2020-05-24 09:19:42

by Zhang, Qiang

[permalink] [raw]
Subject: [PATCH] workqueue: Fix double kfree(rescuer) in destroy_workqueue()

From: Zhang Qiang <[email protected]>

When destroy_workqueue if rescuer worker exist,wq->rescuer pointer be
kfree. if sanity checks passed. the func call_rcu(&wq->rcu, rcu_free_wq)
will be called if the wq->flags & WQ_UNBOUND is false,in rcu_free_wq
func wq->rescuer pointer was kfree again.

Signed-off-by: Zhang Qiang <[email protected]>
---
kernel/workqueue.c | 1 -
1 file changed, 1 deletion(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 891ccad5f271..a2451cdcd503 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -3491,7 +3491,6 @@ static void rcu_free_wq(struct rcu_head *rcu)
else
free_workqueue_attrs(wq->unbound_attrs);

- kfree(wq->rescuer);
kfree(wq);
}

--
2.17.0


2020-05-24 15:36:17

by Markus Elfring

[permalink] [raw]
Subject: Re: [PATCH] workqueue: Fix double kfree(rescuer) in destroy_workqueue()

> When destroy_workqueue if rescuer worker exist,wq->rescuer pointer be
> kfree. if sanity checks passed. the func call_rcu(&wq->rcu, rcu_free_wq)
> will be called if the wq->flags & WQ_UNBOUND is false,in rcu_free_wq
> func wq->rescuer pointer was kfree again.

1. I suggest to improve also this change description.
Do you try to explain here that a call of the function “free_workqueue_attrs”
(or “free_percpu”) would perform sufficient clean-up of system resources
in this use case?

2. You proposed to delete the function call “kfree(wq->rescuer)” from
the implementation of the function “rcu_free_wq”.
https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/kernel/workqueue.c?id=c11d28ab4a691736e30b49813fb801847bd44e83#n3482
https://elixir.bootlin.com/linux/v5.7-rc6/source/kernel/workqueue.c#L3482

This function name should be specified also in the patch subject,
shouldn't it?

3. Would you like to add the tag “Fixes” to the commit message?

Regards,
Markus

2020-05-25 06:31:02

by Zhang, Qiang

[permalink] [raw]
Subject: Re: [PATCH] workqueue: Fix double kfree(rescuer) in destroy_workqueue()

Sorry I didn't describe clearly

I describe the meaning as follows:

destroy_workqueue
if(wq->rescuer)
struct worker *rescuer = wq->rescuer
kfree(rescuer)
..................
if (!(wq->flags & WQ_UNBOUND))
call_rcu(&wq->rcu, rcu_free_wq)
......................
rcu_free_wq
kfree(wq->rescuer)

there are double free.

On 5/24/20 11:33 PM, Markus Elfring wrote:
>> When destroy_workqueue if rescuer worker exist,wq->rescuer pointer be
>> kfree. if sanity checks passed. the func call_rcu(&wq->rcu, rcu_free_wq)
>> will be called if the wq->flags & WQ_UNBOUND is false,in rcu_free_wq
>> func wq->rescuer pointer was kfree again.
>
> 1. I suggest to improve also this change description.
> Do you try to explain here that a call of the function “free_workqueue_attrs”
> (or “free_percpu”) would perform sufficient clean-up of system resources
> in this use case?
>
> 2. You proposed to delete the function call “kfree(wq->rescuer)” from
> the implementation of the function “rcu_free_wq”.
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/kernel/workqueue.c?id=c11d28ab4a691736e30b49813fb801847bd44e83#n3482
> https://elixir.bootlin.com/linux/v5.7-rc6/source/kernel/workqueue.c#L3482
>
> This function name should be specified also in the patch subject,
> shouldn't it?
>
> 3. Would you like to add the tag “Fixes” to the commit message?
>
> Regards,
> Markus
>

2020-05-25 06:48:05

by Zhang, Qiang

[permalink] [raw]
Subject: Re: [PATCH] workqueue: Fix double kfree(rescuer) in destroy_workqueue()

Sorry I didn't describe clearly

I describe the meaning as follows:

destroy_workqueue:
if(wq->rescuer)
struct worker *rescuer = wq->rescuer
kfree(rescuer) //first kfree


if (!(wq->flags & WQ_UNBOUND))
call_rcu(&wq->rcu, rcu_free_wq)

rcu_free_wq
kfree(wq->rescuer) //second kfree

there are double free.

On 5/24/20 11:33 PM, Markus Elfring wrote:
>> When destroy_workqueue if rescuer worker exist,wq->rescuer pointer be
>> kfree. if sanity checks passed. the func call_rcu(&wq->rcu, rcu_free_wq)
>> will be called if the wq->flags & WQ_UNBOUND is false,in rcu_free_wq
>> func wq->rescuer pointer was kfree again.
>
> 1. I suggest to improve also this change description.
> Do you try to explain here that a call of the function “free_workqueue_attrs”
> (or “free_percpu”) would perform sufficient clean-up of system resources
> in this use case?
>
> 2. You proposed to delete the function call “kfree(wq->rescuer)” from
> the implementation of the function “rcu_free_wq”.
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/kernel/workqueue.c?id=c11d28ab4a691736e30b49813fb801847bd44e83#n3482
> https://elixir.bootlin.com/linux/v5.7-rc6/source/kernel/workqueue.c#L3482
>
> This function name should be specified also in the patch subject,
> shouldn't it?
>
> 3. Would you like to add the tag “Fixes” to the commit message?
>
> Regards,
> Markus
>

2020-05-25 08:06:14

by Markus Elfring

[permalink] [raw]
Subject: Re: [PATCH] workqueue: Fix double kfree(rescuer) in destroy_workqueue()

> Sorry I didn't describe clearly
>
> I describe the meaning as follows:

Can it help to adjust the change description in the way
that a duplicate memory release should be deleted from the implementation
of the callback function “rcu_free_wq”?

Which commit should be referenced for the tag “Fixes”?

Regards,
Markus