lkml.org 
[lkml]   [2010]   [Mar]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
Subject[PATCH -mm 2/4] cputimers: make sure thread_group_cputime() can't count the same thread twice lockless
- change __exit_signal() to do __unhash_process() before we accumulate
the counters in ->signal

- add a couple of barriers into thread_group_cputime() and __exit_signal()
to make sure thread_group_cputime() can never account the same thread
twice if it races with exit.

If any thread T was already accounted in ->signal, next_thread() or
pid_alive() must see the result of __unhash_process(T).

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---

kernel/exit.c | 14 +++++++++-----
kernel/posix-cpu-timers.c | 6 ++++++
2 files changed, 15 insertions(+), 5 deletions(-)

--- 34-rc1/kernel/exit.c~cpuacct_2_thread_group_cputime_rcu_safe 2010-03-29 18:03:17.000000000 +0200
+++ 34-rc1/kernel/exit.c 2010-03-29 18:29:35.000000000 +0200
@@ -88,6 +88,8 @@ static void __exit_signal(struct task_st
rcu_read_lock_held() ||
lockdep_is_held(&tasklist_lock));
spin_lock(&sighand->siglock);
+ __unhash_process(tsk, group_dead);
+ sig->nr_threads--;

posix_cpu_timers_exit(tsk);
if (group_dead) {
@@ -111,9 +113,14 @@ static void __exit_signal(struct task_st
* The group leader stays around as a zombie as long
* as there are other threads. When it gets reaped,
* the exit.c code will add its counts into these totals.
- * We won't ever get here for the group leader, since it
- * will have been the last reference on the signal_struct.
+ *
+ * Make sure that this thread can't be accounted twice
+ * by thread_group_cputime() under rcu. If it sees the
+ * the result of accounting below it must see the result
+ * of __unhash_process()->__list_del(thread_group) above.
*/
+ smp_wmb();
+
sig->utime = cputime_add(sig->utime, tsk->utime);
sig->stime = cputime_add(sig->stime, tsk->stime);
sig->gtime = cputime_add(sig->gtime, tsk->gtime);
@@ -127,9 +134,6 @@ static void __exit_signal(struct task_st
sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
}

- sig->nr_threads--;
- __unhash_process(tsk, group_dead);
-
/*
* Do this under ->siglock, we can race with another thread
* doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
--- 34-rc1/kernel/posix-cpu-timers.c~cpuacct_2_thread_group_cputime_rcu_safe 2010-03-29 18:09:15.000000000 +0200
+++ 34-rc1/kernel/posix-cpu-timers.c 2010-03-29 18:29:35.000000000 +0200
@@ -239,6 +239,12 @@ void thread_group_cputime(struct task_st
times->utime = sig->utime;
times->stime = sig->stime;
times->sum_exec_runtime = sig->sum_sched_runtime;
+ /*
+ * This pairs with wmb() in __exit_signal(). If any thread was
+ * already accounted in tsk->signal, while_each_thread() must
+ * not see it.
+ */
+ smp_rmb();

rcu_read_lock();
/* make sure we can trust tsk->thread_group list */


\
 
 \ /
  Last update: 2010-03-29 20:19    [W:0.114 / U:0.420 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site