Dave Jones reported a bug with futex_lock_pi() that his trinity test
exposed. Sometime between queue_me() and taking the q.lock_ptr, the
lock_ptr became NULL, resulting in a crash.
While futex_wake() is careful to not call wake_futex() on futex_q's with
a pi_state or an rt_waiter (which are either waiting for a
futex_unlock_pi() or a PI futex_requeue()), futex_wake_op() and
futex_requeue() do not perform the same test.
Update futex_wake_op() and futex_requeue() to test for q.pi_state and
q.rt_waiter and abort with -EINVAL if detected. To ensure any future
breakage is caught, add a WARN() to wake_futex() if the same condition
is true.
This fix has seen 3 hours of testing with "trinity -c futex" on an
x86_64 VM with 4 CPUS.
Signed-off-by: Darren Hart <[email protected]>
Reported-by: Dave Jones <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: John Kacur <[email protected]>
Cc: [email protected]
---
kernel/futex.c | 20 +++++++++++++++++++-
1 file changed, 19 insertions(+), 1 deletion(-)
diff --git a/kernel/futex.c b/kernel/futex.c
index 3717e7b..5699b21 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -840,6 +840,11 @@ static void wake_futex(struct futex_q *q)
{
struct task_struct *p = q->task;
+ if (q->pi_state || q->rt_waiter) {
+ WARN(1, "%s: refusing to wake PI futex\n", __FUNCTION__);
+ return;
+ }
+
/*
* We set q->lock_ptr = NULL _before_ we wake up the task. If
* a non-futex wake up happens on another CPU then the task
@@ -1075,6 +1080,10 @@ retry_private:
plist_for_each_entry_safe(this, next, head, list) {
if (match_futex (&this->key, &key1)) {
+ if (this->pi_state || this->rt_waiter) {
+ ret = -EINVAL;
+ goto out_unlock;
+ }
wake_futex(this);
if (++ret >= nr_wake)
break;
@@ -1087,6 +1096,10 @@ retry_private:
op_ret = 0;
plist_for_each_entry_safe(this, next, head, list) {
if (match_futex (&this->key, &key2)) {
+ if (this->pi_state || this->rt_waiter) {
+ ret = -EINVAL;
+ goto out_unlock;
+ }
wake_futex(this);
if (++op_ret >= nr_wake2)
break;
@@ -1095,6 +1108,7 @@ retry_private:
ret += op_ret;
}
+out_unlock:
double_unlock_hb(hb1, hb2);
out_put_keys:
put_futex_key(&key2);
@@ -1384,9 +1398,13 @@ retry_private:
/*
* FUTEX_WAIT_REQEUE_PI and FUTEX_CMP_REQUEUE_PI should always
* be paired with each other and no other futex ops.
+ *
+ * We should never be requeueing a futex_q with a pi_state,
+ * which is awaiting a futex_unlock_pi().
*/
if ((requeue_pi && !this->rt_waiter) ||
- (!requeue_pi && this->rt_waiter)) {
+ (!requeue_pi && this->rt_waiter) ||
+ this->pi_state) {
ret = -EINVAL;
break;
}
--
1.7.11.7
Commit-ID: 0e8f7a5954be13d0c8dcbca3204a9e962498c46e
Gitweb: http://git.kernel.org/tip/0e8f7a5954be13d0c8dcbca3204a9e962498c46e
Author: Darren Hart <[email protected]>
AuthorDate: Tue, 20 Nov 2012 23:36:45 -0800
Committer: Thomas Gleixner <[email protected]>
CommitDate: Wed, 21 Nov 2012 21:05:34 +0100
futex: Avoid wake_futex for a PI futex_q
Dave Jones reported a bug with futex_lock_pi() that his trinity test
exposed. Sometime between queue_me() and taking the q.lock_ptr, the
lock_ptr became NULL, resulting in a crash.
While futex_wake() is careful to not call wake_futex() on futex_q's with
a pi_state or an rt_waiter (which are either waiting for a
futex_unlock_pi() or a PI futex_requeue()), futex_wake_op() and
futex_requeue() do not perform the same test.
Update futex_wake_op() and futex_requeue() to test for q.pi_state and
q.rt_waiter and abort with -EINVAL if detected. To ensure any future
breakage is caught, add a WARN() to wake_futex() if the same condition
is true.
This fix has seen 3 hours of testing with "trinity -c futex" on an
x86_64 VM with 4 CPUS.
