Refactor software GSO code.

Software GSO implementation currently has a complicated code path with
implicit assumptions that all packets to WritePackets carry same Data
and it does this to avoid allocations on the path etc. But this makes it
hard to reuse the WritePackets API.

This change breaks all such assumptions by introducing a new Vectorised
View API ReadToVV which can be used to cleanly split a VV into multiple
independent VVs. Further this change also makes packet buffers linkable
to form an intrusive list. This allows us to get rid of the array of
packet buffers that are passed in the WritePackets API call and replace
it with a list of packet buffers.

While this code does introduce some more allocations in the benchmarks
it doesn't cause any degradation.

Updates #231

PiperOrigin-RevId: 304731742

1 files changed, 16 insertions, 6 deletions

diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go
index 0a448b57c..2e6f42b92 100644
--- a/pkg/sentry/kernel/kernel.go
+++ b/pkg/sentry/kernel/kernel.go
@@ -564,15 +564,25 @@ func (ts *TaskSet) unregisterEpollWaiters() {
 
 	ts.mu.RLock()
 	defer ts.mu.RUnlock()
+
+	// Tasks that belong to the same process could potentially point to the
+	// same FDTable. So we retain a map of processed ones to avoid
+	// processing the same FDTable multiple times.
+	processed := make(map[*FDTable]struct{})
 	for t := range ts.Root.tids {
 		// We can skip locking Task.mu here since the kernel is paused.
-		if t.fdTable != nil {
-			t.fdTable.forEach(func(_ int32, file *fs.File, _ *vfs.FileDescription, _ FDFlags) {
-				if e, ok := file.FileOperations.(*epoll.EventPoll); ok {
-					e.UnregisterEpollWaiters()
-				}
-			})
+		if t.fdTable == nil {
+			continue
+		}
+		if _, ok := processed[t.fdTable]; ok {
+			continue
 		}
+		t.fdTable.forEach(func(_ int32, file *fs.File, _ *vfs.FileDescription, _ FDFlags) {
+			if e, ok := file.FileOperations.(*epoll.EventPoll); ok {
+				e.UnregisterEpollWaiters()
+			}
+		})
+		processed[t.fdTable] = struct{}{}
 	}
 }