Module Name: src
Committed By: riz
Date: Mon Oct 1 23:07:08 UTC 2012
Modified Files:
src/sys/kern [netbsd-6]: kern_exit.c kern_lwp.c sys_lwp.c
src/sys/sys [netbsd-6]: lwp.h
Log Message:
Pull up following revision(s) (requested by rmind in ticket #583):
sys/kern/sys_lwp.c: revision 1.55
sys/sys/lwp.h: revision 1.164
sys/kern/kern_exit.c: revision 1.242
sys/kern/kern_lwp.c: revision 1.173
exit_lwps, lwp_wait: fix a race condition by re-trying if p_lock was dropped
in a case of process exit. Necessary to re-flag all LWPs for exit, as their
state might have changed or new LWPs spawned.
Should fix PR/46168 and PR/46402.
To generate a diff of this commit:
cvs rdiff -u -r1.236.2.1 -r1.236.2.2 src/sys/kern/kern_exit.c
cvs rdiff -u -r1.166 -r1.166.2.1 src/sys/kern/kern_lwp.c
cvs rdiff -u -r1.52.14.1 -r1.52.14.2 src/sys/kern/sys_lwp.c
cvs rdiff -u -r1.159.2.1 -r1.159.2.2 src/sys/sys/lwp.h
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
Modified files:
Index: src/sys/kern/kern_exit.c
diff -u src/sys/kern/kern_exit.c:1.236.2.1 src/sys/kern/kern_exit.c:1.236.2.2
--- src/sys/kern/kern_exit.c:1.236.2.1 Thu Apr 12 17:05:36 2012
+++ src/sys/kern/kern_exit.c Mon Oct 1 23:07:07 2012
@@ -1,4 +1,4 @@
-/* $NetBSD: kern_exit.c,v 1.236.2.1 2012/04/12 17:05:36 riz Exp $ */
+/* $NetBSD: kern_exit.c,v 1.236.2.2 2012/10/01 23:07:07 riz Exp $ */
/*-
* Copyright (c) 1998, 1999, 2006, 2007, 2008 The NetBSD Foundation, Inc.
@@ -67,7 +67,7 @@
*/
#include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: kern_exit.c,v 1.236.2.1 2012/04/12 17:05:36 riz Exp $");
+__KERNEL_RCSID(0, "$NetBSD: kern_exit.c,v 1.236.2.2 2012/10/01 23:07:07 riz Exp $");
#include "opt_ktrace.h"
#include "opt_perfctrs.h"
@@ -598,17 +598,12 @@ exit1(struct lwp *l, int rv)
void
exit_lwps(struct lwp *l)
{
- struct proc *p;
- struct lwp *l2;
- int error;
- lwpid_t waited;
+ proc_t *p = l->l_proc;
+ lwp_t *l2;
int nlocks;
KERNEL_UNLOCK_ALL(l, &nlocks);
- p = l->l_proc;
- KASSERT(mutex_owned(p->p_lock));
-
#ifdef KERN_SA
if (p->p_sa != NULL) {
struct sadata_vp *vp;
@@ -642,6 +637,8 @@ exit_lwps(struct lwp *l)
#endif
retry:
+ KASSERT(mutex_owned(p->p_lock));
+
/*
* Interrupt LWPs in interruptable sleep, unsuspend suspended
* LWPs and then wait for everyone else to finish.
@@ -656,30 +653,20 @@ exit_lwps(struct lwp *l)
l2->l_stat == LSSUSPENDED || l2->l_stat == LSSTOP) {
/* setrunnable() will release the lock. */
setrunnable(l2);
- DPRINTF(("exit_lwps: Made %d.%d runnable\n",
- p->p_pid, l2->l_lid));
continue;
}
lwp_unlock(l2);
}
+
+ /*
+ * Wait for every LWP to exit. Note: LWPs can get suspended/slept
+ * behind us or there may even be new LWPs created. Therefore, a
+ * full retry is required on error.
+ */
while (p->p_nlwps > 1) {
- DPRINTF(("exit_lwps: waiting for %d LWPs (%d zombies)\n",
- p->p_nlwps, p->p_nzlwps));
- error = lwp_wait1(l, 0, &waited, LWPWAIT_EXITCONTROL);
- if (p->p_nlwps == 1)
- break;
- if (error == EDEADLK) {
- /*
- * LWPs can get suspended/slept behind us.
- * (eg. sa_setwoken)
- * kick them again and retry.
- */
+ if (lwp_wait(l, 0, NULL, true)) {
goto retry;
}
- if (error)
- panic("exit_lwps: lwp_wait1 failed with error %d",
- error);
- DPRINTF(("exit_lwps: Got LWP %d from lwp_wait1()\n", waited));
}
KERNEL_LOCK(nlocks, l);
Index: src/sys/kern/kern_lwp.c
diff -u src/sys/kern/kern_lwp.c:1.166 src/sys/kern/kern_lwp.c:1.166.2.1
--- src/sys/kern/kern_lwp.c:1.166 Sat Feb 11 23:16:17 2012
+++ src/sys/kern/kern_lwp.c Mon Oct 1 23:07:08 2012
@@ -1,4 +1,4 @@
-/* $NetBSD: kern_lwp.c,v 1.166 2012/02/11 23:16:17 martin Exp $ */
+/* $NetBSD: kern_lwp.c,v 1.166.2.1 2012/10/01 23:07:08 riz Exp $ */
/*-
* Copyright (c) 2001, 2006, 2007, 2008, 2009 The NetBSD Foundation, Inc.