Reported-by: Dave Jones <[email protected]>
Signed-off-by: Darren Hart <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: John Kacur <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/3b25c8ba053760892871713ff6e81660433f6734.1353483196.git.dvhart@linux.intel.com
Signed-off-by: Thomas Gleixner <[email protected]>
---
kernel/futex.c | 20 +++++++++++++++++++-
1 file changed, 19 insertions(+), 1 deletion(-)
diff --git a/kernel/futex.c b/kernel/futex.c
index 3717e7b..5699b21 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -840,6 +840,11 @@ static void wake_futex(struct futex_q *q)
{
struct task_struct *p = q->task;
+ if (q->pi_state || q->rt_waiter) {
+ WARN(1, "%s: refusing to wake PI futex\n", __FUNCTION__);
+ return;
+ }
+
/*
* We set q->lock_ptr = NULL _before_ we wake up the task. If
* a non-futex wake up happens on another CPU then the task
@@ -1075,6 +1080,10 @@ retry_private:
plist_for_each_entry_safe(this, next, head, list) {
if (match_futex (&this->key, &key1)) {
+ if (this->pi_state || this->rt_waiter) {
+ ret = -EINVAL;
+ goto out_unlock;
+ }
wake_futex(this);
if (++ret >= nr_wake)
break;
@@ -1087,6 +1096,10 @@ retry_private:
op_ret = 0;
plist_for_each_entry_safe(this, next, head, list) {
if (match_futex (&this->key, &key2)) {
+ if (this->pi_state || this->rt_waiter) {
+ ret = -EINVAL;
+ goto out_unlock;
+ }
wake_futex(this);
if (++op_ret >= nr_wake2)
break;
@@ -1095,6 +1108,7 @@ retry_private:
ret += op_ret;
}
+out_unlock:
double_unlock_hb(hb1, hb2);
out_put_keys:
put_futex_key(&key2);
@@ -1384,9 +1398,13 @@ retry_private:
/*
* FUTEX_WAIT_REQEUE_PI and FUTEX_CMP_REQUEUE_PI should always
* be paired with each other and no other futex ops.
+ *
+ * We should never be requeueing a futex_q with a pi_state,
+ * which is awaiting a futex_unlock_pi().
*/
if ((requeue_pi && !this->rt_waiter) ||
- (!requeue_pi && this->rt_waiter)) {
+ (!requeue_pi && this->rt_waiter) ||
+ this->pi_state) {
ret = -EINVAL;
break;
}
On 11/21/2012 02:34 PM, Andrew Morton wrote:
> On Tue, 20 Nov 2012 23:36:45 -0800
> Darren Hart <[email protected]> wrote:
>
>> Dave Jones reported a bug with futex_lock_pi() that his trinity test
>> exposed. Sometime between queue_me() and taking the q.lock_ptr, the
>> lock_ptr became NULL, resulting in a crash.
>>
>> While futex_wake() is careful to not call wake_futex() on futex_q's with
>> a pi_state or an rt_waiter (which are either waiting for a
>> futex_unlock_pi() or a PI futex_requeue()), futex_wake_op() and
>> futex_requeue() do not perform the same test.
>>
>> Update futex_wake_op() and futex_requeue() to test for q.pi_state and
>> q.rt_waiter and abort with -EINVAL if detected. To ensure any future
>> breakage is caught, add a WARN() to wake_futex() if the same condition
>> is true.
>>
>> This fix has seen 3 hours of testing with "trinity -c futex" on an
>> x86_64 VM with 4 CPUS.
>>
>> ...
>>
>> --- a/kernel/futex.c
>> +++ b/kernel/futex.c
>> @@ -840,6 +840,11 @@ static void wake_futex(struct futex_q *q)
>> {
>> struct task_struct *p = q->task;
>>
>> + if (q->pi_state || q->rt_waiter) {
>> + WARN(1, "%s: refusing to wake PI futex\n", __FUNCTION__);
>> + return;
>> + }
>
> There's no need to display __FUNCTION__ because WARN() gives a
> backtrace, and we can more neatly use the WARN() return value:
>
> --- a/kernel/futex.c~futex-avoid-wake_futex-for-a-pi-futex_q-fix
> +++ a/kernel/futex.c
> @@ -843,10 +843,8 @@ static void wake_futex(struct futex_q *q
> {
> struct task_struct *p = q->task;
>
> - if (q->pi_state || q->rt_waiter) {
> - WARN(1, "%s: refusing to wake PI futex\n", __FUNCTION__);
> + if (WARN(q->pi_state || q->rt_waiter, "refusing to wake PI futex\n"))
> return;
> - }
Thanks, that's better. Duh. That block of code used to be a lot more
complex during my debug sessions, I should have caught that and boiled
it down one step further.
--
Darren
>
> /*
> * We set q->lock_ptr = NULL _before_ we wake up the task. If
> _
>
>>
>> ...
>>
--
Darren Hart
Intel Open Source Technology Center
Yocto Project - Technical Lead - Linux Kernel