rclone/vfs/write.go
Nick Craig-Wood d377842395 vfs: make write without cache more efficient
This updates the out of sequence write code to be more efficient using
a conditional lock with a timeout.
2019-08-08 12:37:50 +01:00

343 lines
9.8 KiB
Go

package vfs
import (
"context"
"io"
"os"
"sync"
"sync/atomic"
"time"
"github.com/rclone/rclone/fs"
"github.com/rclone/rclone/fs/operations"
)
// WriteFileHandle is an open for write handle on a File
type WriteFileHandle struct {
baseHandle
mu sync.Mutex
cond *sync.Cond // cond lock for out of sequence writes
closed bool // set if handle has been closed
remote string
pipeWriter *io.PipeWriter
o fs.Object
result chan error
file *File
writeCalled bool // set the first time Write() is called
offset int64
opened bool
flags int
truncated bool
}
// Check interfaces
var (
_ io.Writer = (*WriteFileHandle)(nil)
_ io.WriterAt = (*WriteFileHandle)(nil)
_ io.Closer = (*WriteFileHandle)(nil)
)
func newWriteFileHandle(d *Dir, f *File, remote string, flags int) (*WriteFileHandle, error) {
fh := &WriteFileHandle{
remote: remote,
flags: flags,
result: make(chan error, 1),
file: f,
}
fh.cond = sync.NewCond(&fh.mu)
fh.file.addWriter(fh)
return fh, nil
}
// returns whether it is OK to truncate the file
func (fh *WriteFileHandle) safeToTruncate() bool {
return fh.truncated || fh.flags&os.O_TRUNC != 0 || !fh.file.exists()
}
// openPending opens the file if there is a pending open
//
// call with the lock held
func (fh *WriteFileHandle) openPending() (err error) {
if fh.opened {
return nil
}
if !fh.safeToTruncate() {
fs.Errorf(fh.remote, "WriteFileHandle: Can't open for write without O_TRUNC on existing file without --vfs-cache-mode >= writes")
return EPERM
}
var pipeReader *io.PipeReader
pipeReader, fh.pipeWriter = io.Pipe()
go func() {
// NB Rcat deals with Stats.Transferring etc
o, err := operations.Rcat(context.TODO(), fh.file.d.f, fh.remote, pipeReader, time.Now())
if err != nil {
fs.Errorf(fh.remote, "WriteFileHandle.New Rcat failed: %v", err)
}
// Close the pipeReader so the pipeWriter fails with ErrClosedPipe
_ = pipeReader.Close()
fh.o = o
fh.result <- err
}()
fh.file.setSize(0)
fh.truncated = true
fh.file.d.addObject(fh.file) // make sure the directory has this object in it now
fh.opened = true
return nil
}
// String converts it to printable
func (fh *WriteFileHandle) String() string {
if fh == nil {
return "<nil *WriteFileHandle>"
}
fh.mu.Lock()
defer fh.mu.Unlock()
if fh.file == nil {
return "<nil *WriteFileHandle.file>"
}
return fh.file.String() + " (w)"
}
// Node returns the Node assocuated with this - satisfies Noder interface
func (fh *WriteFileHandle) Node() Node {
fh.mu.Lock()
defer fh.mu.Unlock()
return fh.file
}
// WriteAt writes len(p) bytes from p to the underlying data stream at offset
// off. It returns the number of bytes written from p (0 <= n <= len(p)) and
// any error encountered that caused the write to stop early. WriteAt must
// return a non-nil error if it returns n < len(p).
//
// If WriteAt is writing to a destination with a seek offset, WriteAt should
// not affect nor be affected by the underlying seek offset.
//
// Clients of WriteAt can execute parallel WriteAt calls on the same
// destination if the ranges do not overlap.
//
// Implementations must not retain p.
func (fh *WriteFileHandle) WriteAt(p []byte, off int64) (n int, err error) {
fh.mu.Lock()
defer fh.mu.Unlock()
return fh.writeAt(p, off)
}
// Implementatino of WriteAt - call with lock held
func (fh *WriteFileHandle) writeAt(p []byte, off int64) (n int, err error) {
// defer log.Trace(fh.remote, "len=%d off=%d", len(p), off)("n=%d, fh.off=%d, err=%v", &n, &fh.offset, &err)
if fh.closed {
fs.Errorf(fh.remote, "WriteFileHandle.Write: error: %v", EBADF)
return 0, ECLOSED
}
if fh.offset != off {
// Set a background timer so we don't wait forever
timeout := time.NewTimer(10 * time.Second)
done := make(chan struct{})
abort := int32(0)
go func() {
select {
case <-timeout.C:
// set abort flag an give all the waiting goroutines a kick on timeout
atomic.StoreInt32(&abort, 1)
fh.cond.Broadcast()
case <-done:
}
}()
// Wait for an in-sequence write or abort
for fh.offset != off && atomic.LoadInt32(&abort) == 0 {
// fs.Debugf(fh.remote, "waiting for in-sequence write to %d", off)
fh.cond.Wait()
}
// tidy up end timer
close(done)
timeout.Stop()
}
if fh.offset != off {
fs.Errorf(fh.remote, "WriteFileHandle.Write: can't seek in file without --vfs-cache-mode >= writes")
return 0, ESPIPE
}
if err = fh.openPending(); err != nil {
return 0, err
}
fh.writeCalled = true
n, err = fh.pipeWriter.Write(p)
fh.offset += int64(n)
fh.file.setSize(fh.offset)
if err != nil {
fs.Errorf(fh.remote, "WriteFileHandle.Write error: %v", err)
return 0, err
}
// fs.Debugf(fh.remote, "WriteFileHandle.Write OK (%d bytes written)", n)
fh.cond.Broadcast() // wake everyone up waiting for an in-sequence read
return n, nil
}
// Write writes len(p) bytes from p to the underlying data stream. It returns
// the number of bytes written from p (0 <= n <= len(p)) and any error
// encountered that caused the write to stop early. Write must return a non-nil
// error if it returns n < len(p). Write must not modify the slice data, even
// temporarily.
