lkml.org 
[lkml]   [2016]   [Nov]   [14]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
Subject[PATCH 2/2] sched/autogroup: a zombie thread must not use autogroup->tg
Exactly because for_each_thread() in autogroup_move_group() can't see it
and update its ->sched_task_group before _put() and possibly free().

So the exiting task needs another sched_move_task() before exit_notify()
and we need to re-introduce the PF_EXITING (or similar) check removed by
the previous change for another reason.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: stable@vger.kernel.org
---
include/linux/sched.h | 2 ++
kernel/exit.c | 1 +
kernel/sched/auto_group.c | 19 +++++++++++++++++++
3 files changed, 22 insertions(+)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 348f51b..e9c009d 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -2567,6 +2567,7 @@ extern void sched_autogroup_create_attach(struct task_struct *p);
extern void sched_autogroup_detach(struct task_struct *p);
extern void sched_autogroup_fork(struct signal_struct *sig);
extern void sched_autogroup_exit(struct signal_struct *sig);
+extern void sched_autogroup_exit_task(struct task_struct *p);
#ifdef CONFIG_PROC_FS
extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
@@ -2576,6 +2577,7 @@ static inline void sched_autogroup_create_attach(struct task_struct *p) { }
static inline void sched_autogroup_detach(struct task_struct *p) { }
static inline void sched_autogroup_fork(struct signal_struct *sig) { }
static inline void sched_autogroup_exit(struct signal_struct *sig) { }
+static inline void sched_autogroup_exit_task(struct task_struct *p) { }
#endif

extern int yield_to(struct task_struct *p, bool preempt);
diff --git a/kernel/exit.c b/kernel/exit.c
index f3dd46d..76e263e 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -837,6 +837,7 @@ void __noreturn do_exit(long code)
*/
perf_event_exit_task(tsk);

+ sched_autogroup_exit_task(tsk);
cgroup_exit(tsk);

/*
diff --git a/kernel/sched/auto_group.c b/kernel/sched/auto_group.c
index ad2b19a..f1c8fd5 100644
--- a/kernel/sched/auto_group.c
+++ b/kernel/sched/auto_group.c
@@ -115,10 +115,26 @@ bool task_wants_autogroup(struct task_struct *p, struct task_group *tg)
* If we race with autogroup_move_group() the caller can use the old
* value of signal->autogroup but in this case sched_move_task() will
* be called again before autogroup_kref_put().
+ *
+ * However, there is no way sched_autogroup_exit_task() could tell us
+ * to avoid autogroup->tg, so we abuse PF_EXITING flag for this case.
*/
+ if (p->flags & PF_EXITING)
+ return false;
+
return true;
}

+void sched_autogroup_exit_task(struct task_struct *p)
+{
+ /*
+ * We are going to call exit_notify() and autogroup_move_group() can't
+ * see this thread after that: we can no longer use signal->autogroup.
+ * See the PF_EXITING check in task_wants_autogroup().
+ */
+ sched_move_task(p);
+}
+
static void
autogroup_move_group(struct task_struct *p, struct autogroup *ag)
{
@@ -142,6 +158,9 @@ autogroup_move_group(struct task_struct *p, struct autogroup *ag)
* In the latter case for_each_thread() can not miss a migrating thread,
* cpu_cgroup_attach() must not be possible after cgroup_exit() and it
* can't be removed from thread list, we hold ->siglock.
+ *
+ * If an exiting thread was already removed from thread list we rely on
+ * sched_autogroup_exit_task().
*/
for_each_thread(p, t)
sched_move_task(t);
--
2.5.0
\
 
 \ /
  Last update: 2016-11-14 19:46    [W:0.098 / U:0.168 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site