[src/trunk]: src/sys/kern In mi_switch(), also call pserialize_switchpoint() ...

[bouyer <bouyer%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/f5558e19c373
branches:  trunk
changeset: 991651:f5558e19c373
user:      bouyer <bouyer%NetBSD.org@localhost>
date:      Tue Jul 24 15:09:37 2018 +0000

description:
In mi_switch(), also call pserialize_switchpoint() if we're not switching
to another lwp, as proposed on
http://mail-index.netbsd.org/tech-kern/2018/07/20/msg023709.html

Without it, on a SMP machine with few processes running (e.g while
running sysinst), pserialize could hang for a long time until all
CPUs got a LWP to run (or, eventually, forever).
Tested on Xen domUs with 4 CPUs, and on a 64-threads AMD machine.

diffstat:

 sys/kern/kern_synch.c |  5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

diffs (26 lines):

diff -r 37eaef011b3d -r f5558e19c373 sys/kern/kern_synch.c
--- a/sys/kern/kern_synch.c     Tue Jul 24 13:48:48 2018 +0000
+++ b/sys/kern/kern_synch.c     Tue Jul 24 15:09:37 2018 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: kern_synch.c,v 1.316 2018/07/12 10:46:48 maxv Exp $    */
+/*     $NetBSD: kern_synch.c,v 1.317 2018/07/24 15:09:37 bouyer Exp $  */
 
 /*-
  * Copyright (c) 1999, 2000, 2004, 2006, 2007, 2008, 2009
@@ -69,7 +69,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: kern_synch.c,v 1.316 2018/07/12 10:46:48 maxv Exp $");
+__KERNEL_RCSID(0, "$NetBSD: kern_synch.c,v 1.317 2018/07/24 15:09:37 bouyer Exp $");
 
 #include "opt_kstack.h"
 #include "opt_dtrace.h"
@@ -745,6 +745,7 @@
                retval = 1;
        } else {
                /* Nothing to do - just unlock and return. */
+               pserialize_switchpoint();
                mutex_spin_exit(spc->spc_mutex);
                lwp_unlock(l);
                retval = 0;