//
// Implementations must not retain p.
func (fh *WriteFileHandle) Write(p []byte) (n int, err error) {
fh.mu.Lock()
defer fh.mu.Unlock()
// Since we can't seek, just call WriteAt with the current offset
return fh.writeAt(p, fh.offset)
}
// WriteString a string to the file
func (fh *WriteFileHandle) WriteString(s string) (n int, err error) {
return fh.Write([]byte(s))
}
// Offset returns the offset of the file pointer
func (fh *WriteFileHandle) Offset() (offset int64) {
fh.mu.Lock()
defer fh.mu.Unlock()
return fh.offset
}
// close the file handle returning EBADF if it has been
// closed already.
//
// Must be called with fh.mu held
func (fh *WriteFileHandle) close() (err error) {
if fh.closed {
return ECLOSED
}
fh.closed = true
// leave writer open until file is transferred
defer func() {
fh.file.delWriter(fh, false)
fh.file.finishWriterClose()
}()
// If file not opened and not safe to truncate then then leave file intact
if !fh.opened && !fh.safeToTruncate() {
return nil
}
if err = fh.openPending(); err != nil {
return err
}
writeCloseErr := fh.pipeWriter.Close()
err = <-fh.result
if err == nil {
fh.file.setObject(fh.o)
err = writeCloseErr
}
return err
}
// Close closes the file
func (fh *WriteFileHandle) Close() error {
fh.mu.Lock()
defer fh.mu.Unlock()
return fh.close()
}
// Flush is called on each close() of a file descriptor. So if a
// filesystem wants to return write errors in close() and the file has
// cached dirty data, this is a good place to write back data and
// return any errors. Since many applications ignore close() errors
// this is not always useful.
//
// NOTE: The flush() method may be called more than once for each
// open(). This happens if more than one file descriptor refers to an
// opened file due to dup(), dup2() or fork() calls. It is not
// possible to determine if a flush is final, so each flush should be
// treated equally. Multiple write-flush sequences are relatively
// rare, so this shouldn't be a problem.
//
// Filesystems shouldn't assume that flush will always be called after
// some writes, or that if will be called at all.
func (fh *WriteFileHandle) Flush() error {
fh.mu.Lock()
defer fh.mu.Unlock()
if fh.closed {
fs.Debugf(fh.remote, "WriteFileHandle.Flush nothing to do")
return nil
}
// fs.Debugf(fh.remote, "WriteFileHandle.Flush")
// If Write hasn't been called then ignore the Flush - Release
// will pick it up
if !fh.writeCalled {
fs.Debugf(fh.remote, "WriteFileHandle.Flush unwritten handle, writing 0 bytes to avoid race conditions")
_, err := fh.writeAt([]byte{}, fh.offset)
return err
}
err := fh.close()
if err != nil {
fs.Errorf(fh.remote, "WriteFileHandle.Flush error: %v", err)
} else {
// fs.Debugf(fh.remote, "WriteFileHandle.Flush OK")
}
return err
}
// Release is called when we are finished with the file handle
//
// It isn't called directly from userspace so the error is ignored by
// the kernel
func (fh *WriteFileHandle) Release() error {
fh.mu.Lock()
defer fh.mu.Unlock()
if fh.closed {
fs.Debugf(fh.remote, "WriteFileHandle.Release nothing to do")
return nil
}
fs.Debugf(fh.remote, "WriteFileHandle.Release closing")
err := fh.close()
if err != nil {
fs.Errorf(fh.remote, "WriteFileHandle.Release error: %v", err)
} else {
// fs.Debugf(fh.remote, "WriteFileHandle.Release OK")
}
return err
}
// Stat returns info about the file
func (fh *WriteFileHandle) Stat() (os.FileInfo, error) {
fh.mu.Lock()
defer fh.mu.Unlock()
return fh.file, nil
}
// Truncate file to given size
func (fh *WriteFileHandle) Truncate(size int64) (err error) {
fh.mu.Lock()
defer fh.mu.Unlock()
if fh.closed {
return ECLOSED
}
if size != fh.offset {
fs.Errorf(fh.remote, "WriteFileHandle: Truncate: Can't change size without --vfs-cache-mode >= writes")
return EPERM
}
// File is correct size
if size == 0 {
fh.truncated = true
}
return nil
}
// Read reads up to len(p) bytes into p.
func (fh *WriteFileHandle) Read(p []byte) (n int, err error) {
fs.Errorf(fh.remote, "WriteFileHandle: Read: Can't read and write to file without --vfs-cache-mode >= minimal")
return 0, EPERM
}
// ReadAt reads len(p) bytes into p starting at offset off in the
// underlying input source. It returns the number of bytes read (0 <=
// n <= len(p)) and any error encountered.
func (fh *WriteFileHandle) ReadAt(p []byte, off int64) (n int, err error) {
fs.Errorf(fh.remote, "WriteFileHandle: ReadAt: Can't read and write to file without --vfs-cache-mode >= minimal")
return 0, EPERM
}
// Sync commits the current contents of the file to stable storage. Typically,
// this means flushing the file system's in-memory copy of recently written
// data to disk.
func (fh *WriteFileHandle) Sync() error {
return nil
}