From 75c483ec6443f731cc6f2149c4738547eb602c6f Mon Sep 17 00:00:00 2001
From: Paul Buetow <paul@buetow.org>
Date: Fri, 8 May 2026 23:18:52 +0300
Subject: swap global filter in place to skip BPF reattach

Changing the global filter used to call stopTrace + beginTraceCmd, which
detached and re-attached every tracepoint and re-loaded the BPF object.
On heavily loaded I/O systems that took several seconds and showed an
'Attaching tracepoints...' overlay each time. The probe set never depends
on the global filter (ShouldIAttachTracepoint only reads CLI regex flags),
so the restart was gratuitous.

Now the eventloop holds its filter behind atomic.Pointer with SetFilter /
Filter accessors, and the trace starter registers el.SetFilter via the
runtime bindings as a SetLiveFilterSetter callback. applyGlobalFilter
and undoGlobalFilter call runtime.applyLiveFilter first; only if no
trace is running do they fall back to the full restart path.
---
 internal/eventloop.go | 27 +++++++++++++++++++++++++--
 1 file changed, 25 insertions(+), 2 deletions(-)

(limited to 'internal/eventloop.go')

diff --git a/internal/eventloop.go b/internal/eventloop.go
index 87f99ed..fc82ee4 100644
--- a/internal/eventloop.go
+++ b/internal/eventloop.go
@@ -2,6 +2,7 @@ package internal
 
 import (
 	"fmt"
+	"sync/atomic"
 	"time"
 
 	"ior/internal/event"
@@ -38,7 +39,11 @@ type eventLoopConfig struct {
 type rawEventHandler func(raw []byte, ch chan<- *event.Pair)
 
 type eventLoop struct {
-	filter         globalfilter.Filter
+	// filterPtr holds the active global filter. Stored as atomic.Pointer so
+	// the TUI can swap filters in place via SetFilter without tearing down
+	// and reattaching the BPF probes (the previous behavior caused a multi-
+	// second 'Attaching tracepoints' overlay every time the filter changed).
+	filterPtr      atomic.Pointer[globalfilter.Filter]
 	pairs          pairTracker           // enter/exit pairing state and inter-syscall duration tracking
 	pendingHandles *pendingHandleTracker // TID → pathname from name_to_handle_at, for open_by_handle_at correlation
 	fdTracker      *fdTracker            // fd table and procfs resolution cache
@@ -57,6 +62,24 @@ type eventLoop struct {
 	done                    chan struct{}
 }
 
+// Filter returns a snapshot of the currently active global filter. Each call
+// loads a single atomic pointer and returns the underlying value, so the
+// caller observes a consistent filter even if SetFilter races concurrently.
+func (e *eventLoop) Filter() globalfilter.Filter {
+	if p := e.filterPtr.Load(); p != nil {
+		return *p
+	}
+	return globalfilter.Filter{}
+}
+
+// SetFilter atomically replaces the active global filter. The replacement is
+// cloned so the caller can keep mutating its own filter without affecting
+// what the eventloop sees.
+func (e *eventLoop) SetFilter(filter globalfilter.Filter) {
+	cloned := filter.Clone()
+	e.filterPtr.Store(&cloned)
+}
+
 func newEventLoop(cfg eventLoopConfig) (*eventLoop, error) {
 	fdState := configuredFDTracker(cfg.fdTracker)
 	commState := configuredCommResolver(cfg.commResolver)
@@ -65,7 +88,6 @@ func newEventLoop(cfg eventLoopConfig) (*eventLoop, error) {
 	}
 
 	el := &eventLoop{
-		filter:         cfg.filter.Clone(),
 		pairs:          newPairTracker(),
 		pendingHandles: newPendingHandleTracker(),
 		fdTracker:      fdState,
@@ -75,6 +97,7 @@ func newEventLoop(cfg eventLoopConfig) (*eventLoop, error) {
 		cfg:            cfg,
 		done:           make(chan struct{}),
 	}
+	el.SetFilter(cfg.filter)
 	el.initRawHandlers()
 	el.configureOutputCallback()
 	el.seedTrackedPidComm()
-- 
cgit v1.2.3