -// For all of the following errnos, if the syscall is not interrupted by a -// signal delivered to a user handler, the syscall is restarted. -const ( - // ERESTARTSYS is returned by an interrupted syscall to indicate that it - // should be converted to EINTR if interrupted by a signal delivered to a - // user handler without SA_RESTART set, and restarted otherwise. - ERESTARTSYS = SyscallRestartErrno(512) - - // ERESTARTNOINTR is returned by an interrupted syscall to indicate that it - // should always be restarted. - ERESTARTNOINTR = SyscallRestartErrno(513) - - // ERESTARTNOHAND is returned by an interrupted syscall to indicate that it - // should be converted to EINTR if interrupted by a signal delivered to a - // user handler, and restarted otherwise. - ERESTARTNOHAND = SyscallRestartErrno(514) - - // ERESTART_RESTARTBLOCK is returned by an interrupted syscall to indicate - // that it should be restarted using a custom function. The interrupted - // syscall must register a custom restart function by calling - // Task.SetRestartSyscallFn. - ERESTART_RESTARTBLOCK = SyscallRestartErrno(516) -) - var vsyscallCount = metric.MustCreateNewUint64Metric("/kernel/vsyscall_count", false /* sync */, "Number of times vsyscalls were invoked by the application") -// Error implements error.Error. -func (e SyscallRestartErrno) Error() string { - // Descriptions are borrowed from strace. - switch e { - case ERESTARTSYS: - return "to be restarted if SA_RESTART is set" - case ERESTARTNOINTR: - return "to be restarted" - case ERESTARTNOHAND: - return "to be restarted if no handler" - case ERESTART_RESTARTBLOCK: - return "interrupted by signal" - default: - return "(unknown interrupt error)" - } -} - -// SyscallRestartErrnoFromReturn returns the SyscallRestartErrno represented by -// rv, the value in a syscall return register. -func SyscallRestartErrnoFromReturn(rv uintptr) (SyscallRestartErrno, bool) { - switch int(rv) { - case -int(ERESTARTSYS): - return ERESTARTSYS, true - case -int(ERESTARTNOINTR): - return ERESTARTNOINTR, true - case -int(ERESTARTNOHAND): - return ERESTARTNOHAND, true - case -int(ERESTART_RESTARTBLOCK): - return ERESTART_RESTARTBLOCK, true - default: - return 0, false - } -} - // SyscallRestartBlock represents the restart block for a syscall restartable // with a custom function. It encapsulates the state required to restart a // syscall across a S/R. @@ -447,7 +380,7 @@ func ExtractErrno(err error, sysno int) int { return 0 case syscall.Errno: return int(err) - case SyscallRestartErrno: + case syserror.SyscallRestartErrno: return int(err) case *memmap.BusError: // Bus errors may generate SIGBUS, but for syscalls they still diff --git a/pkg/sentry/syscalls/linux/error.go b/pkg/sentry/syscalls/linux/error.go index 46060f6f5..dab6207c0 100644 --- a/pkg/sentry/syscalls/linux/error.go +++ b/pkg/sentry/syscalls/linux/error.go @@ -147,7 +147,7 @@ func handleIOErrorImpl(t *kernel.Task, partialResult bool, err, intr error, op s } switch err.(type) { - case kernel.SyscallRestartErrno: + case syserror.SyscallRestartErrno: // Identical to the EINTR case. return true, nil } diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 1bc9b184e..256422689 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -184,7 +184,7 @@ func openAt(t *kernel.Task, dirFD int32, addr usermem.Addr, flags uint) (fd uint file, err := d.Inode.GetFile(t, d, fileFlags) if err != nil { - return syserror.ConvertIntr(err, kernel.ERESTARTSYS) + return syserror.ConvertIntr(err, syserror.ERESTARTSYS) } defer file.DecRef(t) @@ -414,7 +414,7 @@ func createAt(t *kernel.Task, dirFD int32, addr usermem.Addr, flags uint, mode l // Create a new fs.File. newFile, err = found.Inode.GetFile(t, found, fileFlags) if err != nil { - return syserror.ConvertIntr(err, kernel.ERESTARTSYS) + return syserror.ConvertIntr(err, syserror.ERESTARTSYS) } defer newFile.DecRef(t) case syserror.ENOENT: diff --git a/pkg/sentry/syscalls/linux/sys_futex.go b/pkg/sentry/syscalls/linux/sys_futex.go index 9d1b2edb1..12b2fa690 100644 --- a/pkg/sentry/syscalls/linux/sys_futex.go +++ b/pkg/sentry/syscalls/linux/sys_futex.go @@ -74,7 +74,7 @@ func futexWaitAbsolute(t *kernel.Task, clockRealtime