pkg/repro: fix potential deadlock

If an instance fails to boot 3 times in a row,
we drop it on the floor. If we drop all instances this way,
the repro process deadlocks infinitly.
Retry infinitly instead. There is nothing else good we can do
in this case. If the instances becomes alive again, repro will resume.

1 files changed, 5 insertions, 13 deletions

diff --git a/pkg/repro/repro.go b/pkg/repro/repro.go
index 9f0faf29c..5cfa7acda 100644
--- a/pkg/repro/repro.go
+++ b/pkg/repro/repro.go
@@ -612,8 +612,8 @@ func (ctx *context) testCProg(p *prog.Prog, duration time.Duration, opts csource
 }
 
 func (ctx *context) returnInstance(inst *reproInstance) {
-	ctx.bootRequests <- inst.index
 	inst.execProg.Close()
+	ctx.bootRequests <- inst.index
 }
 
 func (ctx *context) reproLogf(level int, format string, args ...interface{}) {
@@ -660,27 +660,19 @@ func (ctx *context) createInstances(cfg *mgrconfig.Config, vmPool *vm.Pool) {
 		go func() {
 			defer wg.Done()
 
-			var inst *instance.ExecProgInstance
-			maxTry := 3
-			for try := 0; try < maxTry; try++ {
+			for try := 0; ; try++ {
 				select {
 				case <-vm.Shutdown:
-					try = maxTry
-					continue
+					return
 				default:
 				}
-				var err error
-				inst, err = instance.CreateExecProgInstance(vmPool, vmIndex, cfg,
+				inst, err := instance.CreateExecProgInstance(vmPool, vmIndex, cfg,
 					ctx.reporter, &instance.OptionalConfig{Logf: ctx.reproLogf})
 				if err != nil {
-					ctx.reproLogf(0, "failed to init instance: %v, attempt %d/%d",
-						err, try+1, maxTry)
+					ctx.reproLogf(0, "failed to boot instance (try %v): %v", try+1, err)
 					time.Sleep(10 * time.Second)
 					continue
 				}
-				break
-			}
-			if inst != nil {
 				ctx.instances <- &reproInstance{execProg: inst, index: vmIndex}
 			}
 		}()