fix: surface rb.Poll errors instead of silently dropping them

The libbpfgo RingBuffer.Poll() previously launched a background goroutine
that could fail with a fatal C-level errno (e.g. EBADF, EPERM) and silently
exit, causing BPF events to stop flowing with no diagnostic.

Changes:
- Patch ../libbpfgo (replace directive) so RingBuffer.Poll() returns a
  buffered <-chan error that the background goroutine sends to on failure.
- In setupEventChannel, capture the error channel and spawn a monitoring
  goroutine that logs any poll error to stderr.
- Update go.mod with the replace directive and drop the now-unused remote
  checksum entries from go.sum.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

1 files changed, 11 insertions, 1 deletions

diff --git a/internal/ior_bpfsetup.go b/internal/ior_bpfsetup.go
index 885d321..7575700 100644
--- a/internal/ior_bpfsetup.go
+++ b/internal/ior_bpfsetup.go
@@ -88,13 +88,23 @@ func setupBPFModule(parentCtx context.Context, cfg flags.Config) (*bpf.Module, *
 }
 
 // setupEventChannel initialises the BPF ring-buffer and returns the event channel.
+// It starts polling the ring buffer and logs any fatal poll error to stderr so
+// that a background ring-buffer failure does not go completely unnoticed.
 func setupEventChannel(bpfModule *bpf.Module) (chan []byte, error) {
 	ch := make(chan []byte, appconfig.DefaultChannelBufferSize)
 	rb, err := bpfModule.InitRingBuf("event_map", ch)
 	if err != nil {
 		return nil, err
 	}
-	rb.Poll(300)
+	// Poll returns a channel that receives any fatal error from the background
+	// polling goroutine. Monitor it in a goroutine so the error is logged and
+	// does not silently drop events.
+	pollErrCh := rb.Poll(300)
+	go func() {
+		if err := <-pollErrCh; err != nil {
+			fmt.Fprintf(os.Stderr, "ior: ring buffer poll failed: %v\n", err)
+		}
+	}()
 	return ch, nil
 }