bool, ts linux.Timespec, fo } t.Futex().WaitComplete(w, t) - return 0, syserror.ConvertIntr(err, kernel.ERESTARTSYS) + return 0, syserror.ConvertIntr(err, syserror.ERESTARTSYS) } // futexWaitDuration performs a FUTEX_WAIT, blocking until the wait is @@ -110,7 +110,7 @@ func futexWaitDuration(t *kernel.Task, duration time.Duration, forever bool, add // The wait duration was absolute, restart with the original arguments. if forever { - return 0, kernel.ERESTARTSYS + return 0, syserror.ERESTARTSYS } // The wait duration was relative, restart with the remaining duration. @@ -121,7 +121,7 @@ func futexWaitDuration(t *kernel.Task, duration time.Duration, forever bool, add val: val, mask: mask, }) - return 0, kernel.ERESTART_RESTARTBLOCK + return 0, syserror.ERESTART_RESTARTBLOCK } func futexLockPI(t *kernel.Task, ts linux.Timespec, forever bool, addr usermem.Addr, private bool) error { @@ -149,7 +149,7 @@ func futexLockPI(t *kernel.Task, ts linux.Timespec, forever bool, addr usermem.A } t.Futex().WaitComplete(w, t) - return syserror.ConvertIntr(err, kernel.ERESTARTSYS) + return syserror.ConvertIntr(err, syserror.ERESTARTSYS) } func tryLockPI(t *kernel.Task, addr usermem.Addr, private bool) error { diff --git a/pkg/sentry/syscalls/linux/sys_getdents.go b/pkg/sentry/syscalls/linux/sys_getdents.go index f5699e55d..59004cefe 100644 --- a/pkg/sentry/syscalls/linux/sys_getdents.go +++ b/pkg/sentry/syscalls/linux/sys_getdents.go @@ -82,7 +82,7 @@ func getdents(t *kernel.Task, fd int32, addr usermem.Addr, size int, f func(*dir ds := newDirentSerializer(f, w, t.Arch(), size) rerr := dir.Readdir(t, ds) - switch err := handleIOError(t, ds.Written() > 0, rerr, kernel.ERESTARTSYS, "getdents", dir); err { + switch err := handleIOError(t, ds.Written() > 0, rerr, syserror.ERESTARTSYS, "getdents", dir); err { case nil: dir.Dirent.InotifyEvent(linux.IN_ACCESS, 0) return uintptr(ds.Written()), nil diff --git a/pkg/sentry/syscalls/linux/sys_lseek.go b/pkg/sentry/syscalls/linux/sys_lseek.go index 1c38f8f4f..0046347cb 100644 --- a/pkg/sentry/syscalls/linux/sys_lseek.go +++ b/pkg/sentry/syscalls/linux/sys_lseek.go @@ -48,7 +48,7 @@ func Lseek(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall } offset, serr := file.Seek(t, sw, offset) - err := handleIOError(t, false /* partialResult */, serr, kernel.ERESTARTSYS, "lseek", file) + err := handleIOError(t, false /* partialResult */, serr, syserror.ERESTARTSYS, "lseek", file) if err != nil { return 0, nil, err } diff --git a/pkg/sentry/syscalls/linux/sys_mmap.go b/pkg/sentry/syscalls/linux/sys_mmap.go index 72786b032..d0109baa4 100644 --- a/pkg/sentry/syscalls/linux/sys_mmap.go +++ b/pkg/sentry/syscalls/linux/sys_mmap.go @@ -267,7 +267,7 @@ func Msync(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall }) // MSync calls fsync, the same interrupt conversion rules apply, see // mm/msync.c, fsync POSIX.1-2008. - return 0, nil, syserror.ConvertIntr(err, kernel.ERESTARTSYS) + return 0, nil, syserror.ConvertIntr(err, syserror.ERESTARTSYS) } // Mlock implements linux syscall mlock(2). diff --git a/pkg/sentry/syscalls/linux/sys_poll.go b/pkg/sentry/syscalls/linux/sys_poll.go index 3435bdf77..789e2ed5b 100644 --- a/pkg/sentry/syscalls/linux/sys_poll.go +++ b/pkg/sentry/syscalls/linux/sys_poll.go @@ -410,7 +410,7 @@ func poll(t *kernel.Task, pfdAddr usermem.Addr, nfds uint, timeout time.Duration nfds: nfds, timeout: remainingTimeout, }) - return 0, kernel.ERESTART_RESTARTBLOCK + return 0, syserror.ERESTART_RESTARTBLOCK } return n, err } @@ -464,7 +464,7 @@ func Ppoll(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall // Note that this means that if err is nil but copyErr is not, copyErr is // ignored. This is consistent with Linux. if err == syserror.EINTR && copyErr == nil { - err = kernel.ERESTARTNOHAND + err = syserror.ERESTARTNOHAND } return n, nil, err } @@ -494,7 +494,7 @@ func Select(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal copyErr := copyOutTimevalRemaining(t, startNs, timeout, timevalAddr) // See comment in Ppoll. if err == syserror.EINTR && copyErr == nil { - err = kernel.ERESTARTNOHAND + err = syserror.ERESTARTNOHAND } return n, nil, err } @@ -539,7 +539,7 @@ func