[src/trunk]: src/sys/kern PR kern/41923: assertion "cur != owner" failed

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

details:   https://anonhg.NetBSD.org/src/rev/5a7f53a2bf1b
branches:  trunk
changeset: 747393:5a7f53a2bf1b
user:      bouyer <bouyer%NetBSD.org@localhost>
date:      Sun Sep 13 14:38:20 2009 +0000

description:
PR kern/41923: assertion "cur != owner" failed
In the for(;;) loop of turnstile_block(), the lock owner can change while
cur's lock is released (cur's lock is also the tschain_t's mutex).
Remove the KASSERT about owner being invariant and try to deal with the
fact that the owner can change instead.
http://mail-index.netbsd.org/tech-kern/2009/08/24/msg005957.html
and followups.

diffstat:

 sys/kern/kern_turnstile.c |  29 +++++++++++++++++++++++++----
 1 files changed, 25 insertions(+), 4 deletions(-)

diffs (72 lines):

diff -r 2fe60078e798 -r 5a7f53a2bf1b sys/kern/kern_turnstile.c
--- a/sys/kern/kern_turnstile.c Sun Sep 13 14:30:21 2009 +0000
+++ b/sys/kern/kern_turnstile.c Sun Sep 13 14:38:20 2009 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: kern_turnstile.c,v 1.24 2009/03/21 13:11:14 ad Exp $   */
+/*     $NetBSD: kern_turnstile.c,v 1.25 2009/09/13 14:38:20 bouyer Exp $       */
 
 /*-
  * Copyright (c) 2002, 2006, 2007, 2009 The NetBSD Foundation, Inc.
@@ -60,7 +60,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: kern_turnstile.c,v 1.24 2009/03/21 13:11:14 ad Exp $");
+__KERNEL_RCSID(0, "$NetBSD: kern_turnstile.c,v 1.25 2009/09/13 14:38:20 bouyer Exp $");
 
 #include <sys/param.h>
 #include <sys/lockdebug.h>
@@ -253,6 +253,7 @@
        sq = &ts->ts_sleepq[q];
        ts->ts_waiters[q]++;
        sleepq_enter(sq, l, tc->tc_mutex);
+       /* now tc->tc_mutex is also cur->l_mutex and l->l_mutex */
        LOCKDEBUG_BARRIER(tc->tc_mutex, 1);
        l->l_kpriority = true;
        obase = l->l_kpribase;
@@ -275,6 +276,7 @@
         * compiling a kernel with LOCKDEBUG to pinpoint the problem.
         */
        prio = lwp_eprio(l);
+
        for (;;) {
                bool dolock;
 
@@ -285,9 +287,24 @@
                if (owner == NULL)
                        break;
 
-               KASSERT(l != owner);
-               KASSERT(cur != owner);
+               /* The owner may have changed as we have dropped the tc lock */
+               if (cur == owner) {
+                       /*
+                        * we own the lock: stop here, sleepq_block()
+                        * should wake up immediatly
+                        */
+                       break;
+               }
 
+               if (l == owner) {
+                       /* owner has changed, restart from curlwp */
+                       lwp_unlock(l);
+                       l = cur;
+                       lwp_lock(l);
+                       prio = lwp_eprio(l);
+                       continue;
+               }
+                       
                if (l->l_mutex != owner->l_mutex)
                        dolock = true;
                else
@@ -295,6 +312,10 @@
                if (dolock && !lwp_trylock(owner)) {
                        /*
                         * restart from curlwp.
+                        * Note that there may be a livelock here:
+                        * the owner may try grabing cur's lock (which is
+                        * the tc lock) while we're trying to grab
+                        * the owner's lock.
                         */
                        lwp_unlock(l);
                        l = cur;