Re: [Linux-kernel-mentees] [PATCH net] rds: Prevent kernel-infoleak in rds_notify_queue_get()

[HÃkon Bugge]

> On 31 Jul 2020, at 13:59, Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> 
> On Fri, Jul 31, 2020 at 01:14:09PM +0200, HÃkon Bugge wrote:
>> 
>> 
>>> On 31 Jul 2020, at 11:59, Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
>>> 
>>> On Fri, Jul 31, 2020 at 07:53:01AM +0300, Leon Romanovsky wrote:
>>>> On Thu, Jul 30, 2020 at 03:20:26PM -0400, Peilin Ye wrote:
>>>>> rds_notify_queue_get() is potentially copying uninitialized kernel stack
>>>>> memory to userspace since the compiler may leave a 4-byte hole at the end
>>>>> of `cmsg`.
>>>>> 
>>>>> In 2016 we tried to fix this issue by doing `= { 0 };` on `cmsg`, which
>>>>> unfortunately does not always initialize that 4-byte hole. Fix it by using
>>>>> memset() instead.
>>>> 
>>>> Of course, this is the difference between "{ 0 }" and "{}" initializations.
>>>> 
>>> 
>>> No, there is no difference.  Even struct assignments like:
>>> 
>>> 	foo = *bar;
>>> 
>>> can leave struct holes uninitialized.  Depending on the compiler the
>>> assignment can be implemented as a memset() or as a series of struct
>>> member assignments.
>> 
>> What about:
>> 
>> struct rds_rdma_notify {
>> 	__u64                      user_token;
>> 	__s32                      status;
>> } __attribute__((packed));
> 
> Why is this still a discussion at all?
> 
> Try it and see, run pahole and see if there are holes in this structure
> (odds are no), you don't need us to say what is happening here...

An older posting had this:

$ pahole -C "rds_rdma_notify" net/rds/recv.o
struct rds_rdma_notify {
	__u64                      user_token;           /*     0     8 */
	__s32                      status;               /*     8     4 */

	/* size: 16, cachelines: 1, members: 2 */
	/* padding: 4 */
	/* last cacheline: 16 bytes */
};


Thxs, HÃkon