Re: [PATCH 1/2] splice: export do_tee()

From: Jens Axboe
Date: Mon May 04 2020 - 12:39:07 EST


On 5/4/20 10:36 AM, Pavel Begunkov wrote:
> On 04/05/2020 17:03, Pavel Begunkov wrote:
>> On 04/05/2020 16:43, Jens Axboe wrote:
>>> On 5/4/20 6:31 AM, Pavel Begunkov wrote:
>>>> On 04/05/2020 14:09, Jann Horn wrote:
>>>>> On Sat, May 2, 2020 at 2:10 PM Pavel Begunkov <asml.silence@xxxxxxxxx> wrote:
>>>>>> export do_tee() for use in io_uring
>>>>> [...]
>>>>>> diff --git a/fs/splice.c b/fs/splice.c
>>>>> [...]
>>>>>> * The 'flags' used are the SPLICE_F_* variants, currently the only
>>>>>> * applicable one is SPLICE_F_NONBLOCK.
>>>>>> */
>>>>>> -static long do_tee(struct file *in, struct file *out, size_t len,
>>>>>> - unsigned int flags)
>>>>>> +long do_tee(struct file *in, struct file *out, size_t len, unsigned int flags)
>>>>>> {
>>>>>> struct pipe_inode_info *ipipe = get_pipe_info(in);
>>>>>> struct pipe_inode_info *opipe = get_pipe_info(out);
>>>>>
>>>>> AFAICS do_tee() in its current form is not something you should be
>>>>> making available to anything else, because the file mode checks are
>>>>> performed in sys_tee() instead of in do_tee(). (And I don't see any
>>>>> check for file modes in your uring patch, but maybe I missed it?) If
>>>>> you want to make do_tee() available elsewhere, please refactor the
>>>>> file mode checks over into do_tee().
>>>>
>>>> Overlooked it indeed. Glad you found it
>>>
>>> Yeah indeed, that's a glaring oversight on my part too. Will you send
>>> a patch for 5.7-rc as well for splice?
>>
>> Absolutely
>
> The right way would be to do as Jann proposed, but would you prefer an
> io_uring.c local fix for-5.7 and then a proper one? I assume it could
> be easier to manage.

Let's just do a proper one for 5.7 as well.

--
Jens Axboe