Hi experts, Looks like the print* helper functions may run into deadlocks if the printer Goroutine is rescheduled to a different M to execute 'printunlock' as the "if mp.printlock==0" checking is likely to fail if M has changed.
Wondering is it intentional behavior? As print* are (almostly if not always) used in runtime during STW or guarded by high level locks. I didn't figure out how to produce a deadlock by user testcase, but the change below could easily deadlock 'go_bootstrap' when running './src/make.bash'. A similar issue was opened in https://github.com/golang/go/issues/54786, but no follow-up for months. Thanks. ``` diff --git a/src/runtime/mgcsweep.go b/src/runtime/mgcsweep.go index 6ccf090ac5..3336823dc0 100644 --- a/src/runtime/mgcsweep.go +++ b/src/runtime/mgcsweep.go @@ -155,6 +155,7 @@ func (a *activeSweep) begin() sweepLocker { return sweepLocker{mheap_.sweepgen, false} } if a.state.CompareAndSwap(state, state+1) { + println("activeSweep.begin, old:", state, ",new:", state+1) return sweepLocker{mheap_.sweepgen, true} } } @@ -172,6 +173,7 @@ func (a *activeSweep) end(sl sweepLocker) { throw("mismatched begin/end of activeSweep") } if a.state.CompareAndSwap(state, state-1) { + println("activeSweep.end, old:", state, ",new:", state-1) if state != sweepDrainedMask { return } ``` -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/golang-nuts/e48221e3-0709-49e1-8cbd-314044558c56n%40googlegroups.com.
