...
Run Format

Source file src/os/file_unix.go

Documentation: os

  // Copyright 2009 The Go Authors. All rights reserved.
  // Use of this source code is governed by a BSD-style
  // license that can be found in the LICENSE file.
  
  // +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
  
  package os
  
  import (
  	"internal/poll"
  	"runtime"
  	"syscall"
  )
  
  // fixLongPath is a noop on non-Windows platforms.
  func fixLongPath(path string) string {
  	return path
  }
  
  func rename(oldname, newname string) error {
  	fi, err := Lstat(newname)
  	if err == nil && fi.IsDir() {
  		// There are two independent errors this function can return:
  		// one for a bad oldname, and one for a bad newname.
  		// At this point we've determined the newname is bad.
  		// But just in case oldname is also bad, prioritize returning
  		// the oldname error because that's what we did historically.
  		if _, err := Lstat(oldname); err != nil {
  			if pe, ok := err.(*PathError); ok {
  				err = pe.Err
  			}
  			return &LinkError{"rename", oldname, newname, err}
  		}
  		return &LinkError{"rename", oldname, newname, syscall.EEXIST}
  	}
  	err = syscall.Rename(oldname, newname)
  	if err != nil {
  		return &LinkError{"rename", oldname, newname, err}
  	}
  	return nil
  }
  
  // file is the real representation of *File.
  // The extra level of indirection ensures that no clients of os
  // can overwrite this data, which could cause the finalizer
  // to close the wrong file descriptor.
  type file struct {
  	pfd      poll.FD
  	name     string
  	dirinfo  *dirInfo // nil unless directory being read
  	nonblock bool     // whether we set nonblocking mode
  }
  
  // Fd returns the integer Unix file descriptor referencing the open file.
  // The file descriptor is valid only until f.Close is called or f is garbage collected.
  func (f *File) Fd() uintptr {
  	if f == nil {
  		return ^(uintptr(0))
  	}
  
  	// If we put the file descriptor into nonblocking mode,
  	// then set it to blocking mode before we return it,
  	// because historically we have always returned a descriptor
  	// opened in blocking mode. The File will continue to work,
  	// but any blocking operation will tie up a thread.
  	if f.nonblock {
  		syscall.SetNonblock(f.pfd.Sysfd, false)
  	}
  
  	return uintptr(f.pfd.Sysfd)
  }
  
  // NewFile returns a new File with the given file descriptor and
  // name. The returned value will be nil if fd is not a valid file
  // descriptor.
  func NewFile(fd uintptr, name string) *File {
  	return newFile(fd, name, false)
  }
  
  // newFile is like NewFile, but if pollable is true it tries to add the
  // file to the runtime poller.
  func newFile(fd uintptr, name string, pollable bool) *File {
  	fdi := int(fd)
  	if fdi < 0 {
  		return nil
  	}
  	f := &File{&file{
  		pfd: poll.FD{
  			Sysfd:         fdi,
  			IsStream:      true,
  			ZeroReadIsEOF: true,
  		},
  		name: name,
  	}}
  
  	// Don't try to use kqueue with regular files on FreeBSD.
  	// It crashes the system unpredictably while running all.bash.
  	// Issue 19093.
  	if runtime.GOOS == "freebsd" {
  		pollable = false
  	}
  
  	if err := f.pfd.Init("file", pollable); err != nil {
  		// An error here indicates a failure to register
  		// with the netpoll system. That can happen for
  		// a file descriptor that is not supported by
  		// epoll/kqueue; for example, disk files on
  		// GNU/Linux systems. We assume that any real error
  		// will show up in later I/O.
  	} else if pollable {
  		// We successfully registered with netpoll, so put
  		// the file into nonblocking mode.
  		if err := syscall.SetNonblock(fdi, true); err == nil {
  			f.nonblock = true
  		}
  	}
  
  	runtime.SetFinalizer(f.file, (*file).close)
  	return f
  }
  
  // Auxiliary information if the File describes a directory
  type dirInfo struct {
  	buf  []byte // buffer for directory I/O
  	nbuf int    // length of buf; return value from Getdirentries
  	bufp int    // location of next record in buf.
  }
  
  // epipecheck raises SIGPIPE if we get an EPIPE error on standard
  // output or standard error. See the SIGPIPE docs in os/signal, and
  // issue 11845.
  func epipecheck(file *File, e error) {
  	if e == syscall.EPIPE && (file.pfd.Sysfd == 1 || file.pfd.Sysfd == 2) {
  		sigpipe()
  	}
  }
  
  // DevNull is the name of the operating system's ``null device.''
  // On Unix-like systems, it is "/dev/null"; on Windows, "NUL".
  const DevNull = "/dev/null"
  
  // OpenFile is the generalized open call; most users will use Open
  // or Create instead. It opens the named file with specified flag
  // (O_RDONLY etc.) and perm, (0666 etc.) if applicable. If successful,
  // methods on the returned File can be used for I/O.
  // If there is an error, it will be of type *PathError.
  func OpenFile(name string, flag int, perm FileMode) (*File, error) {
  	chmod := false
  	if !supportsCreateWithStickyBit && flag&O_CREATE != 0 && perm&ModeSticky != 0 {
  		if _, err := Stat(name); IsNotExist(err) {
  			chmod = true
  		}
  	}
  
  	var r int
  	for {
  		var e error
  		r, e = syscall.Open(name, flag|syscall.O_CLOEXEC, syscallMode(perm))
  		if e == nil {
  			break
  		}
  
