Streamline error handling and locking, with fix for close() race

This commit is contained in:
Jakob Borg 2014-05-04 17:40:40 +02:00
parent 10e8861f14
commit 482795bab0
1 changed files with 118 additions and 120 deletions

View File

@ -64,24 +64,24 @@ type Connection interface {
}
type rawConnection struct {
sync.RWMutex
id string
receiver Model
reader io.ReadCloser
cr *countingReader
xr *xdr.Reader
writer io.WriteCloser
cw *countingWriter
wb *bufio.Writer
xw *xdr.Writer
wmut sync.Mutex
closed bool
id string
receiver Model
reader io.ReadCloser
cr *countingReader
xr *xdr.Reader
writer io.WriteCloser
cw *countingWriter
wb *bufio.Writer
xw *xdr.Writer
closed chan struct{}
awaiting map[int]chan asyncResult
nextID int
indexSent map[string]map[string][2]int64
hasSentIndex bool
hasRecvdIndex bool
imut sync.Mutex
}
type asyncResult struct {
@ -115,7 +115,6 @@ func NewConnection(nodeID string, reader io.Reader, writer io.Writer, receiver M
cw: cw,
wb: wb,
xw: xdr.NewWriter(wb),
closed: make(chan struct{}),
awaiting: make(map[int]chan asyncResult),
indexSent: make(map[string]map[string][2]int64),
}
@ -132,11 +131,11 @@ func (c *rawConnection) ID() string {
// Index writes the list of file information to the connected peer node
func (c *rawConnection) Index(repo string, idx []FileInfo) {
c.Lock()
if c.isClosed() {
c.Unlock()
return
}
c.imut.Lock()
var msgType int
if c.indexSent[repo] == nil {
// This is the first time we send an index.
@ -159,14 +158,15 @@ func (c *rawConnection) Index(repo string, idx []FileInfo) {
idx = diff
}
header{0, c.nextID, msgType}.encodeXDR(c.xw)
_, err := IndexMessage{repo, idx}.encodeXDR(c.xw)
if err == nil {
err = c.flush()
}
id := c.nextID
c.nextID = (c.nextID + 1) & 0xfff
c.hasSentIndex = true
c.Unlock()
c.imut.Unlock()
c.wmut.Lock()
header{0, id, msgType}.encodeXDR(c.xw)
IndexMessage{repo, idx}.encodeXDR(c.xw)
err := c.flush()
c.wmut.Unlock()
if err != nil {
c.close(err)
@ -176,28 +176,30 @@ func (c *rawConnection) Index(repo string, idx []FileInfo) {
// Request returns the bytes for the specified block after fetching them from the connected peer.
func (c *rawConnection) Request(repo string, name string, offset int64, size int) ([]byte, error) {
c.Lock()
if c.isClosed() {
c.Unlock()
return nil, ErrClosed
}
c.imut.Lock()
id := c.nextID
c.nextID = (c.nextID + 1) & 0xfff
rc := make(chan asyncResult)
if _, ok := c.awaiting[c.nextID]; ok {
if _, ok := c.awaiting[id]; ok {
panic("id taken")
}
c.awaiting[c.nextID] = rc
header{0, c.nextID, messageTypeRequest}.encodeXDR(c.xw)
_, err := RequestMessage{repo, name, uint64(offset), uint32(size)}.encodeXDR(c.xw)
if err == nil {
err = c.flush()
}
c.awaiting[id] = rc
c.imut.Unlock()
c.wmut.Lock()
header{0, id, messageTypeRequest}.encodeXDR(c.xw)
RequestMessage{repo, name, uint64(offset), uint32(size)}.encodeXDR(c.xw)
err := c.flush()
c.wmut.Unlock()
if err != nil {
c.Unlock()
c.close(err)
return nil, err
}
c.nextID = (c.nextID + 1) & 0xfff
c.Unlock()
res, ok := <-rc
if !ok {
@ -208,46 +210,47 @@ func (c *rawConnection) Request(repo string, name string, offset int64, size int
// ClusterConfig send the cluster configuration message to the peer and returns any error
func (c *rawConnection) ClusterConfig(config ClusterConfigMessage) {
c.Lock()
defer c.Unlock()
if c.isClosed() {
return
}
header{0, c.nextID, messageTypeClusterConfig}.encodeXDR(c.xw)
c.imut.Lock()
id := c.nextID
c.nextID = (c.nextID + 1) & 0xfff
c.imut.Unlock()
c.wmut.Lock()
header{0, id, messageTypeClusterConfig}.encodeXDR(c.xw)
config.encodeXDR(c.xw)
err := c.flush()
c.wmut.Unlock()
_, err := config.encodeXDR(c.xw)
if err == nil {
err = c.flush()
}
if err != nil {
c.close(err)
}
}
func (c *rawConnection) ping() bool {
c.Lock()
if c.isClosed() {
c.Unlock()
return false
}
c.imut.Lock()
id := c.nextID
c.nextID = (c.nextID + 1) & 0xfff
rc := make(chan asyncResult, 1)
c.awaiting[c.nextID] = rc
header{0, c.nextID, messageTypePing}.encodeXDR(c.xw)
c.awaiting[id] = rc
c.imut.Unlock()
c.wmut.Lock()
header{0, id, messageTypePing}.encodeXDR(c.xw)
err := c.flush()
c.wmut.Unlock()
if err != nil {
c.Unlock()
c.close(err)
return false
} else if c.xw.Error() != nil {
c.Unlock()
c.close(c.xw.Error())
return false
}
c.nextID = (c.nextID + 1) & 0xfff
