* [PATCH] futex: Drop now unnecessary check in exit_pi_state()
@ 2017-11-03 22:30 Julia Cartwright
2017-11-06 10:23 ` Peter Zijlstra
0 siblings, 1 reply; 2+ messages in thread
From: Julia Cartwright @ 2017-11-03 22:30 UTC (permalink / raw)
To: Peter Zijlstra, Ingo Molnar, Thomas Gleixner
Cc: linux-kernel, Gratian Crisan, Darren Hart
This check was an attempt to protect against a race with put_pi_state()
by ensuring that the pi_state_list was consistent across the
unlock/lock of pi_lock.
However, as of commit 153fbd1226fb3 ("futex: Fix more put_pi_state() vs.
exit_pi_state_list() races"), this check is no longer necessary because
we now hold a reference to the pi_state object across the unlock/lock of
pi_lock. This reference guarantees that a put_pi_state() on another CPU
won't rip the pi_state object from the list when we drop pi_lock.
Cc: Gratian Crisan <gratian.crisan@ni.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Darren Hart <dvhart@infradead.org>
Signed-off-by: Julia Cartwright <julia@ni.com>
---
I'm not sure my analysis is 100% correct here, so please carefully think
through it, as I'm sure you all always do when futex patches hit your
mailbox :).
Julia
kernel/futex.c | 16 ++--------------
1 file changed, 2 insertions(+), 14 deletions(-)
diff --git a/kernel/futex.c b/kernel/futex.c
index ca5bb9cba5cf..e127ec0555b6 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -889,7 +889,7 @@ static struct task_struct *futex_find_get_task(pid_t pid)
*/
void exit_pi_state_list(struct task_struct *curr)
{
- struct list_head *next, *head = &curr->pi_state_list;
+ struct list_head *head = &curr->pi_state_list;
struct futex_pi_state *pi_state;
struct futex_hash_bucket *hb;
union futex_key key = FUTEX_KEY_INIT;
@@ -903,8 +903,7 @@ void exit_pi_state_list(struct task_struct *curr)
*/
raw_spin_lock_irq(&curr->pi_lock);
while (!list_empty(head)) {
- next = head->next;
- pi_state = list_entry(next, struct futex_pi_state, list);
+ pi_state = list_first_entry(head, struct futex_pi_state, list);
key = pi_state->key;
hb = hash_futex(&key);
@@ -929,17 +928,6 @@ void exit_pi_state_list(struct task_struct *curr)
spin_lock(&hb->lock);
raw_spin_lock_irq(&pi_state->pi_mutex.wait_lock);
raw_spin_lock(&curr->pi_lock);
- /*
- * We dropped the pi-lock, so re-check whether this
- * task still owns the PI-state:
- */
- if (head->next != next) {
- /* retain curr->pi_lock for the loop invariant */
- raw_spin_unlock(&pi_state->pi_mutex.wait_lock);
- spin_unlock(&hb->lock);
- put_pi_state(pi_state);
- continue;
- }
WARN_ON(pi_state->owner != curr);
WARN_ON(list_empty(&pi_state->list));
--
2.14.2
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH] futex: Drop now unnecessary check in exit_pi_state()
2017-11-03 22:30 [PATCH] futex: Drop now unnecessary check in exit_pi_state() Julia Cartwright
@ 2017-11-06 10:23 ` Peter Zijlstra
0 siblings, 0 replies; 2+ messages in thread
From: Peter Zijlstra @ 2017-11-06 10:23 UTC (permalink / raw)
To: Julia Cartwright
Cc: Ingo Molnar, Thomas Gleixner, linux-kernel, Gratian Crisan, Darren Hart
On Fri, Nov 03, 2017 at 05:30:28PM -0500, Julia Cartwright wrote:
> This check was an attempt to protect against a race with put_pi_state()
> by ensuring that the pi_state_list was consistent across the
> unlock/lock of pi_lock.
>
> However, as of commit 153fbd1226fb3 ("futex: Fix more put_pi_state() vs.
> exit_pi_state_list() races"), this check is no longer necessary because
> we now hold a reference to the pi_state object across the unlock/lock of
> pi_lock. This reference guarantees that a put_pi_state() on another CPU
> won't rip the pi_state object from the list when we drop pi_lock.
> @@ -929,17 +928,6 @@ void exit_pi_state_list(struct task_struct *curr)
> spin_lock(&hb->lock);
> raw_spin_lock_irq(&pi_state->pi_mutex.wait_lock);
> raw_spin_lock(&curr->pi_lock);
> - /*
> - * We dropped the pi-lock, so re-check whether this
> - * task still owns the PI-state:
> - */
> - if (head->next != next) {
Quite possibly you're right, but I would sleep a whole lot better if
that were to remain a WARN. You never know with this futex stuff :-)
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2017-11-06 10:23 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-11-03 22:30 [PATCH] futex: Drop now unnecessary check in exit_pi_state() Julia Cartwright
2017-11-06 10:23 ` Peter Zijlstra
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).