On 4/14/25 1:58 PM, Uday Shankar wrote:
> +static int ublk_fetch(struct io_uring_cmd *cmd, struct ublk_device *ub,
> + struct ublk_queue *ubq, struct ublk_io *io,
> + const struct ublksrv_io_cmd *ub_cmd,
> + unsigned int issue_flags)
> +{
> + int ret = 0;
> +
> + if (issue_flags & IO_URING_F_NONBLOCK)
> + return -EAGAIN;
> +
> + mutex_lock(&ub->mutex);
This looks like overkill, if we can trylock the mutex that should surely
be fine? And I would imagine succeed most of the time, hence making the
inline/fastpath fine with F_NONBLOCK?
--
Jens Axboe
