LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* [PATCH] vboxguest: check for private_data before trying to close it.
@ 2019-05-28 12:47 Young Xiao
2019-05-28 13:19 ` Hans de Goede
0 siblings, 1 reply; 4+ messages in thread
From: Young Xiao @ 2019-05-28 12:47 UTC (permalink / raw)
To: hdegoede, arnd, gregkh, linux-kernel; +Cc: Young Xiao
vbg_misc_device_close doesn't check that filp->private_data is non-NULL
before trying to close_session, where vbg_core_close_session uses pointer
session whithout checking, too. That can cause an oops in certain error
conditions that can occur on open or lookup before the private_data is set.
This vulnerability is similar to CVE-2011-1771.
Signed-off-by: Young Xiao <92siuyang@gmail.com>
---
drivers/virt/vboxguest/vboxguest_linux.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/drivers/virt/vboxguest/vboxguest_linux.c b/drivers/virt/vboxguest/vboxguest_linux.c
index 6e8c0f1..b03c16f 100644
--- a/drivers/virt/vboxguest/vboxguest_linux.c
+++ b/drivers/virt/vboxguest/vboxguest_linux.c
@@ -88,8 +88,10 @@ static int vbg_misc_device_user_open(struct inode *inode, struct file *filp)
*/
static int vbg_misc_device_close(struct inode *inode, struct file *filp)
{
- vbg_core_close_session(filp->private_data);
- filp->private_data = NULL;
+ if (file->private_data != NULL) {
+ vbg_core_close_session(filp->private_data);
+ filp->private_data = NULL;
+ }
return 0;
}
--
2.7.4
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH] vboxguest: check for private_data before trying to close it.
2019-05-28 12:47 [PATCH] vboxguest: check for private_data before trying to close it Young Xiao
@ 2019-05-28 13:19 ` Hans de Goede
2019-05-28 14:49 ` Yang Xiao
0 siblings, 1 reply; 4+ messages in thread
From: Hans de Goede @ 2019-05-28 13:19 UTC (permalink / raw)
To: Young Xiao, arnd, gregkh, linux-kernel
Hi,
On 28-05-19 14:47, Young Xiao wrote:
> vbg_misc_device_close doesn't check that filp->private_data is non-NULL
> before trying to close_session, where vbg_core_close_session uses pointer
> session whithout checking, too. That can cause an oops in certain error
> conditions that can occur on open or lookup before the private_data is set.
>
> This vulnerability is similar to CVE-2011-1771.
How is this in anyway related to CVE-2011-1771 ????
That CVE is about a filesystems lookup method including a direct open
of the file for performance reasons and if that direct open fails, doing
a fput, which is how it ends up with a file with private_date being NULL,
see:
https://bugzilla.redhat.com/show_bug.cgi?id=703016
"the problem is that CIFS doesn't do O_DIRECT at all, so when you try to open a file with it you get back -EINVAL. CIFS can also do open on lookup in some cases. In that case, fput will be called on the filp, which has not yet had its private_data set."
The vboxguest code on the other hand is not a filessystem and as such
does not have a dentry lookup method at all!
It's close method is part of the file_operations struct for a character
device using the misc_device framework which guarantees that private_data
is always set.
Regards,
Hans
>
> Signed-off-by: Young Xiao <92siuyang@gmail.com>
> ---
> drivers/virt/vboxguest/vboxguest_linux.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/virt/vboxguest/vboxguest_linux.c b/drivers/virt/vboxguest/vboxguest_linux.c
> index 6e8c0f1..b03c16f 100644
> --- a/drivers/virt/vboxguest/vboxguest_linux.c
> +++ b/drivers/virt/vboxguest/vboxguest_linux.c
> @@ -88,8 +88,10 @@ static int vbg_misc_device_user_open(struct inode *inode, struct file *filp)
> */
> static int vbg_misc_device_close(struct inode *inode, struct file *filp)
> {
> - vbg_core_close_session(filp->private_data);
> - filp->private_data = NULL;
> + if (file->private_data != NULL) {
> + vbg_core_close_session(filp->private_data);
> + filp->private_data = NULL;
> + }
> return 0;
> }
>
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] vboxguest: check for private_data before trying to close it.
2019-05-28 13:19 ` Hans de Goede
@ 2019-05-28 14:49 ` Yang Xiao
2019-05-28 14:51 ` Hans de Goede
0 siblings, 1 reply; 4+ messages in thread
From: Yang Xiao @ 2019-05-28 14:49 UTC (permalink / raw)
To: Hans de Goede; +Cc: arnd, gregkh, linux-kernel
Based on your explanation, file->private_data can not be NULL before
call vbg_core_close_session method all the time, since
file->private_data is always set.
Am I right?
On Tue, May 28, 2019 at 9:19 PM Hans de Goede <hdegoede@redhat.com> wrote:
>
> Hi,
>
> On 28-05-19 14:47, Young Xiao wrote:
> > vbg_misc_device_close doesn't check that filp->private_data is non-NULL
> > before trying to close_session, where vbg_core_close_session uses pointer
> > session whithout checking, too. That can cause an oops in certain error
> > conditions that can occur on open or lookup before the private_data is set.
> >
> > This vulnerability is similar to CVE-2011-1771.
>
> How is this in anyway related to CVE-2011-1771 ????
