[src/trunk]: src/sys/kern - DIAGNOSTIC: check for leaked kernel_lock in mi_sw...

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

details:   https://anonhg.NetBSD.org/src/rev/9049e6991837
branches:  trunk
changeset: 968618:9049e6991837
user:      ad <ad%NetBSD.org@localhost>
date:      Wed Jan 22 13:19:33 2020 +0000

description:
- DIAGNOSTIC: check for leaked kernel_lock in mi_switch().

- Now that ci_biglock_wanted is set later, explicitly disable preemption
  while acquiring kernel_lock.  It was blocked in a roundabout way
  previously.

Reported-by: syzbot+43111d810160fb4b978b%syzkaller.appspotmail.com@localhost
Reported-by: syzbot+f5b871bd00089bf97286%syzkaller.appspotmail.com@localhost
Reported-by: syzbot+cd1f15eee5b1b6d20078%syzkaller.appspotmail.com@localhost
Reported-by: syzbot+fb945a331dabd0b6ba9e%syzkaller.appspotmail.com@localhost
Reported-by: syzbot+53a0c2342b361db25240%syzkaller.appspotmail.com@localhost
Reported-by: syzbot+552222a952814dede7d1%syzkaller.appspotmail.com@localhost
Reported-by: syzbot+c7104a72172b0f9093a4%syzkaller.appspotmail.com@localhost
Reported-by: syzbot+efbd30c6ca0f7d8440e8%syzkaller.appspotmail.com@localhost
Reported-by: syzbot+330a421bd46794d8b750%syzkaller.appspotmail.com@localhost

diffstat:

 sys/kern/kern_lock.c  |  14 ++++++++++----
 sys/kern/kern_synch.c |   8 ++++----
 2 files changed, 14 insertions(+), 8 deletions(-)

diffs (79 lines):

diff -r 81559fe1df7f -r 9049e6991837 sys/kern/kern_lock.c
--- a/sys/kern/kern_lock.c      Wed Jan 22 12:44:54 2020 +0000
+++ b/sys/kern/kern_lock.c      Wed Jan 22 13:19:33 2020 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: kern_lock.c,v 1.165 2020/01/17 20:26:22 ad Exp $       */
+/*     $NetBSD: kern_lock.c,v 1.166 2020/01/22 13:19:33 ad Exp $       */
 
 /*-
  * Copyright (c) 2002, 2006, 2007, 2008, 2009, 2020 The NetBSD Foundation, Inc.
@@ -31,7 +31,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: kern_lock.c,v 1.165 2020/01/17 20:26:22 ad Exp $");
+__KERNEL_RCSID(0, "$NetBSD: kern_lock.c,v 1.166 2020/01/22 13:19:33 ad Exp $");
 
 #include <sys/param.h>
 #include <sys/proc.h>
@@ -212,9 +212,10 @@
        owant = ci->ci_biglock_wanted;
 
        /*
-        * Spin until we acquire the lock.  Once we have it, record the
-        * time spent with lockstat.
+        * Stay pinned to the CPU and spin until we acquire the lock.  Once
+        * we have it, record the time spent with lockstat.
         */
+       l->l_nopreempt++;
        LOCKSTAT_ENTER(lsflag);
        LOCKSTAT_START_TIMER(lsflag, spintime);
 
@@ -238,6 +239,11 @@
                s = splvm();
        } while (!__cpu_simple_lock_try(kernel_lock));
 
+       /*
+        * Got it; not re-enable preemption, although we now can't do a
+        * preemption as kernel_lock is held!
+        */
+       l->l_nopreempt--;
        ci->ci_biglock_count = nlocks;
        l->l_blcnt = nlocks;
        splx(s);
diff -r 81559fe1df7f -r 9049e6991837 sys/kern/kern_synch.c
--- a/sys/kern/kern_synch.c     Wed Jan 22 12:44:54 2020 +0000
+++ b/sys/kern/kern_synch.c     Wed Jan 22 13:19:33 2020 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: kern_synch.c,v 1.336 2020/01/09 16:35:03 ad Exp $      */
+/*     $NetBSD: kern_synch.c,v 1.337 2020/01/22 13:19:33 ad Exp $      */
 
 /*-
  * Copyright (c) 1999, 2000, 2004, 2006, 2007, 2008, 2009, 2019
@@ -69,7 +69,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: kern_synch.c,v 1.336 2020/01/09 16:35:03 ad Exp $");
+__KERNEL_RCSID(0, "$NetBSD: kern_synch.c,v 1.337 2020/01/22 13:19:33 ad Exp $");
 
 #include "opt_kstack.h"
 #include "opt_dtrace.h"
@@ -353,8 +353,7 @@
                        break;
                }
                s = splsched();
-               if (__predict_false(l->l_blcnt != 0 ||
-                   curcpu()->ci_biglock_wanted != NULL)) {
+               if (__predict_false(l->l_blcnt != 0)) {
                        /* Hold or want kernel_lock, code is not MT safe. */
                        splx(s);
                        if ((dop & DOPREEMPT_COUNTED) == 0) {
@@ -531,6 +530,7 @@
        KASSERT(lwp_locked(l, NULL));
        KASSERT(kpreempt_disabled());
        KASSERT(mutex_owned(curcpu()->ci_schedstate.spc_mutex));
+       KASSERTMSG(l->l_blcnt == 0, "kernel_lock leaked");
 
        kstack_check_magic(l);