the definitive tun offloads branch (#1704)

util/cpupin_linux.go

@@ -0,0 +1,49 @@
+//go:build linux && !android && !e2e_testing
+
+package util
+
+import (
+	"runtime"
+
+	"golang.org/x/sys/unix"
+)
+
+// PinThreadToCPU restricts the calling OS thread to the given CPU via
+// sched_setaffinity(2). Combined with runtime.LockOSThread on the
+// goroutine, this prevents the kernel from migrating us across CPUs and
+// in turn keeps every sendmmsg from this goroutine going through the
+// same XPS-selected TX ring, eliminating the wire-side reorder that
+// otherwise fragments one nebula flow across multiple rings.
+func PinThreadToCPU(cpu int) error {
+	runtime.LockOSThread()
+	var set unix.CPUSet
+	set.Zero()
+	set.Set(cpu)
+	if err := unix.SchedSetaffinity(0, &set); err != nil {
+		// Without the affinity the thread lock buys no TX-ring stability;
+		// don't leave the goroutine wedded to one OS thread for nothing.
+		runtime.UnlockOSThread()
+		return err
+	}
+	return nil
+}
+
+// AllowedCPUs returns the CPU IDs the calling process is currently allowed to
+// run on, as reported by sched_getaffinity(2). Under a cgroup cpuset or a
+// `taskset` mask the allowed IDs are frequently not the contiguous range
+// 0..NumCPU-1 (e.g. pinned to CPUs 4-7: NumCPU reports 4 while the valid IDs
+// are 4,5,6,7). Callers that need a real CPU to pin to must choose from this
+// set rather than assuming i % NumCPU is runnable, or every pin fails.
+func AllowedCPUs() ([]int, error) {
+	var set unix.CPUSet
+	if err := unix.SchedGetaffinity(0, &set); err != nil {
+		return nil, err
+	}
+	cpus := make([]int, 0, set.Count())
+	for cpu := 0; cpu < len(set)*64; cpu++ {
+		if set.IsSet(cpu) {
+			cpus = append(cpus, cpu)
+		}
+	}
+	return cpus, nil
+}

util/cpupin_other.go

@@ -0,0 +1,18 @@
+//go:build !linux || android || e2e_testing
+
+package util
+
+// PinThreadToCPU is a no-op outside Linux: only Linux exposes a stable
+// per-thread CPU affinity API and only Linux has XPS-driven TX ring
+// selection in the first place. On every other platform there's nothing
+// to fix here.
+func PinThreadToCPU(_ int) error {
+	return nil
+}
+
+// AllowedCPUs has no meaningful answer off Linux (no sched_getaffinity), so it
+// reports "unknown" by returning a nil slice and nil error. Callers treat an
+// empty result as "fall back to the default CPU choice".
+func AllowedCPUs() ([]int, error) {
+	return nil, nil
+}