  		// On OS X, sigaction(2) doesn't guarantee that SA_RESTART will cause
  		// open(2) to be restarted for regular files. This is easy to reproduce on
  		// fuse file systems (see http://golang.org/issue/11180).
  		if runtime.GOOS == "darwin" && e == syscall.EINTR {
  			continue
  		}
  
  		return nil, &PathError{"open", name, e}
  	}
  
  	// open(2) itself won't handle the sticky bit on *BSD and Solaris
  	if chmod {
  		Chmod(name, perm)
  	}
  
  	// There's a race here with fork/exec, which we are
  	// content to live with. See ../syscall/exec_unix.go.
  	if !supportsCloseOnExec {
  		syscall.CloseOnExec(r)
  	}
  
  	return newFile(uintptr(r), name, true), nil
  }
  
  // Close closes the File, rendering it unusable for I/O.
  // It returns an error, if any.
  func (f *File) Close() error {
  	if f == nil {
  		return ErrInvalid
  	}
  	return f.file.close()
  }
  
  func (file *file) close() error {
  	if file == nil {
  		return syscall.EINVAL
  	}
  	var err error
  	if e := file.pfd.Close(); e != nil {
  		if e == poll.ErrFileClosing {
  			e = ErrClosed
  		}
  		err = &PathError{"close", file.name, e}
  	}
  
  	// no need for a finalizer anymore
  	runtime.SetFinalizer(file, nil)
  	return err
  }
  
  // read reads up to len(b) bytes from the File.
  // It returns the number of bytes read and an error, if any.
  func (f *File) read(b []byte) (n int, err error) {
  	n, err = f.pfd.Read(b)
  	runtime.KeepAlive(f)
  	return n, err
  }
  
  // pread reads len(b) bytes from the File starting at byte offset off.
  // It returns the number of bytes read and the error, if any.
  // EOF is signaled by a zero count with err set to nil.
  func (f *File) pread(b []byte, off int64) (n int, err error) {
  	n, err = f.pfd.Pread(b, off)
  	runtime.KeepAlive(f)
  	return n, err
  }
  
  // write writes len(b) bytes to the File.
  // It returns the number of bytes written and an error, if any.
  func (f *File) write(b []byte) (n int, err error) {
  	n, err = f.pfd.Write(b)
  	runtime.KeepAlive(f)
  	return n, err
  }
  
  // pwrite writes len(b) bytes to the File starting at byte offset off.
  // It returns the number of bytes written and an error, if any.
  func (f *File) pwrite(b []byte, off int64) (n int, err error) {
  	n, err = f.pfd.Pwrite(b, off)
  	runtime.KeepAlive(f)
  	return n, err
  }
  
  // seek sets the offset for the next Read or Write on file to offset, interpreted
  // according to whence: 0 means relative to the origin of the file, 1 means
  // relative to the current offset, and 2 means relative to the end.
  // It returns the new offset and an error, if any.
  func (f *File) seek(offset int64, whence int) (ret int64, err error) {
  	ret, err = f.pfd.Seek(offset, whence)
  	runtime.KeepAlive(f)
  	return ret, err
  }
  
  // Truncate changes the size of the named file.
  // If the file is a symbolic link, it changes the size of the link's target.
  // If there is an error, it will be of type *PathError.
  func Truncate(name string, size int64) error {
  	if e := syscall.Truncate(name, size); e != nil {
  		return &PathError{"truncate", name, e}
  	}
  	return nil
  }
  
  // Remove removes the named file or directory.
  // If there is an error, it will be of type *PathError.
  func Remove(name string) error {
  	// System call interface forces us to know
  	// whether name is a file or directory.
  	// Try both: it is cheaper on average than
  	// doing a Stat plus the right one.
  	e := syscall.Unlink(name)
  	if e == nil {
  		return nil
  	}
  	e1 := syscall.Rmdir(name)
  	if e1 == nil {
  		return nil
  	}
  
  	// Both failed: figure out which error to return.
  	// OS X and Linux differ on whether unlink(dir)
  	// returns EISDIR, so can't use that. However,
  	// both agree that rmdir(file) returns ENOTDIR,
  	// so we can use that to decide which error is real.
  	// Rmdir might also return ENOTDIR if given a bad
  	// file path, like /etc/passwd/foo, but in that case,
  	// both errors will be ENOTDIR, so it's okay to
  	// use the error from unlink.
  	if e1 != syscall.ENOTDIR {
  		e = e1
  	}
  	return &PathError{"remove", name, e}
  }
  
  func tempDir() string {
  	dir := Getenv("TMPDIR")
  	if dir == "" {
  		if runtime.GOOS == "android" {
  			dir = "/data/local/tmp"
  		} else {
  			dir = "/tmp"
  		}
  	}
  	return dir
  }
  
  // Link creates newname as a hard link to the oldname file.
  // If there is an error, it will be of type *LinkError.
  func Link(oldname, newname string) error {
  	e := syscall.Link(oldname, newname)
  	if e != nil {
  		return &LinkError{"link", oldname, newname, e}
  	}
  	return nil
  }
  
  // Symlink creates newname as a symbolic link to oldname.
  // If there is an error, it will be of type *LinkError.
  func Symlink(oldname, newname string) error {
  	e := syscall.Symlink(oldname, newname)
  	if e != nil {
  		return &LinkError{"symlink", oldname, newname, e}
  	}
  	return nil
  }
  

View as plain text