Pselect(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca copyErr := copyOutTimespecRemaining(t, startNs, timeout, timespecAddr) // See comment in Ppoll. if err == syserror.EINTR && copyErr == nil { - err = kernel.ERESTARTNOHAND + err = syserror.ERESTARTNOHAND } return n, nil, err } diff --git a/pkg/sentry/syscalls/linux/sys_read.go b/pkg/sentry/syscalls/linux/sys_read.go index 3bbc3fa4b..f655d3db1 100644 --- a/pkg/sentry/syscalls/linux/sys_read.go +++ b/pkg/sentry/syscalls/linux/sys_read.go @@ -71,7 +71,7 @@ func Read(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallC n, err := readv(t, file, dst) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "read", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "read", file) } // Readahead implements readahead(2). @@ -151,7 +151,7 @@ func Pread64(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca n, err := preadv(t, file, dst, offset) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "pread64", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "pread64", file) } // Readv implements linux syscall readv(2). @@ -181,7 +181,7 @@ func Readv(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall n, err := readv(t, file, dst) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "readv", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "readv", file) } // Preadv implements linux syscall preadv(2). @@ -222,7 +222,7 @@ func Preadv(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal n, err := preadv(t, file, dst, offset) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "preadv", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "preadv", file) } // Preadv2 implements linux syscall preadv2(2). @@ -280,12 +280,12 @@ func Preadv2(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca if offset == -1 { n, err := readv(t, file, dst) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "preadv2", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "preadv2", file) } n, err := preadv(t, file, dst, offset) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "preadv2", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "preadv2", file) } func readv(t *kernel.Task, f *fs.File, dst usermem.IOSequence) (int64, error) { diff --git a/pkg/sentry/syscalls/linux/sys_signal.go b/pkg/sentry/syscalls/linux/sys_signal.go index 20cb1a5cb..e748d33d8 100644 --- a/pkg/sentry/syscalls/linux/sys_signal.go +++ b/pkg/sentry/syscalls/linux/sys_signal.go @@ -348,7 +348,7 @@ func Sigaltstack(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.S // Pause implements linux syscall pause(2). func Pause(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { - return 0, nil, syserror.ConvertIntr(t.Block(nil), kernel.ERESTARTNOHAND) + return 0, nil, syserror.ConvertIntr(t.Block(nil), syserror.ERESTARTNOHAND) } // RtSigpending implements linux syscall rt_sigpending(2). @@ -496,7 +496,7 @@ func RtSigsuspend(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel. t.SetSavedSignalMask(oldmask) // Perform the wait. - return 0, nil, syserror.ConvertIntr(t.Block(nil), kernel.ERESTARTNOHAND) + return 0, nil, syserror.ConvertIntr(t.Block(nil), syserror.ERESTARTNOHAND) } // RestartSyscall implements the linux syscall restart_syscall(2). diff --git a/pkg/sentry/syscalls/linux/sys_socket.go b/pkg/sentry/syscalls/linux/sys_socket.go index fec1c1974..38f573c14 100644 --- a/pkg/sentry/syscalls/linux/sys_socket.go +++ b/pkg/sentry/syscalls/linux/sys_socket.go @@ -285,7 +285,7 @@ func Connect(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca } blocking := !file.Flags().NonBlocking - return 0, nil, syserror.ConvertIntr(s.Connect(t, a, blocking).ToError(), kernel.ERESTARTSYS) + return 0, nil, syserror.ConvertIntr(s.Connect(t, a, blocking).ToError(), syserror.ERESTARTSYS) } // accept is the implementation of the accept syscall. It is called by accept @@ -316,7 +316,7 @@ func accept(t *kernel.Task, fd int32, addr usermem.Addr, addrLen usermem.Addr, f peerRequested := addrLen != 0 nfd, peer, peerLen, e := s.Accept(t, peerRequested, flags, blocking) if e != nil { - return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS) + return 0, syserror.ConvertIntr(e.ToError(), syserror.ERESTARTSYS) } if peerRequested { // NOTE(magi): Linux does not give you an error if it can't @@ -771,7 +771,7 @@ func recvSingleMsg(t *kernel.Task, s socket.Socket, msgPtr usermem.Addr, flags i if msg.ControlLen == 0 && msg.NameLen == 0 { n, mflags, _, _, cms, err := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, false, 0) if err != nil { - return 0, syserror.ConvertIntr(err.ToError(), kernel.ERESTARTSYS) + return 0, syserror.ConvertIntr(err.ToError(), syserror.ERESTARTSYS) } if !cms.Unix.Empty() { mflags |= linux.MSG_CTRUNC @@ -793,7 +793,7 @@ func recvSingleMsg(t *kernel.Task, s socket.Socket, msgPtr usermem.Addr, flags i } n, mflags, sender, senderLen, cms, e := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, msg.NameLen != 0, msg.ControlLen) if e != nil { - return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS) + return 0, syserror.ConvertIntr(e.ToError(), syserror.ERESTARTSYS) } defer cms.Release(t) @@ -882,7 +882,7 @@ func recvFrom(t *kernel.Task, fd int32, bufPtr usermem.Addr, bufLen uint64, flag n, _, sender, senderLen, cm, e := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, nameLenPtr != 0, 0) cm.Release(t) if e != nil { - return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS) + return 0, syserror.ConvertIntr(e.ToError(), syserror.ERESTARTSYS) } // Copy the address to the caller. @@ -1064,7 +1064,7 @@ func sendSingleMsg(t *kernel.Task, s socket.Socket, file *fs.File, msgPtr userme // Call the syscall implementation. n, e := s.SendMsg(t, src, to, int(flags), haveDeadline, deadline, controlMessages) - err = handleIOError(t, n != 0, e.ToError(), kernel.ERESTARTSYS, "sendmsg", file) + err = handleIOError(t, n != 0, e.ToError(), syserror.ERESTARTSYS, "sendmsg", file) if err != nil { controlMessages.Release(t) } @@ -1124,7 +1124,7 @@ func sendTo(t *kernel.Task, fd int32, bufPtr usermem.Addr, bufLen uint64, flags // Call the syscall implementation. n, e := s.SendMsg(t, src, to, int(flags), haveDeadline, deadline, socket.ControlMessages{Unix: control.New(t, s, nil)}) - return uintptr(n), handleIOError(t, n != 0, e.ToError(), kernel.ERESTARTSYS, "sendto", file) + return uintptr(n), handleIOError(t, n != 0, e.ToError(), syserror.ERESTARTSYS, "sendto", file) } // SendTo implements the linux syscall sendto(2). diff --git a/pkg/sentry/syscalls/linux/sys_splice.go b/pkg/sentry/syscalls/linux/sys_splice.go index b8846a10a..c69941feb 100644 --- a/pkg/sentry/syscalls/linux/sys_splice.go +++ b/pkg/sentry/syscalls/linux/sys_splice.go @@ -170,7 +170,7 @@ func Sendfile(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysc // We can only pass a single file to handleIOError, so pick inFile // arbitrarily. This is used only for debugging purposes. - return uintptr(n), nil, handleIOError(t, false, err, kernel.ERESTARTSYS, "sendfile", inFile) + return uintptr(n), nil, handleIOError(t, false, err, syserror.ERESTARTSYS, "sendfile", inFile) } // Splice implements splice(2). @@ -280,7 +280,7 @@ func Splice(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal } // See above; inFile is chosen arbitrarily here. - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "splice", inFile) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "splice", inFile) } // Tee imlements tee(2). @@ -333,5 +333,5 @@ func Tee(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallCo } // See above; inFile is chosen arbitrarily here. - return uintptr(n), nil, handleIOError(t, false, err, kernel.ERESTARTSYS, "tee", inFile) + return uintptr(n), nil, handleIOError(t, false, err, syserror.ERESTARTSYS, "tee", inFile) } diff --git a/pkg/sentry/syscalls/linux/sys_sync.go b/pkg/sentry/syscalls/linux/sys_sync.go index f2c0e5069..048a21c6e 100644 --- a/pkg/sentry/syscalls/linux/sys_sync.go +++ b/pkg/sentry/syscalls/linux/sys_sync.go @@ -57,7 +57,7 @@ func Fsync(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall defer file.DecRef(t) err := file.Fsync(t, 0, fs.FileMaxOffset, fs.SyncAll) - return 0, nil, syserror.ConvertIntr(err, kernel.ERESTARTSYS) + return 0, nil, syserror.ConvertIntr(err, syserror.ERESTARTSYS) } // Fdatasync implements linux syscall fdatasync(2). @@ -73,7 +73,7 @@ func Fdatasync(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys defer