c.Unlock()
res, ok := <-rc
return ok && res.err == nil
@ -258,40 +261,47 @@ type flusher interface {
}
func (c *rawConnection) flush() error {
c.wb.Flush()
if err := c.xw.Error(); err != nil {
return err
}
if err := c.wb.Flush(); err != nil {
return err
}
if f, ok := c.writer.(flusher); ok {
return f.Flush()
}
return nil
}
func (c *rawConnection) close(err error) {
c.Lock()
select {
case <-c.closed:
c.Unlock()
c.imut.Lock()
c.wmut.Lock()
defer c.imut.Unlock()
defer c.wmut.Unlock()
if c.closed {
return
default:
}
close(c.closed)
c.closed = true
for _, ch := range c.awaiting {
close(ch)
}
c.awaiting = nil
c.writer.Close()
c.reader.Close()
c.Unlock()
c.receiver.Close(c.id, err)
}
func (c *rawConnection) isClosed() bool {
select {
case <-c.closed:
return true
default:
return false
}
c.wmut.Lock()
defer c.wmut.Unlock()
return c.closed
}
func (c *rawConnection) readerLoop() {
@ -299,8 +309,8 @@ loop:
for !c.isClosed() {
var hdr header
hdr.decodeXDR(c.xr)
if c.xr.Error() != nil {
c.close(c.xr.Error())
if err := c.xr.Error(); err != nil {
c.close(err)
break loop
}
if hdr.version != 0 {
@ -312,8 +322,8 @@ loop:
case messageTypeIndex:
var im IndexMessage
im.decodeXDR(c.xr)
if c.xr.Error() != nil {
c.close(c.xr.Error())
if err := c.xr.Error(); err != nil {
c.close(err)
break loop
} else {
@ -326,15 +336,12 @@ loop:
go c.receiver.Index(c.id, im.Repository, im.Files)
}
c.Lock()
c.hasRecvdIndex = true
c.Unlock()
case messageTypeIndexUpdate:
var im IndexMessage
im.decodeXDR(c.xr)
if c.xr.Error() != nil {
c.close(c.xr.Error())
if err := c.xr.Error(); err != nil {
c.close(err)
break loop
} else {
go c.receiver.IndexUpdate(c.id, im.Repository, im.Files)
@ -343,8 +350,8 @@ loop:
case messageTypeRequest:
var req RequestMessage
req.decodeXDR(c.xr)
if c.xr.Error() != nil {
c.close(c.xr.Error())
if err := c.xr.Error(); err != nil {
c.close(err)
break loop
}
go c.processRequest(hdr.msgID, req)
@ -352,16 +359,16 @@ loop:
case messageTypeResponse:
data := c.xr.ReadBytesMax(256 * 1024) // Sufficiently larger than max expected block size
if c.xr.Error() != nil {
c.close(c.xr.Error())
if err := c.xr.Error(); err != nil {
c.close(err)
break loop
}
go func(hdr header, err error) {
c.Lock()
c.imut.Lock()
rc, ok := c.awaiting[hdr.msgID]
delete(c.awaiting, hdr.msgID)
c.Unlock()
c.imut.Unlock()
if ok {
rc <- asyncResult{data, err}
@ -370,37 +377,34 @@ loop:
}(hdr, c.xr.Error())
case messageTypePing:
c.Lock()
c.wmut.Lock()
header{0, hdr.msgID, messageTypePong}.encodeXDR(c.xw)
err := c.flush()
c.Unlock()
c.wmut.Unlock()
if err != nil {
c.close(err)
break loop
} else if c.xw.Error() != nil {
c.close(c.xw.Error())
break loop
}
case messageTypePong:
c.RLock()
c.imut.Lock()
rc, ok := c.awaiting[hdr.msgID]
c.RUnlock()
if ok {
rc <- asyncResult{}
close(rc)
go func() {
rc <- asyncResult{}
close(rc)
}()
c.Lock()
delete(c.awaiting, hdr.msgID)
c.Unlock()
}
c.imut.Unlock()
case messageTypeClusterConfig:
var cm ClusterConfigMessage
cm.decodeXDR(c.xr)
if c.xr.Error() != nil {
c.close(c.xr.Error())
if err := c.xr.Error(); err != nil {
c.close(err)
break loop
} else {
go c.receiver.ClusterConfig(c.id, cm)
@ -416,15 +420,14 @@ loop:
func (c *rawConnection) processRequest(msgID int, req RequestMessage) {
data, _ := c.receiver.Request(c.id, req.Repository, req.Name, int64(req.Offset), int(req.Size))
c.Lock()
c.wmut.Lock()
header{0, msgID, messageTypeResponse}.encodeXDR(c.xw)
_, err := c.xw.WriteBytes(data)
if err == nil {
err = c.flush()
}
c.Unlock()
c.xw.WriteBytes(data)
err := c.flush()
c.wmut.Unlock()
buffers.Put(data)
if err != nil {
c.close(err)
}
@ -434,27 +437,22 @@ func (c *rawConnection) pingerLoop() {
var rc = make(chan bool, 1)
ticker := time.Tick(pingIdleTime / 2)
for {
if c.isClosed() {
return
}
select {
case <-ticker:
c.RLock()
ready := c.hasRecvdIndex && c.hasSentIndex
c.RUnlock()
if ready {
go func() {
rc <- c.ping()
}()
select {
case ok := <-rc:
if !ok {
c.close(fmt.Errorf("ping failure"))
}
case <-time.After(pingTimeout):
c.close(fmt.Errorf("ping timeout"))
go func() {
rc <- c.ping()
}()
select {
case ok := <-rc:
if !ok {
c.close(fmt.Errorf("ping failure"))
}
case <-time.After(pingTimeout):
c.close(fmt.Errorf("ping timeout"))
}
case <-c.closed:
return
}
}
}