Re: [PATCH 5.19 2/3] Revert "io_uring: add buffer selection support to IORING_OP_NOP"

[Jens Axboe <axboe@kernel.dk>]

On 6/14/22 12:21 PM, Dylan Yudaken wrote:
> On Tue, 2022-06-14 at 17:51 +0100, Pavel Begunkov wrote:
>> This reverts commit 3d200242a6c968af321913b635fc4014b238cba4.
>>
>> Buffer selection with nops was used for debugging and benchmarking
>> but
>> is useless in real life. Let's revert it before it's released.
>>
>> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
>> ---
>>  fs/io_uring.c | 15 +--------------
>>  1 file changed, 1 insertion(+), 14 deletions(-)
>>
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index bf556f77d4ab..1b95c6750a81 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -1114,7 +1114,6 @@ static const struct io_op_def io_op_defs[] = {
>>         [IORING_OP_NOP] = {
>>                 .audit_skip             = 1,
>>                 .iopoll                 = 1,
>> -               .buffer_select          = 1,
>>         },
>>         [IORING_OP_READV] = {
>>                 .needs_file             = 1,
>> @@ -5269,19 +5268,7 @@ static int io_nop_prep(struct io_kiocb *req,
>> const struct io_uring_sqe *sqe)
>>   */
>>  static int io_nop(struct io_kiocb *req, unsigned int issue_flags)
>>  {
>> -       unsigned int cflags;
>> -       void __user *buf;
>> -
>> -       if (req->flags & REQ_F_BUFFER_SELECT) {
>> -               size_t len = 1;
>> -
>> -               buf = io_buffer_select(req, &len, issue_flags);
>> -               if (!buf)
>> -                       return -ENOBUFS;
>> -       }
>> -
>> -       cflags = io_put_kbuf(req, issue_flags);
>> -       __io_req_complete(req, issue_flags, 0, cflags);
>> +       __io_req_complete(req, issue_flags, 0, 0);
>>         return 0;
>>  }
>>  
> 
> The liburing test case I added in "buf-ring: add tests that cycle
> through the provided buffer ring" relies on this.

Good point.

> I don't mind either way if this is kept or that liburing patch is
> reverted, but it should be consistent. What do you think?

It was useful for benchmarking as well, but it'd be a trivial patch to
do for targeted testing.

I'm fine with killing it, but can also be persuaded not to ;-)

-- 
Jens Axboe