file.DecRef(t) err := file.Fsync(t, 0, fs.FileMaxOffset, fs.SyncData) - return 0, nil, syserror.ConvertIntr(err, kernel.ERESTARTSYS) + return 0, nil, syserror.ConvertIntr(err, syserror.ERESTARTSYS) } // SyncFileRange implements linux syscall sync_file_rage(2) @@ -135,7 +135,7 @@ func SyncFileRange(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel err = file.Fsync(t, offset, fs.FileMaxOffset, fs.SyncData) } - return 0, nil, syserror.ConvertIntr(err, kernel.ERESTARTSYS) + return 0, nil, syserror.ConvertIntr(err, syserror.ERESTARTSYS) } // LINT.ThenChange(vfs2/sync.go) diff --git a/pkg/sentry/syscalls/linux/sys_thread.go b/pkg/sentry/syscalls/linux/sys_thread.go index 2d16e4933..101096038 100644 --- a/pkg/sentry/syscalls/linux/sys_thread.go +++ b/pkg/sentry/syscalls/linux/sys_thread.go @@ -262,7 +262,7 @@ func parseCommonWaitOptions(wopts *kernel.WaitOptions, options int) error { wopts.Events |= kernel.EventGroupContinue } if options&linux.WNOHANG == 0 { - wopts.BlockInterruptErr = kernel.ERESTARTSYS + wopts.BlockInterruptErr = syserror.ERESTARTSYS } if options&linux.WNOTHREAD == 0 { wopts.SiblingChildren = true diff --git a/pkg/sentry/syscalls/linux/sys_time.go b/pkg/sentry/syscalls/linux/sys_time.go index 2d2aa0819..a2a24a027 100644 --- a/pkg/sentry/syscalls/linux/sys_time.go +++ b/pkg/sentry/syscalls/linux/sys_time.go @@ -213,7 +213,7 @@ func clockNanosleepUntil(t *kernel.Task, c ktime.Clock, ts linux.Timespec) error return nil } - return syserror.ConvertIntr(err, kernel.ERESTARTNOHAND) + return syserror.ConvertIntr(err, syserror.ERESTARTNOHAND) } // clockNanosleepFor blocks for a specified duration. @@ -254,7 +254,7 @@ func clockNanosleepFor(t *kernel.Task, c ktime.Clock, dur time.Duration, rem use duration: remaining, rem: rem, }) - return kernel.ERESTART_RESTARTBLOCK + return syserror.ERESTART_RESTARTBLOCK default: panic(fmt.Sprintf("Impossible BlockWithTimer error %v", err)) } diff --git a/pkg/sentry/syscalls/linux/sys_write.go b/pkg/sentry/syscalls/linux/sys_write.go index 485526e28..95bfe6606 100644 --- a/pkg/sentry/syscalls/linux/sys_write.go +++ b/pkg/sentry/syscalls/linux/sys_write.go @@ -71,7 +71,7 @@ func Write(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall n, err := writev(t, file, src) t.IOUsage().AccountWriteSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "write", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "write", file) } // Pwrite64 implements linux syscall pwrite64(2). @@ -118,7 +118,7 @@ func Pwrite64(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysc n, err := pwritev(t, file, src, offset) t.IOUsage().AccountWriteSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "pwrite64", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "pwrite64", file) } // Writev implements linux syscall writev(2). @@ -148,7 +148,7 @@ func Writev(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal n, err := writev(t, file, src) t.IOUsage().AccountWriteSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "writev", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "writev", file) } // Pwritev implements linux syscall pwritev(2). @@ -189,7 +189,7 @@ func Pwritev(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca n, err := pwritev(t, file, src, offset) t.IOUsage().AccountWriteSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "pwritev", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "pwritev", file) } // Pwritev2 implements linux syscall pwritev2(2). @@ -250,12 +250,12 @@ func Pwritev2(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysc if offset == -1 { n, err := writev(t, file, src) t.IOUsage().AccountWriteSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "pwritev2", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "pwritev2", file) } n, err := pwritev(t, file, src, offset) t.IOUsage().AccountWriteSyscall(n) - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "pwritev2", file) + return uintptr(n), nil, handleIOError(t, n != 0, err, syserror.ERESTARTSYS, "pwritev2", file) } func writev(t *kernel.Task, f *fs.File, src usermem.IOSequence) (int64, error) { diff --git a/pkg/sentry/syscalls/linux/vfs2/poll.go b/pkg/sentry/syscalls/linux/vfs2/poll.go index 7b9d5e18a..79ad64039 