On 07/11/2020 13:20, Pavel Begunkov wrote:
> We don't even allow not plain data msg_control, which is disallowed in
> __sys_{send,revb}msg_sock(). So no need in fs for IORING_OP_SENDMSG and
> IORING_OP_RECVMSG. fs->lock is less contanged not as much as before, but
> there are cases that can be, e.g. IOSQE_ASYNC.
This one is still good to go. If anyone needs fs, etc. for msg_control,
IMHO it should be done in a different way not penalising others.
i.e. grabbing it in io_prep_async_work().
>
> Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
> ---
> fs/io_uring.c | 6 ++----
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index 2e435b336927..8d721a652d61 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -849,8 +849,7 @@ static const struct io_op_def io_op_defs[] = {
> .pollout = 1,
> .needs_async_data = 1,
> .async_size = sizeof(struct io_async_msghdr),
> - .work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG |
> - IO_WQ_WORK_FS,
> + .work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG,
> },
> [IORING_OP_RECVMSG] = {
> .needs_file = 1,
> @@ -859,8 +858,7 @@ static const struct io_op_def io_op_defs[] = {
> .buffer_select = 1,
> .needs_async_data = 1,
> .async_size = sizeof(struct io_async_msghdr),
> - .work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG |
> - IO_WQ_WORK_FS,
> + .work_flags = IO_WQ_WORK_MM | IO_WQ_WORK_BLKCG,
> },
> [IORING_OP_TIMEOUT] = {
> .needs_async_data = 1,
>
--
Pavel Begunkov
