From mboxrd@z Thu Jan  1 00:00:00 1970
From: Alexey Kodanev <alexey.kodanev@oracle.com>
Date: Tue, 15 Mar 2016 14:29:56 +0300
Subject: [LTP] [PATCH] device-drivers/uaccess: add a workaround for
 Intel's SMAP feature
In-Reply-To: <56E7EF8C.3010702@oracle.com>
References: <1458033917-13891-1-git-send-email-wangxg.fnst@cn.fujitsu.com>
 <56E7D636.9040804@cn.fujitsu.com> <56E7EF8C.3010702@oracle.com>
Message-ID: <56E7F234.60709@oracle.com>
List-Id: <ltp.lists.linux.it>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: ltp@lists.linux.it


On 03/15/2016 02:18 PM, Alexey Kodanev wrote:
> Hi,
> On 03/15/2016 12:30 PM, Xiaoguang Wang wrote:
>> hello,
>>
>> Sorry, I do not have a machine with this SMAP feature, anybody can
>> help to test this patch, thanks very much.
>>
>> Regards,
>> Xiaoguang Wang
>>
>> On 03/15/2016 05:25 PM, Xiaoguang Wang wrote:
>>> When Intel's new "Supervisor Mode Access Prevention" (SMAP) feature is
>>> enabled, kernel will be forbidden to visit user space memory, please
>>> see url: https://lwn.net/Articles/517475/ for detailed infomation.
>>>
>>> Here, we explicitly call stac()/clac() to enable/disable SMAP feature.
>>>
>>> Signed-off-by: Xiaoguang Wang <wangxg.fnst@cn.fujitsu.com>
>>> ---
>>> testcases/kernel/device-drivers/uaccess/ltp_uaccess.c | 8 ++++++++
>>>   1 file changed, 8 insertions(+)
>>>
>>> diff --git a/testcases/kernel/device-drivers/uaccess/ltp_uaccess.c 
>>> b/testcases/kernel/device-drivers/uaccess/ltp_uaccess.c
>>> index a9a06e9..35a9aa0 100644
>>> --- a/testcases/kernel/device-drivers/uaccess/ltp_uaccess.c
>>> +++ b/testcases/kernel/device-drivers/uaccess/ltp_uaccess.c
>>> @@ -68,6 +68,9 @@ static ssize_t sys_tcase(struct device *dev,
>>>       str = (char *) ptr;
>>>       test_result = 0;
>>>   +#ifdef CONFIG_X86_SMAP
>>> +    stac();
>>> +#endif

Also you don't need ifdefs as smap.h has a check:

...
#else /* CONFIG_X86_SMAP */

static inline void clac(void) { }
static inline void stac(void) { }
...

>>>
>>>       switch (tc) {
>>>       case TC_READ_USER:
>>>           if (copy_from_user(buffer, str, str_size))
>>> @@ -90,6 +93,11 @@ static ssize_t sys_tcase(struct device *dev,
>>>       break;
>>>       }
>>>   +#ifdef CONFIG_X86_SMAP
>>> +    clac();
>>> +#endif
>
>
> copy_from_user() and others have stac() / clac() already. Are you sure we
> need to invoke them there again?
>
> Thanks,
> Alexey
>
>