[PATCH] sched/core: Fix a potential double fetch bug on attr->size

[PATCH] sched/core: Fix a potential double fetch bug on attr->size

[Meng Xu]

From: Meng Xu <mengxu.gatech@gmail.com>

`attr->size` after the second fetch `copy_from_user(attr, uattr, size)`,
can be different from what is initially fetched in and checked
`get_user(size, &uattr->size)` by racing condition in the userspace.

The issue and the patch are both similar to commit f12f42a
(in kernel/events/core.c).

Signed-off-by: Meng Xu <mengxu.gatech@gmail.com>
---
 kernel/sched/core.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 0869b20..c22d2b4 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4349,6 +4349,8 @@ static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *a
 	if (ret)
 		return -EFAULT;
 
+	attr->size = size;
+
 	/*
 	 * XXX: Do we want to be lenient like existing syscalls; or do we want
 	 * to be strict and return an error on out-of-bounds values?
-- 
2.7.4

[PATCH] sched/core: Fix a potential double fetch bug on attr->size

[Meng Xu]

From: Meng Xu <mengxu.gatech@gmail.com>

`attr->size` after the second fetch `copy_from_user(attr, uattr, size)`,
can be different from what is initially fetched in and checked
`get_user(size, &uattr->size)` by racing condition in the userspace.

The issue and the patch are both similar to commit f12f42a
(in kernel/events/core.c).

Signed-off-by: Meng Xu <mengxu.gatech@gmail.com>
---
 kernel/sched/core.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 0869b20..c22d2b4 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4349,6 +4349,8 @@ static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *a
 	if (ret)
 		return -EFAULT;
 
+	attr->size = size;
+
 	/*
 	 * XXX: Do we want to be lenient like existing syscalls; or do we want
 	 * to be strict and return an error on out-of-bounds values?
-- 
2.7.4

Re: [PATCH] sched/core: Fix a potential double fetch bug on attr->size

[Peter Zijlstra]

On Sat, Sep 23, 2017 at 10:05:56PM -0400, Meng Xu wrote:
> Hi Peter and Ingo,
> 
> As a reminder, this is a very similar issue to perf_copy_attr (see following patch)
> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?id=f12f42acdbb577a12eecfcebbbec41c81505c4dc <https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?id=f12f42acdbb577a12eecfcebbbec41c81505c4dc>
> 
> Will it be fixed as well?

If someone were to actually send me the patch.. probably.

> > On Aug 29, 2017, at 3:10 PM, Meng Xu <meng.xu@gatech.edu> wrote:
> > 
> > From: Meng Xu <mengxu.gatech@gmail.com>
> > 
> > `attr->size` after the second fetch `copy_from_user(attr, uattr, size)`,
> > can be different from what is initially fetched in and checked
> > `get_user(size, &uattr->size)` by racing condition in the userspace.
> > 
> > The issue and the patch are both similar to commit f12f42a
> > (in kernel/events/core.c).
> > 
> > Signed-off-by: Meng Xu <mengxu.gatech@gmail.com>
> > ---
> > kernel/sched/core.c | 2 ++
> > 1 file changed, 2 insertions(+)
> > 
> > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > index 0869b20..c22d2b4 100644
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -4349,6 +4349,8 @@ static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *a
> > 	if (ret)
> > 		return -EFAULT;
> > 
> > +	attr->size = size;
> > +
> > 	/*
> > 	 * XXX: Do we want to be lenient like existing syscalls; or do we want
> > 	 * to be strict and return an error on out-of-bounds values?
> > -- 
> > 2.7.4
> > 
> 

Re: [PATCH] sched/core: Fix a potential double fetch bug on attr->size

[Meng Xu]

Hi Peter,

I am sorry, I thought the patch is included in the forwarded
email. I just resent the patch. Please check.

Best Regards,
Meng

On 09/25/2017 03:31 AM, Peter Zijlstra wrote:
> On Sat, Sep 23, 2017 at 10:05:56PM -0400, Meng Xu wrote:
>> Hi Peter and Ingo,
>>
>> As a reminder, this is a very similar issue to perf_copy_attr (see following patch)
>> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?id=f12f42acdbb577a12eecfcebbbec41c81505c4dc <https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?id=f12f42acdbb577a12eecfcebbbec41c81505c4dc>
>>
>> Will it be fixed as well?
> If someone were to actually send me the patch.. probably.
>
>>> On Aug 29, 2017, at 3:10 PM, Meng Xu <meng.xu@gatech.edu> wrote:
>>>
>>> From: Meng Xu <mengxu.gatech@gmail.com>
>>>
>>> `attr->size` after the second fetch `copy_from_user(attr, uattr, size)`,
>>> can be different from what is initially fetched in and checked
>>> `get_user(size, &uattr->size)` by racing condition in the userspace.
>>>
>>> The issue and the patch are both similar to commit f12f42a
>>> (in kernel/events/core.c).
>>>
>>> Signed-off-by: Meng Xu <mengxu.gatech@gmail.com>
>>> ---
>>> kernel/sched/core.c | 2 ++
>>> 1 file changed, 2 insertions(+)
>>>
>>> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
>>> index 0869b20..c22d2b4 100644
>>> --- a/kernel/sched/core.c
>>> +++ b/kernel/sched/core.c
>>> @@ -4349,6 +4349,8 @@ static int sched_copy_attr(struct sched_attr __user *uattr, struct sched_attr *a
>>> 	if (ret)
>>> 		return -EFAULT;
>>>
>>> +	attr->size = size;
>>> +
>>> 	/*
>>> 	 * XXX: Do we want to be lenient like existing syscalls; or do we want
>>> 	 * to be strict and return an error on out-of-bounds values?
>>> -- 
>>> 2.7.4
>>>