>
> That CVE is about a filesystems lookup method including a direct open
> of the file for performance reasons and if that direct open fails, doing
> a fput, which is how it ends up with a file with private_date being NULL,
> see:
>
> https://bugzilla.redhat.com/show_bug.cgi?id=703016
>
> "the problem is that CIFS doesn't do O_DIRECT at all, so when you try to open a file with it you get back -EINVAL. CIFS can also do open on lookup in some cases. In that case, fput will be called on the filp, which has not yet had its private_data set."
>
> The vboxguest code on the other hand is not a filessystem and as such
> does not have a dentry lookup method at all!
>
> It's close method is part of the file_operations struct for a character
> device using the misc_device framework which guarantees that private_data
> is always set.
>
> Regards,
>
> Hans
>
>
>
>
>
> >
> > Signed-off-by: Young Xiao <92siuyang@gmail.com>
> > ---
> > drivers/virt/vboxguest/vboxguest_linux.c | 6 ++++--
> > 1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/virt/vboxguest/vboxguest_linux.c b/drivers/virt/vboxguest/vboxguest_linux.c
> > index 6e8c0f1..b03c16f 100644
> > --- a/drivers/virt/vboxguest/vboxguest_linux.c
> > +++ b/drivers/virt/vboxguest/vboxguest_linux.c
> > @@ -88,8 +88,10 @@ static int vbg_misc_device_user_open(struct inode *inode, struct file *filp)
> > */
> > static int vbg_misc_device_close(struct inode *inode, struct file *filp)
> > {
> > - vbg_core_close_session(filp->private_data);
> > - filp->private_data = NULL;
> > + if (file->private_data != NULL) {
> > + vbg_core_close_session(filp->private_data);
> > + filp->private_data = NULL;
> > + }
> > return 0;
> > }
> >
> >
--
Best regards!
Young
-----------------------------------------------------------
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] vboxguest: check for private_data before trying to close it.
2019-05-28 14:49 ` Yang Xiao
@ 2019-05-28 14:51 ` Hans de Goede
0 siblings, 0 replies; 4+ messages in thread
From: Hans de Goede @ 2019-05-28 14:51 UTC (permalink / raw)
To: Yang Xiao; +Cc: arnd, gregkh, linux-kernel
Hi,
On 28-05-19 16:49, Yang Xiao wrote:
> Based on your explanation, file->private_data can not be NULL before
> call vbg_core_close_session method all the time, since
> file->private_data is always set.
> Am I right?
Yes.
Regards,
Hans
>
> On Tue, May 28, 2019 at 9:19 PM Hans de Goede <hdegoede@redhat.com> wrote:
>>
>> Hi,
>>
>> On 28-05-19 14:47, Young Xiao wrote:
>>> vbg_misc_device_close doesn't check that filp->private_data is non-NULL
>>> before trying to close_session, where vbg_core_close_session uses pointer
>>> session whithout checking, too. That can cause an oops in certain error
>>> conditions that can occur on open or lookup before the private_data is set.
>>>
>>> This vulnerability is similar to CVE-2011-1771.
>>
>> How is this in anyway related to CVE-2011-1771 ????
>>
>> That CVE is about a filesystems lookup method including a direct open
>> of the file for performance reasons and if that direct open fails, doing
>> a fput, which is how it ends up with a file with private_date being NULL,
>> see:
>>
>> https://bugzilla.redhat.com/show_bug.cgi?id=703016
>>
>> "the problem is that CIFS doesn't do O_DIRECT at all, so when you try to open a file with it you get back -EINVAL. CIFS can also do open on lookup in some cases. In that case, fput will be called on the filp, which has not yet had its private_data set."
>>
>> The vboxguest code on the other hand is not a filessystem and as such
>> does not have a dentry lookup method at all!
>>
>> It's close method is part of the file_operations struct for a character
>> device using the misc_device framework which guarantees that private_data
>> is always set.
>>
>> Regards,
>>
>> Hans
>>
>>
>>
>>
>>
>>>
>>> Signed-off-by: Young Xiao <92siuyang@gmail.com>
>>> ---
>>> drivers/virt/vboxguest/vboxguest_linux.c | 6 ++++--
>>> 1 file changed, 4 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/virt/vboxguest/vboxguest_linux.c b/drivers/virt/vboxguest/vboxguest_linux.c
>>> index 6e8c0f1..b03c16f 100644
>>> --- a/drivers/virt/vboxguest/vboxguest_linux.c
>>> +++ b/drivers/virt/vboxguest/vboxguest_linux.c
>>> @@ -88,8 +88,10 @@ static int vbg_misc_device_user_open(struct inode *inode, struct file *filp)
>>> */
>>> static int vbg_misc_device_close(struct inode *inode, struct file *filp)
>>> {
>>> - vbg_core_close_session(filp->private_data);
>>> - filp->private_data = NULL;
>>> + if (file->private_data != NULL) {
>>> + vbg_core_close_session(filp->private_data);
>>> + filp->private_data = NULL;
>>> + }
>>> return 0;
>>> }
>>>
>>>
>
>
>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2019-05-28 14:51 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-05-28 12:47 [PATCH] vboxguest: check for private_data before trying to close it Young Xiao
2019-05-28 13:19 ` Hans de Goede
2019-05-28 14:49 ` Yang Xiao
2019-05-28 14:51 ` Hans de Goede
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).