On 30/01/2020 02:31, Jens Axboe wrote:
> On 1/29/20 4:19 PM, Pavel Begunkov wrote:
>> On 30/01/2020 01:44, Jens Axboe wrote:
>>> On 1/29/20 3:37 PM, Pavel Begunkov wrote:
>>>> FYI, for-next
>>>>
>>>> fs/io_uring.c: In function ‘io_epoll_ctl’:
>>>> fs/io_uring.c:2661:22: error: ‘IO_WQ_WORK_NEEDS_FILES’ undeclared (first use in
>>>> this function)
>>>>  2661 |   req->work.flags |= IO_WQ_WORK_NEEDS_FILES;
>>>>       |                      ^~~~~~~~~~~~~~~~~~~~~~
>>>> fs/io_uring.c:2661:22: note: each undeclared identifier is reported only once
>>>> for each function it appears in
>>>> make[1]: *** [scripts/Makefile.build:266: fs/io_uring.o] Error 1
>>>> make: *** [Makefile:1693: fs] Error 2
>>>
>>> Oops thanks, forgot that the epoll bits aren't in the 5.6 main branch
>>> yet, but they are in for-next. I'll fix it up there, thanks.
>>>
>>
>> Great. Also, it seems revert of ("io_uring: only allow submit from owning task
>> ") didn't get into for-next nor for-5.6/io_uring-vfs.
> 
> That's on purpose, didn't want to fold that in since it's already in
> master. Once this goes out to Linus (tomorrow/Friday), then it'll
> be resolved there.

I see, thanks

> 
> For local testing, I've been reverting it manually.
> 

-- 
Pavel Begunkov