While I'm at it, also enable large reads.

For googlecloudplatform/gcsfuse#27.

fuseops/convert.go

@@ -36,7 +36,10 @@ func Convert(r bazilfuse.Request, logger *log.Logger) (o Op) {
 
 	switch typed := r.(type) {
 	case *bazilfuse.InitRequest:
-		to := &InitOp{}
+		to := &InitOp{
+			maxReadahead: typed.MaxReadahead,
+		}
+
 		o = to
 		co = &to.commonOp
 

fuseops/ops.go

@@ -48,6 +48,8 @@ type Op interface {
 // mount to succeed.
 type InitOp struct {
 	commonOp
+
+	maxReadahead uint32
 }
 
 func (o *InitOp) Respond(err error) {
@@ -58,7 +60,7 @@ func (o *InitOp) Respond(err error) {
 
 	resp := &bazilfuse.InitResponse{}
 
-	// Unconditionally enable large WriteFile ops from the kernel on Linux.
+	// Ask the Linux kernel for larger write requests.
 	//
 	// As of 2015-03-26, the behavior in the kernel is:
 	//
@@ -77,6 +79,26 @@ func (o *InitOp) Respond(err error) {
 	resp.Flags |= bazilfuse.InitBigWrites
 	resp.MaxWrite = maxWrite
 
+	// Ask the Linux kernel for larger read requests.
+	//
+	// As of 2015-03-26, the behavior in the kernel is:
+	//
+	//  *  (http://goo.gl/bQ1f1i, http://goo.gl/HwBrR6) Set the local variable
+	//     ra_pages to be init_response->max_readahead divided by the page size.
+	//
+	//  *  (http://goo.gl/gcIsSh, http://goo.gl/LKV2vA) Set
+	//     backing_dev_info::ra_pages to the min of that value and what was sent
+	//     in the request's max_readahead field.
+	//
+	//  *  (http://goo.gl/u2SqzH) Use backing_dev_info::ra_pages when deciding
+	//     how much to read ahead.
+	//
+	//  *  (http://goo.gl/JnhbdL) Don't read ahead at all if that field is zero.
+	//
+	// Reading a page at a time is a drag. Ask for as much as the kernel is
+	// willing to give us.
+	resp.MaxReadahead = o.maxReadahead
+
 	// Respond.
 	o.commonOp.logger.Printf("Responding: %v", &resp)
 	o.r.(*bazilfuse.InitRequest).Respond(resp)