waf: end the asks to GeoJS, the DNSBL zones, AbuseIPDB and routewatch at the stop (closes #439)
check / check (push) Failing after 1m58s

Each of the four keeps the cancel function of every ask it started, and a
wait group of their goroutines. Its new method (GeoJS.StopAsking,
DNSBL.StopAsking, AbuseIPDB.StopChecking, routewatch Client.StopAsking)
refuses further asks, ends those under way, and returns once their
goroutines have. An ask ended that way is no failure: no alert, no failure
count, no wait before the next. Each provider in internal/sww adds the
method as a stop hook, so the last write of the state files follows it.

Model: opus-5-5
This commit was merged in pull request #463.
This commit is contained in:
clawbot committed 2026-11-19 19:02:12 +01:00
1 parent 805c64a256
commit e7cc0adbdb
9 files changed
+380 -67

No files matched your search

+47 -14
View File
@@ -99,8 +99,13 @@ type GeoJS struct {
// waiting are the clients without an answer: those to ask GeoJS about,
// and those it is being asked about.
waiting map[netip.Prefix]*wait
// asking is true while a request to GeoJS is under way.
asking bool
// asking is true while a request to GeoJS is under way, and endAsking
// ends it. stopped is true once StopAsking has been called, and
// askGoroutines waits for the goroutine that makes the requests.
asking bool
endAsking context.CancelFunc
stopped bool
askGoroutines sync.WaitGroup
// retryDelay is how long GeoJS is left alone after its last failure,
// zero after an answer; retryAt is when it may be asked again.
retryDelay time.Duration
@@ -210,6 +215,23 @@ func (g *GeoJS) LookUpWithoutWaiting(
return answer
}
// StopAsking ends the request to GeoJS under way, and returns once the
// goroutine that made it has. A request it ends is no failure: Failed is not
// called, and GeoJS is not left alone after it. After StopAsking, GeoJS is
// asked nothing.
func (g *GeoJS) StopAsking() {
g.mu.Lock()
g.stopped = true
if g.asking {
g.endAsking()
}
g.mu.Unlock()
g.askGoroutines.Wait()
}
// Kept returns client's answer, if one is kept, without asking GeoJS.
func (g *GeoJS) Kept(client netip.Prefix) (Answer, bool) {
g.mu.Lock()
@@ -294,7 +316,8 @@ func (g *GeoJS) answerOrWait(
}
if !g.asking {
// GeoJS is left alone after a failure, so no answer can come.
// GeoJS is left alone after a failure, or asked nothing after
// StopAsking, so no answer can come.
w.late = true
}
@@ -323,22 +346,28 @@ func (g *GeoJS) kept(client netip.Prefix) (Answer, bool) {
}
// ask starts asking GeoJS about the waiting clients, unless a request to
// it is under way or it is left alone after a failure. The requests to
// GeoJS are for every client waiting, so they go on when the caller whose
// ctx is given stops waiting.
// it is under way, it is left alone after a failure, or StopAsking has
// been called. The requests to GeoJS are for every client waiting, so they
// go on when the caller whose ctx is given stops waiting, until StopAsking.
func (g *GeoJS) ask(ctx context.Context) {
if g.asking || g.now().Before(g.retryAt) {
if g.asking || g.stopped || g.now().Before(g.retryAt) {
return
}
g.asking = true
askCtx, end := context.WithCancel(context.WithoutCancel(ctx))
g.asking, g.endAsking = true, end
go g.askAboutWaiting(context.WithoutCancel(ctx))
g.askGoroutines.Go(func() {
defer end()
g.askAboutWaiting(askCtx)
})
}
// askAboutWaiting asks GeoJS about the waiting clients, one request at a
// time, until none is left or GeoJS fails. Each answer kept is given to
// Answered, outside the lock, since Answered takes locks of its own.
// time, until none is left, GeoJS fails, or StopAsking ends ctx. Each
// answer kept is given to Answered, outside the lock, since Answered takes
// locks of its own.
func (g *GeoJS) askAboutWaiting(ctx context.Context) {
for {
clients := g.nextClients()
@@ -347,6 +376,10 @@ func (g *GeoJS) askAboutWaiting(ctx context.Context) {
}
given, err := g.request(ctx, clients)
if err != nil && ctx.Err() != nil {
continue // ended by StopAsking, which is no failure
}
kept, answered := g.keep(clients, given, err)
if g.answered != nil {
@@ -362,13 +395,13 @@ func (g *GeoJS) askAboutWaiting(ctx context.Context) {
}
// nextClients returns up to maxPerRequest of the waiting clients. When
// none is waiting, it returns none and notes that no request to GeoJS is
// under way.
// none is waiting, or StopAsking has been called, it returns none and
// notes that no request to GeoJS is under way.
func (g *GeoJS) nextClients() []netip.Prefix {
g.mu.Lock()
defer g.mu.Unlock()
if len(g.waiting) == 0 {
if len(g.waiting) == 0 || g.stopped {
g.asking = false
return nil
+38
View File
@@ -318,6 +318,44 @@ func TestCallerWhoseContextEndsEarlyDoesNotStopTheRequest(t *testing.T) {
})
}
func TestStopAskingEndsTheRequestUnderWayAsNoFailureAndAsksNoMore(
t *testing.T,
) {
t.Parallel()
synctest.Test(t, func(t *testing.T) {
service, _, g, reports := startWithReports()
clients := newClients()
service.set(hanging)
g.LookUpWithoutWaiting(t.Context(), clients())
waitForRequests(t, service, 1)
// The request is ended, not waited out, and is no failure.
began := time.Now()
g.StopAsking()
if waited := time.Since(began); waited != 0 {
t.Errorf("StopAsking returned after %s, want at once", waited)
}
reports.wantFailures(t)
// A new client is then not asked about, and its lookup does not
// wait.
service.set(answering)
wantCountry(t, g, clients(), "")
if waited := time.Since(began); waited != 0 {
t.Errorf("the lookup returned after %s, want at once", waited)
}
waitForRequests(t, service, 1)
reports.wantFailures(t)
})
}
func TestRequestOver10SecondsIsAFailure(t *testing.T) {
t.Parallel()