@@ -211,7 +211,7 @@
*/
#include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: kern_lwp.c,v 1.166 2012/02/11 23:16:17 martin Exp $");
+__KERNEL_RCSID(0, "$NetBSD: kern_lwp.c,v 1.166.2.1 2012/10/01 23:07:08 riz Exp $");
#include "opt_ddb.h"
#include "opt_lockdebug.h"
@@ -489,22 +489,21 @@ lwp_unstop(struct lwp *l)
* Must be called with p->p_lock held.
*/
int
-lwp_wait1(struct lwp *l, lwpid_t lid, lwpid_t *departed, int flags)
+lwp_wait(struct lwp *l, lwpid_t lid, lwpid_t *departed, bool exiting)
{
- struct proc *p = l->l_proc;
- struct lwp *l2;
- int nfound, error;
- lwpid_t curlid;
- bool exiting;
+ const lwpid_t curlid = l->l_lid;
+ proc_t *p = l->l_proc;
+ lwp_t *l2;
+ int error;
KASSERT(mutex_owned(p->p_lock));
p->p_nlwpwait++;
l->l_waitingfor = lid;
- curlid = l->l_lid;
- exiting = ((flags & LWPWAIT_EXITCONTROL) != 0);
for (;;) {
+ int nfound;
+
/*
* Avoid a race between exit1() and sigexit(): if the
* process is dumping core, then we need to bail out: call
@@ -514,10 +513,7 @@ lwp_wait1(struct lwp *l, lwpid_t lid, lw
if ((p->p_sflag & PS_WCORE) != 0) {
mutex_exit(p->p_lock);
lwp_userret(l);
-#ifdef DIAGNOSTIC
- panic("lwp_wait1");
-#endif
- /* NOTREACHED */
+ KASSERT(false);
}
/*
@@ -607,13 +603,14 @@ lwp_wait1(struct lwp *l, lwpid_t lid, lw
}
/*
- * The kernel is careful to ensure that it can not deadlock
- * when exiting - just keep waiting.
+ * Note: since the lock will be dropped, need to restart on
+ * wakeup to run all LWPs again, e.g. there may be new LWPs.
*/
if (exiting) {
KASSERT(p->p_nlwps > 1);
cv_wait(&p->p_lwpcv, p->p_lock);
- continue;
+ error = EAGAIN;
+ break;
}
/*
Index: src/sys/kern/sys_lwp.c
diff -u src/sys/kern/sys_lwp.c:1.52.14.1 src/sys/kern/sys_lwp.c:1.52.14.2
--- src/sys/kern/sys_lwp.c:1.52.14.1 Mon May 21 15:25:56 2012
+++ src/sys/kern/sys_lwp.c Mon Oct 1 23:07:07 2012
@@ -1,4 +1,4 @@
-/* $NetBSD: sys_lwp.c,v 1.52.14.1 2012/05/21 15:25:56 riz Exp $ */
+/* $NetBSD: sys_lwp.c,v 1.52.14.2 2012/10/01 23:07:07 riz Exp $ */
/*-
* Copyright (c) 2001, 2006, 2007, 2008 The NetBSD Foundation, Inc.
@@ -35,7 +35,7 @@
*/
#include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: sys_lwp.c,v 1.52.14.1 2012/05/21 15:25:56 riz Exp $");
+__KERNEL_RCSID(0, "$NetBSD: sys_lwp.c,v 1.52.14.2 2012/10/01 23:07:07 riz Exp $");
#include <sys/param.h>
#include <sys/systm.h>
@@ -373,19 +373,14 @@ sys__lwp_wait(struct lwp *l, const struc
lwpid_t dep;
mutex_enter(p->p_lock);
- error = lwp_wait1(l, SCARG(uap, wait_for), &dep, 0);
+ error = lwp_wait(l, SCARG(uap, wait_for), &dep, false);
mutex_exit(p->p_lock);
- if (error)
- return error;
-
- if (SCARG(uap, departed)) {
+ if (!error && SCARG(uap, departed)) {
error = copyout(&dep, SCARG(uap, departed), sizeof(dep));
- if (error)
- return error;
}
- return 0;
+ return error;
}
int
Index: src/sys/sys/lwp.h
diff -u src/sys/sys/lwp.h:1.159.2.1 src/sys/sys/lwp.h:1.159.2.2
--- src/sys/sys/lwp.h:1.159.2.1 Mon May 21 15:25:56 2012
+++ src/sys/sys/lwp.h Mon Oct 1 23:07:08 2012
@@ -1,4 +1,4 @@
-/* $NetBSD: lwp.h,v 1.159.2.1 2012/05/21 15:25:56 riz Exp $ */
+/* $NetBSD: lwp.h,v 1.159.2.2 2012/10/01 23:07:08 riz Exp $ */
/*-
* Copyright (c) 2001, 2006, 2007, 2008, 2009, 2010
@@ -320,9 +320,7 @@ void lwp_drainrefs(lwp_t *);
bool lwp_alive(lwp_t *);
lwp_t *lwp_find_first(proc_t *);
-/* Flags for _lwp_wait1 */
-#define LWPWAIT_EXITCONTROL 0x00000001
-int lwp_wait1(lwp_t *, lwpid_t, lwpid_t *, int);
+int lwp_wait(lwp_t *, lwpid_t, lwpid_t *, bool);
void lwp_continue(lwp_t *);
void lwp_unsleep(lwp_t *, bool);
void lwp_unstop(lwp_t *);
