Re: [PATCH] xfs: correct statx's result_mask value

["Su Yanjun <suyj.fnst@cn.fujitsu.com>" <suyj.fnst@cn.fujitsu.com>]

On 1/8/2019 1:07 PM, Darrick J. Wong wrote:
> On Tue, Jan 08, 2019 at 12:58:43PM +0800, Su Yanjun <suyj.fnst@cn.fujitsu.com> wrote:
>>
>> On 1/8/2019 2:04 AM, Eric Sandeen wrote:
>>> On 1/7/19 11:52 AM, Darrick J. Wong wrote:
>>>> On Mon, Jan 07, 2019 at 04:53:10AM -0500, Su Yanjun wrote:
>>>>> For statx syscall, xfs return the wrong result_mask.
>>>>>
>>>>> Signed-off-by: Su Yanjun<suyj.fnst@cn.fujitsu.com>
>>>>> ---
>>>>>    fs/xfs/xfs_iops.c | 3 +++
>>>>>    1 file changed, 3 insertions(+)
>>>>>
>>>>> diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
>>>>> index f48ffd7..3811457 100644
>>>>> --- a/fs/xfs/xfs_iops.c
>>>>> +++ b/fs/xfs/xfs_iops.c
>>>>> @@ -521,6 +521,9 @@ xfs_vn_getattr(
>>>>>    			stat->btime.tv_nsec = ip->i_d.di_crtime.t_nsec;
>>>>>    		}
>>>>>    	}
>>>>> +	
>>>>> +	/* Only return mask that we care */
>>>>> +	stat->result_mask &= request_mask;
>>>> Why not just:
>>>>
>>>> 	stat->result_mask = STATX_BASIC_STATS;
>>>>
>>>> at the top of the function?
>>>>
>>>> I don't see the need to mask off result_mask at all, since we could some
>>>> day elect to return more than what's in request_mask...
>> When we run xfstests with nfs, the generic/423 case runs failed. So i review
>> the nfs'
>> nfs_getattr code it does validate the request_mask.
>>
>> Then i review the xfs' getattr code, it has no such check. Whatever
>> request_mask
>>   is set, the stat's result_mask always the 0x7ff.
> Yes, statx can return more data than what userspace callers ask for:
>
>> Maybe it has Unclear semantics about statx's result_mask.
> "A filesystem may also fill in fields that the caller didn't ask for if
> it has values for them available and the information is available at no
> extra cost.  If this happens, the  corresponding  bits will be set in
> stx_mask."
>
> --D

I get it, then the testcase generic/423 may need update in xfstests.
Thanks for your reply.

>>>> ...waitaminute, are you seeing garbage in the result_mask that's
>>>> returned to userspace?  I also noticed the vfs stat functions declare
>>>> "struct kstat stat;" without explicitly zeroing the structure fields,
>>>> which means (I think) that we can leak stack information if the kernel
>>>> isn't built with the stackleak plugin?
>> No such problem.
>>> A clear problem statement and reproducer steps would be hugely useful
>>> here.
>>>
>>> -Eric
>> Thanks,
>> Su
>>
>>
>