Disabled async reads.

Allowing the kernel to send multiple reads for the same file handle
concurrently interferes with sequential read detection like that in
Aaron Jacobs 8 years ago
parent b5c1fd6ddc
commit 348ed9e731

@ -165,10 +165,6 @@ func (c *Connection) Init() (err error) {
// Tell the kernel not to use pitifully small 4 KiB writes.
initOp.Flags |= fusekernel.InitBigWrites
// Tell the kernel it is free to send further requests while a read request
// is in flight.
initOp.Flags |= fusekernel.InitAsyncRead
c.Reply(ctx, nil)

@ -19,7 +19,6 @@ import (
@ -73,18 +72,6 @@ func (t *InterruptFSTest) StatFoo() {
func (t *InterruptFSTest) InterruptedDuringRead() {
// On Linux, since we have async reads enabled, the kernel sends the read and
// the flush ops in parallel. When the process receives SIGINT, the interrupt
// is delivered only for the flush, probably because that's what the process
// appears to be blocking on. So this test doesn't work.
// Note that this means that cancellation is not delivered for reads on
// Linux. This is unfortunate, but probably worth it due to the significant
// increase in performance.
if runtime.GOOS == "linux" {
var err error