100644 --- a/pkg/sentry/syscalls/linux/vfs2/poll.go +++ b/pkg/sentry/syscalls/linux/vfs2/poll.go @@ -415,7 +415,7 @@ func poll(t *kernel.Task, pfdAddr usermem.Addr, nfds uint, timeout time.Duration nfds: nfds, timeout: remainingTimeout, }) - return 0, kernel.ERESTART_RESTARTBLOCK + return 0, syserror.ERESTART_RESTARTBLOCK } return n, err } @@ -462,7 +462,7 @@ func Ppoll(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall // Note that this means that if err is nil but copyErr is not, copyErr is // ignored. This is consistent with Linux. if err == syserror.EINTR && copyErr == nil { - err = kernel.ERESTARTNOHAND + err = syserror.ERESTARTNOHAND } return n, nil, err } @@ -492,7 +492,7 @@ func Select(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal copyErr := copyOutTimevalRemaining(t, startNs, timeout, timevalAddr) // See comment in Ppoll. if err == syserror.EINTR && copyErr == nil { - err = kernel.ERESTARTNOHAND + err = syserror.ERESTARTNOHAND } return n, nil, err } @@ -533,7 +533,7 @@ func Pselect(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca copyErr := copyOutTimespecRemaining(t, startNs, timeout, timespecAddr) // See comment in Ppoll. if err == syserror.EINTR && copyErr == nil { - err = kernel.ERESTARTNOHAND + err = syserror.ERESTARTNOHAND } return n, nil, err } diff --git a/pkg/sentry/syscalls/linux/vfs2/read_write.go b/pkg/sentry/syscalls/linux/vfs2/read_write.go index a905dae0a..b77b29dcc 100644 --- a/pkg/sentry/syscalls/linux/vfs2/read_write.go +++ b/pkg/sentry/syscalls/linux/vfs2/read_write.go @@ -62,7 +62,7 @@ func Read(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallC n, err := read(t, file, dst, vfs.ReadOptions{}) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, kernel.ERESTARTSYS, "read", file) + return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, syserror.ERESTARTSYS, "read", file) } // Readv implements Linux syscall readv(2). @@ -87,7 +87,7 @@ func Readv(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall n, err := read(t, file, dst, vfs.ReadOptions{}) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, kernel.ERESTARTSYS, "readv", file) + return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, syserror.ERESTARTSYS, "readv", file) } func read(t *kernel.Task, file *vfs.FileDescription, dst usermem.IOSequence, opts vfs.ReadOptions) (int64, error) { @@ -174,7 +174,7 @@ func Pread64(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca n, err := pread(t, file, dst, offset, vfs.ReadOptions{}) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, kernel.ERESTARTSYS, "pread64", file) + return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, syserror.ERESTARTSYS, "pread64", file) } // Preadv implements Linux syscall preadv(2). @@ -205,7 +205,7 @@ func Preadv(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal n, err := pread(t, file, dst, offset, vfs.ReadOptions{}) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, kernel.ERESTARTSYS, "preadv", file) + return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, syserror.ERESTARTSYS, "preadv", file) } // Preadv2 implements Linux syscall preadv2(2). @@ -251,7 +251,7 @@ func Preadv2(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca n, err = pread(t, file, dst, offset, opts) } t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, kernel.ERESTARTSYS, "preadv2", file) + return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, syserror.ERESTARTSYS, "preadv2", file) } func pread(t *kernel.Task, file *vfs.FileDescription, dst usermem.IOSequence, offset int64, opts vfs.ReadOptions) (int64, error) { @@ -332,7 +332,7 @@ func Write(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall n, err := write(t, file, src, vfs.WriteOptions{}) t.IOUsage().AccountWriteSyscall(n) - return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, kernel.ERESTARTSYS, "write", file) + return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, syserror.ERESTARTSYS, "write", file) } // Writev implements Linux syscall writev(2). @@ -357,7 +357,7 @@ func Writev(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal n, err := write(t, file, src, vfs.WriteOptions{}) t.IOUsage().AccountWriteSyscall(n) - return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, kernel.ERESTARTSYS, "writev", file) + return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, syserror.ERESTARTSYS, "writev", file) } func write(t *kernel.Task, file *vfs.FileDescription, src usermem.IOSequence, opts vfs.WriteOptions) (int64, error) { @@ -444,7 +444,7 @@ func Pwrite64(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysc n, err := pwrite(t, file, src, offset, vfs.WriteOptions{}) t.IOUsage().AccountWriteSyscall(n) - return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, kernel.ERESTARTSYS, "pwrite64", file) + return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, syserror.ERESTARTSYS, "pwrite64", file) } // Pwritev implements Linux syscall pwritev(2). @@ -475,7 +475,7 @@ func Pwritev(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca n, err := pwrite(t, file, src, offset, vfs.WriteOptions{}) t.IOUsage().AccountReadSyscall(n) - return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, kernel.ERESTARTSYS, "pwritev", file) + return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, syserror.ERESTARTSYS, "pwritev", file) } // Pwritev2 implements Linux syscall pwritev2(2). @@ -521,7 +521,7 @@ func Pwritev2(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysc n, err = pwrite(t, file, src, offset, opts) } t.IOUsage().AccountWriteSyscall(n) - return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, kernel.ERESTARTSYS, "pwritev2", file) + return uintptr(n), nil, slinux.HandleIOErrorVFS2(t, n != 0, err, syserror.ERESTARTSYS, "pwritev2", file) } func pwrite(t *kernel.Task, file *vfs.FileDescription, src usermem.IOSequence, offset int64, opts vfs.WriteOptions) (int64, error) { diff --git a/pkg/sentry/syscalls/linux/vfs2/socket.go b/pkg/sentry/syscalls/linux/vfs2/socket.go index 4a68c64f3..a5032657a 100644 --- a/pkg/sentry/syscalls/linux/vfs2/socket.go +++ b/pkg/sentry/syscalls/linux/vfs2/socket.go @@ -288,7 +288,7 @@ func Connect(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca } blocking := (file.StatusFlags() & linux.SOCK_NONBLOCK) == 0 - return 0, nil, syserror.ConvertIntr(s.Connect(t, a, blocking).ToError(), kernel.ERESTARTSYS) + return 0, nil, syserror.ConvertIntr(s.Connect(t, a, blocking).ToError(), syserror.ERESTARTSYS) } // accept is the implementation of the accept syscall. It is called by accept @@ -319,7 +319,7 @@ func accept(t *kernel.Task, fd int32, addr usermem.Addr, addrLen usermem.Addr, f peerRequested := addrLen != 0 nfd, peer, peerLen, e := s.Accept(t, peerRequested, flags, blocking) if e != nil { - return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS) + return 0, syserror.ConvertIntr(e.ToError(), syserror.ERESTARTSYS) } if peerRequested { // NOTE(magi): Linux does not give you an error if it can't @@ -774,7 +774,7 @@ func recvSingleMsg(t *kernel.Task, s socket.SocketVFS2, msgPtr usermem.Addr, fla if msg.ControlLen == 0 && msg.NameLen == 0 { n, mflags, _, _, cms, err := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, false, 0) if err != nil { - return 0, syserror.ConvertIntr(err.ToError(), kernel.ERESTARTSYS) + return 0, syserror.ConvertIntr(err.ToError(), syserror.ERESTARTSYS) } if !cms.Unix.Empty() { mflags |= linux.MSG_CTRUNC @@ -796,7 +796,7 @@ func recvSingleMsg(t *kernel.Task, s socket.SocketVFS2, msgPtr usermem.Addr, fla } n, mflags, sender, senderLen, cms, e := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, msg.NameLen != 0, msg.ControlLen) if e != nil { - return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS) + return 0, syserror.ConvertIntr(e.ToError(), syserror.ERESTARTSYS) } defer cms.Release(t) @@ -885,7 +885,7 @@ func recvFrom(t *kernel.Task, fd int32, bufPtr usermem.Addr, bufLen uint64, flag n, _, sender, senderLen, cm, e := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, nameLenPtr != 0, 0) cm.Release(t) if e != nil { - return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS) + return 0, syserror.ConvertIntr(e.ToError(), syserror.ERESTARTSYS) } // Copy the address to the caller. @@ -1067,7 +1067,7 @@ func sendSingleMsg(t *kernel.Task, s socket.SocketVFS2, file *vfs.FileDescriptio // Call the syscall implementation. n, e := s.SendMsg(t, src, to, int(flags), haveDeadline, deadline, controlMessages) - err = slinux.HandleIOErrorVFS2(t, n != 0, e.ToError(), kernel.ERESTARTSYS, "sendmsg", file) + err = slinux.HandleIOErrorVFS2(t, n != 0, e.ToError(), syserror.ERESTARTSYS, "sendmsg", file) if err != nil { controlMessages.Release(t) } @@ -1127,7 +1127,7 @@ func sendTo(t *kernel.Task, fd int32, bufPtr usermem.Addr, bufLen uint64, flags // Call the syscall implementation. n, e := s.SendMsg(t, src, to, int(flags), haveDeadline, deadline, socket.ControlMessages{Unix: control.New(t, s, nil)}) - return uintptr(n), slinux.HandleIOErrorVFS2(t, n != 0, e.ToError(), kernel.ERESTARTSYS, "sendto", file) + return uintptr(n), slinux.HandleIOErrorVFS2(t, n != 0, e.ToError(), syserror.ERESTARTSYS, "sendto", file) } // SendTo implements the linux syscall sendto(2). diff --git a/pkg/sentry/syscalls/linux/vfs2/sync.go b/pkg/sentry/syscalls/linux/vfs2/sync.go index a6491ac37..6e9b599e2 100644 --- a/pkg/sentry/syscalls/linux/vfs2/sync.go +++ b/pkg/sentry/syscalls/linux/vfs2/sync.go @@ -108,7 +108,7 @@ func SyncFileRange(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel if flags&linux.SYNC_FILE_RANGE_WAIT_AFTER != 0 { if err := file.Sync(t); err != nil { - return 0, nil, syserror.ConvertIntr(err, kernel.ERESTARTSYS) + return 0, nil, syserror.ConvertIntr(err, syserror.ERESTARTSYS) } } return 0, nil, nil diff --git a/pkg/syserror/syserror.go b/pkg/syserror/syserror.go index 798e07b01..fe9f50169 100644 --- a/pkg/syserror/syserror.go +++ b/pkg/syserror/syserror.go @@ -153,6 +153,73 @@ func ConvertIntr(err, intr error) error { return err } +// SyscallRestartErrno represents a ERESTART* errno defined in the Linux's kernel +// include/linux/errno.h. These errnos are never returned to userspace +// directly, but are used to communicate the expected behavior of an +// interrupted syscall from the syscall to signal handling. +type SyscallRestartErrno int + +// These numeric values are significant because ptrace syscall exit tracing can +// observe them. +// +// For all of the following errnos, if the syscall is not interrupted by a +// signal delivered to a user handler, the syscall is restarted. +const ( + // ERESTARTSYS is returned by an interrupted syscall to indicate that it + // should be converted to EINTR if interrupted by a signal delivered to a + // user handler without SA_RESTART set, and restarted otherwise. + ERESTARTSYS = SyscallRestartErrno(512) + + // ERESTARTNOINTR is returned by an interrupted syscall to indicate that it + // should always be restarted. + ERESTARTNOINTR = SyscallRestartErrno(513) + + // ERESTARTNOHAND is returned by an interrupted syscall to indicate that it + // should be converted to EINTR if interrupted by a signal delivered to a + // user handler, and restarted otherwise. + ERESTARTNOHAND = SyscallRestartErrno(514) + + // ERESTART_RESTARTBLOCK is returned by an interrupted syscall to indicate + // that it should be restarted using a custom function. The interrupted + // syscall must register a custom restart function by calling + // Task.SetRestartSyscallFn. + ERESTART_RESTARTBLOCK = SyscallRestartErrno(516) +) + +// Error implements error.Error. +func (e SyscallRestartErrno) Error() string { + // Descriptions are borrowed from strace. + switch e { + case ERESTARTSYS: + return "to be restarted if SA_RESTART is set" + case ERESTARTNOINTR: + return "to be restarted" + case ERESTARTNOHAND: + return "to be restarted if no handler" + case ERESTART_RESTARTBLOCK: + return "interrupted by signal" + default: + return "(unknown interrupt error)" + } +} + +// SyscallRestartErrnoFromReturn returns the SyscallRestartErrno represented by +// rv, the value in a syscall return register. +func SyscallRestartErrnoFromReturn(rv uintptr) (SyscallRestartErrno, bool) { + switch int(rv) { + case -int(ERESTARTSYS): + return ERESTARTSYS, true + case -int(ERESTARTNOINTR): + return ERESTARTNOINTR, true + case -int(ERESTARTNOHAND): + return ERESTARTNOHAND, true + case -int(ERESTART_RESTARTBLOCK): + return ERESTART_RESTARTBLOCK, true + default: + return 0, false + } +} + func init() { AddErrorTranslation(ErrWouldBlock, syscall.EWOULDBLOCK) AddErrorTranslation(ErrInterrupted, syscall.EINTR) |